Language resource management -- Semantic annotation framework (SemAF)

The aim of ISO 24617-4:2014 is to propose a consensual annotation scheme for semantic roles; that is to say, a scheme that indicates the role that a participant plays in an event or state, as described mostly by a verb, and typically providing answers to questions such as "?who' did ?what' to ?whom'", and ?when', ?where', ?why', and ?how'. This includes not only the semantic relations between a verb and its arguments but also those relations that are relevant for other predicative elements such as nominalizations, nouns, adjectives, and predicate modifiers; the predicating role of adverbs and the use of coercion fall outside the scope of ISO 24617-4:2014.

Gestion de ressources linguistiques -- Cadre d'annotation sémantique (SemAF)

Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 4. del: Semantične vloge (SemAF-SR)

Cilj tega dela standarda ISO 4617 je predlagati sporazumno shemo označevanja za semantične vloge; tj. shemo, ki označuje vlogo udeleženca v dogodku ali stanju, kot jo večinoma opisuje glagol, in običajno zagotovi odgovore na vprašanja, kot so »’kdo’ je storil ‘kaj’ ‘komu’« ter ‘kdaj’, ‘kje’, ‘zakaj’ in ‘kako’. To zajema ne samo semantične odnose med glagolom in njegovimi argumenti, ampak tudi te odnose, ki so pomembni za druge predikativne elemente, kot so nominalizacije, samostalniki, pridevniki in predikatni modifikatorji; predikativna vloga prislovov in uporaba prisile so zunaj področja uporabe tega dela standarda ISO 24617.
OPOMBA: V jezikoslovju pride do prisile, kadar uporabnik jezika zaradi slovničnega konteksta na novo interpretira celotno ali delno semantiko in/ali oblikovne značilnosti leksema, ki se pojavi v tem kontekstu.

General Information

Status
Published
Publication Date
23-Jul-2014
Current Stage
6060 - International Standard published
Start Date
01-Jul-2014
Completion Date
24-Jul-2014

Buy Standard

Standard
ISO 24617-4:2018
English language
50 pages
sale 10% off
Preview
sale 10% off
Preview

e-Library read for
1 day
Standard
ISO 24617-4:2014 - Language resource management -- Semantic annotation framework (SemAF)
English language
45 pages
sale 15% off
Preview
sale 15% off
Preview
Standard
ISO 24617-4:2018
English language
50 pages
sale 10% off
Preview
sale 10% off
Preview

e-Library read for
1 day
Standard
ISO 24617-4:2018
English language
50 pages
sale 10% off
Preview
sale 10% off
Preview

e-Library read for
1 day

Standards Content (sample)

SLOVENSKI STANDARD
SIST ISO 24617-4:2018
01-september-2018
Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 4.
del: Semantične vloge (SemAF-SR)
Language resource management -- Semantic annotation framework (SemAF) -- Part 4:
Semantic roles (SemAF-SR)

Gestion de ressources linguistiques -- Cadre d'annotation sémantique (SemAF) -- Partie

4: Rôles sémantiques (SemAF-SR)
Ta slovenski standard je istoveten z: ISO 24617-4:2014
ICS:
01.020 Terminologija (načela in Terminology (principles and
koordinacija) coordination)
01.140.20 Informacijske vede Information sciences
35.240.30 Uporabniške rešitve IT v IT applications in information,
informatiki, dokumentiranju in documentation and
založništvu publishing
SIST ISO 24617-4:2018 en,fr,de

2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

---------------------- Page: 1 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 2 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL ISO
STANDARD 24617-4
First edition
2014-08-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 4:
Semantic roles (SemAF-SR)
Gestion de ressources linguistiques — Cadre d’annotation sémantique
(SemAF) —
Partie 4: Rôles sémantiques (SemAF-SR)
Reference number
ISO 24617-4:2014(E)
ISO 2014
---------------------- Page: 3 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2014

All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form

or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior

written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of

the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2014 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Contents Page

Foreword ........................................................................................................................................................................................................................................iv

Introduction ..................................................................................................................................................................................................................................v

1 Scope ................................................................................................................................................................................................................................. 1

2 Terms and definitions ..................................................................................................................................................................................... 1

2.1 Formal semantic units ...................................................................................................................................................................... 1

2.2 Conceptual semantic units and relations ......................................................................................................................... 2

2.3 General linguistic units .................................................................................................................................................................... 3

3 Abbreviated terms .............................................................................................................................................................................................. 4

4 Purpose and justification ............................................................................................................................................................................ 4

5 The nature of semantic roles ................................................................................................................................................................... 4

5.1 General ........................................................................................................................................................................................................... 4

5.2 Typical examples of semantic roles ...................................................................................................................................... 6

6 Metamodel .................................................................................................................................................................................................................. 7

6.1 Key concepts ............................................................................................................................................................................................. 7

6.2 Introduction to a metamodel for semantic role annotation .........................................................................11

6.3 Criteria for distinguishing semantic roles ....................................................................................................................12

6.4 Defining eventuality frames......................................................................................................................................................13

7 Interactions .............................................................................................................................................................................................................14

7.1 Semantic roles and semantic types ....................................................................................................................................14

7.2 Complexities ...........................................................................................................................................................................................14

7.3 The role of context in semantic role assignment ....................................................................................................14

7.4 Fuzzy boundaries between roles ..........................................................................................................................................15

7.5 Multiple classification ....................................................................................................................................................................15

7.6 Inheritance relations between semantic roles .........................................................................................................15

8 Guidelines for developing new semantic role frameworks for languages and/

or domains ...............................................................................................................................................................................................................16

8.1 General ........................................................................................................................................................................................................16

8.2 Mapping from VerbNet to LIRICS .........................................................................................................................................17

Annex A (informative) Specification of ISO semantic roles ........................................................................................................19

Annex B (informative) Review of existing frameworks ..................................................................................................................28

Annex C (informative) Specification of the annotation language ........................................................................................37

Bibliography .............................................................................................................................................................................................................................41

© ISO 2014 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Foreword

ISO (the International Organization for Standardization) is a worldwide federation of national standards

bodies (ISO member bodies). The work of preparing International Standards is normally carried out

through ISO technical committees. Each member body interested in a subject for which a technical

committee has been established has the right to be represented on that committee. International

organizations, governmental and non-governmental, in liaison with ISO, also take part in the work.

ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of

electrotechnical standardization.

The procedures used to develop this document and those intended for its further maintenance are

described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the

different types of ISO documents should be noted. This document was drafted in accordance with the

editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).

Attention is drawn to the possibility that some of the elements of this document may be the subject of

patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of

any patent rights identified during the development of the document will be in the Introduction and/or

on the ISO list of patent declarations received (see www.iso.org/patents).

Any trade name used in this document is information given for the convenience of users and does not

constitute an endorsement.

For an explanation on the meaning of ISO specific terms and expressions related to conformity

assessment, as well as information about ISO’s adherence to the WTO principles in the Technical Barriers

to Trade (TBT) see the following URL: Foreword - Supplementary information

The committee responsible for this document is ISO/TC 37, Terminology and other language and content

resources, Subcommittee SC 4, Language resource management.

ISO 24617 consists of the following parts, under the general title Language resource management —

Semantic annotation framework (SemAF):
— Part 1: Time and events (SemAF-Time, ISO-TimeML)
— Part 2: Dialogue acts
— Part 4: Semantic roles (SemAF-SR)
— Part 5: Discourse structure (SemAF-DS)
— Part 7: Spatial information (ISO-Space)
The following parts are under preparation:
— Part 8: Semantic relations in discourse (SemAF-DRel)

Principles of semantic annotation (SemAF-Basics) will form the subject of future Part 6.

iv © ISO 2014 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Introduction

This part of ISO 24617 aims to specify criteria for defining semantic roles (SRs), and is the outcome of

an agreement that the various semantic role frameworks being used to support data annotation (e.g.

FrameNet, VerbNet, PropBank, EngVallex, and LIRICS, to name only a few examples for English) have

strong underlying compatibilities. The goal is to provide both an explanation of these compatibilities and

a loose mapping between definitions of individual semantic roles, as listed in the different frameworks,

that will benefit the community as a whole.

The current specification has been developed under the aegis of the ISO Semantic Annotation Framework

(SemAF), where it is known as SemAF-SR.
The main parts of ISO 24617-4 consist of the following:
— Scope;
— Normative references;
— Terms and definitions;
— motivation and requirements;
— basic concepts and metamodel specifications;
— examples of mapping existing frameworks to the metamodel.

This part of ISO 24617 contains three informative annexes. In Annex A, the ISO semantic roles are

specified. In Annex B, information is provided both on past and current activities in semantic role

annotation and on tools and frame files. Annex C contains the abstract and concrete syntax for the

metamodel.
© ISO 2014 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 8 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL STANDARD ISO 24617-4:2014(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 4:
Semantic roles (SemAF-SR)
1 Scope

The aim of this part of ISO 24617 is to propose a consensual annotation scheme for semantic roles; that

is to say, a scheme that indicates the role that a participant plays in an event or state, as described mostly

by a verb, and typically providing answers to questions such as “‘who’ did ‘what’ to ‘whom’”, and ‘when’,

‘where’, ‘why’, and ‘how’. This includes not only the semantic relations between a verb and its arguments

but also those relations that are relevant for other predicative elements such as nominalizations, nouns,

adjectives, and predicate modifiers; the predicating role of adverbs and the use of coercion fall outside

the scope of this part of ISO 24617.

NOTE In linguistics, coercion occurs when the grammatical context causes the language-user to reinterpret

[60]

all or parts of the semantic and/or formal features of a lexeme that appear in that context.

2 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
2.1 Formal semantic units
2.1.1
argument

formal semantic unit that is an essential element of a predicate argument structure (2.1.3) and can have

variable instantiations depending on the utterance

Note 1 to entry: An argument corresponds to a participant (2.2.5) of an eventuality (2.2.2) described by the

predicate argument structure (2.1.3).

Note 2 to entry: Arguments typically satisfy certain argument positions and can be described as being syntactico-

semantic notions, whereas participants (2.2.5) are semantico-conceptual. The standard view is that subsets of the

participants associated with an eventuality (2.2.2) are selected as arguments by the verb (or nominal or adjective)

expressing the eventuality (2.2.2). Other participants (2.2.5) are either incorporated or realized as eventuality

modifiers (2.2.4).

Note 3 to entry: Natural language predicates typically have one, two, or three arguments, although they can have

more.
2.1.2
predicate

formal semantic unit that represents a semantic relation between one or more arguments (2.1.1) in a

predicate argument structure (2.1.3)

Note 1 to entry: Predicates are indicated by predicative linguistic elements such as verbs, nouns, and adjectives.

© ISO 2014 – All rights reserved 1
---------------------- Page: 9 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.1.3
predicate argument structure

formal representation of the core semantic content of an utterance, consisting of a predicate (2.1.2)

constant, and its arguments (2.1.1)

Note 1 to entry: In classical logic-based semantics, this corresponds to predicate argument structures in first-

order predicate logic.

Note 2 to entry: One of the arguments (2.1.1) can be a variable uniquely identifying the instance of the predicate

argument structure to allow references to it in other predicate argument structures.

Note 3 to entry: The representation of event semantics is subject to many variations; some of them, such as

in Reference [41], can have separate predicates (2.1.2) for each semantic role (2.2.6) relation. In this case, the

predicate argument structure of an utterance is the sum of the individual predicate semantic role (2.2.6) assertions

representing the semantic content of the utterance.
2.2 Conceptual semantic units and relations
2.2.1
entity
conceptual semantic unit that typically functions as a participant (2.2.5)

Note 1 to entry: An entity is an individual such as a person, organization, physical object, or logical entity, as

well as, on occasion, a number, quantity, dimension, or a reification of an eventuality, a property, or a quality, e.g.

emotion (anger, love), the value of a colour, etc.
2.2.2
eventuality

event, state, process, or action which can have participants (2.2.5) and which is being referred to by a

verbal, adjectival, or nominal description in an utterance

Note 1 to entry: The formal representation of an eventuality is a predicate argument structure (2.1.3).

Note 2 to entry: See ISO 24617-1. An eventuality can also be described as ‘something that can be said to obtain or

hold true, to happen or to occur’, as in ISO 24617-1. As such, they can be actual, hypothetical, or generic, covering

situations such as “You should go home,” or “He might be John’s brother.”
2.2.3
eventuality frame

generalized abstract specification of the word sense (2.3.6) associated with an eventuality (2.2.2) in an

utterance

Note 1 to entry: The frame consists of the specification of (a) a predicate (2.1.2) that can participate in a class

hierarchy if such a hierarchy is specified, and (b) the arguments (2.1.1) that this predicate (2.1.2) expects along

with their semantic roles (2.2.6).
2.2.4
eventuality modifier

particular type of participant (2.2.5) that completes the description of an eventuality (2.2.2) but is

optional and not essential

Note 1 to entry: Eventuality modifiers are distinct from other types of participants in that they are used in

supplying information that is typically more peripheral and more general, for example, situating the eventuality

in time or space.

Note 2 to entry: In FrameNet, these would be peripheral frame elements and in PropBank, ArgM’s.

Note 3 to entry: Eventuality modifiers typically correspond to syntactic adjuncts.

2 © ISO 2014 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.2.5
participant

conceptual semantic unit referred to by one or more lexical items in an utterance, which is or can be

involved in an eventuality (2.2.2)

Note 1 to entry: Both entities (2.2.1) and eventualities (2.2.2) can function as participants.

2.2.6
semantic role
mode of involvement of a participant (2.2.5) in an eventuality (2.2.2)

Note 1 to entry: Semantic roles for specific eventualities are often associated with prototypical semantic relations,

e.g. if John causes a breaking event, he is the Agent; if he uses a hammer, it is the Instrument; and someone who

receives something is a Recipient (see Clause 5 for descriptions).
2.3 General linguistic units
2.3.1
lemma
lemmatized form
conventional form chosen to represent a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.2
lexeme

fundamental unit, generally associated to a set of word forms sharing a common meaning

Note 1 to entry: See ISO 24611.
2.3.3
lexical entry

container for managing a set of word forms and possibly one or several meanings [word senses (2.3.6)]

to describe a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.4
lexicon
resource comprising a collection of lexical entries (2.3.3) for a language
Note 1 to entry: See ISO 24611.
2.3.5
utterance

stretch of speech about which no assumptions have been made in terms of linguistic theory

Note 1 to entry: See Reference [12].
2.3.6
word sense
meaning associated with a lexeme (2.3.2) in a context

Note 1 to entry: The ‘river bank’ sense of bank and the ‘financial institution’ sense of bank are considered to be

two different word senses, or lexical units, with the same word form, or lexeme (2.3.2). I called him on the radio

and Call me a taxi are associated to different word senses of the lexeme (2.3.2)call. Unrelated senses, as in bank,

are called homonyms. Senses of the same word form or lexeme which are clearly related (and can be difficult to

distinguish) are called polysemes, e.g. Coins with an image of the king, preoccupied with body image, evokes a strong

mental image.
© ISO 2014 – All rights reserved 3
---------------------- Page: 11 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
3 Abbreviated terms
EngVallex English Valency Lexicon
LIRICS Linguistic Infrastructure for Interoperable Resources and Systems
PropBank Proposition Bank
SR semantic roles
SRL semantic role labelling
WSD word sense disambiguation
4 Purpose and justification

Semantic roles are arousing increasing interest in the information-processing community because they

make explicit the key conceptual relations of participation between a verb and its arguments; that is to

say, they specify ‘who’ did ‘what’ to ‘whom’, and ‘when’, ‘where’, ‘why’, and ‘how’. For English alone, there

are already several different semantic role frameworks, including FrameNet, VerbNet, LIRICS, EngVallex,

and PropBank. Although these have been developed independently, there are strong underlying

compatibilities between them, and they share a central definition of what a semantic role is, and what

its span is, within an individual sentence. In addition to defining key concepts, this part of ISO 24617

aims to clarify and specify these underlying compatibilities and provide, where possible, a mapping

between similar semantic roles across different frameworks. This mapping is intended to serve as an

illustration of how different semantic role definitions can be linked to each other across frameworks,

and presupposes a specification of clearly defined criteria for distinguishing semantic roles.

The specification will be used in two different situations:
— in annotations where the semantic roles are recorded in annotated corpora;

— as a dynamic structure produced by automatic systems, a process typically known as semantic role

labelling (SRL).
The objectives of this specification are to provide

— a reference set of data categories that define a structured collection of semantic roles with an

explicit semantics,

— a pivot representation based on a framework for defining semantic roles that can facilitate

mapping between different formalisms (alternative semantic role representations/syntactic

theories/eventually different languages) and, in the future, between different languages, and

— guidelines for creating new resources that will be immediately interoperable with pre-existing

resources.
5 The nature of semantic roles
5.1 General

For computers to make effective use of information encoded in text, they must be able to detect the

eventualities that are being described and the eventuality participants. The processing of a sentence like

John threw a ball to Mary in the park should result in the identification of a throwing event involving John

as the Agent of the event, Mary as the Recipient, and the ball as the item being thrown; the location of the

throwing event, or where it took place, is the park. This description of the event specifies the conceptual

relations of participation that the referents of the noun phrases play with respect to the event. The

semantic notions being specified are the roles of the participants in an eventuality (i.e. semantic roles).

4 © ISO 2014 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)

This part of ISO 24617 establishes LIRICS (see Annex A) as a reference set of semantic roles with precise

definitions. Researchers are free to define their own sets of semantic roles, but explicit information on

how they can be mapped to the reference set will make resources more interoperable. Many resources

currently map to PropBank, VerbNet, or FrameNet. Since this part of ISO 24617 includes mappings of

these resources to LIRICS, such mappings already qualify as meeting the requirement of interoperability.

Our throw example seems fairly straightforward, but complexities quickly arise. English, for instance,

allows not only several different syntactic constituents to present the same semantic role, but also

several different semantic roles to be presented by the same syntactic constituent. For decades, a central

concern of linguists has been the elucidation of the process of mapping back and forth between the

syntactic analysis of the sentence and the conceptual structure and relations in the event described. For

example, in the following two sentences,
(1) The flame melted the wax.
(2) The wax melted.

a standard syntactic parser represents the wax as the verb’s direct object in the first sentence and its

subject in the second. There is nothing overt to indicate that it has the same conceptual relation in

both cases despite the fact that it is expressed syntactically in a different way. We can capture this by

annotating the wax as having the same semantic role (or conceptual relation) in both sentences. It would

typically be labelled the Patient, the participant undergoing a change of state. Note that both sentences

are in the active voice, not the passive voice. In The wax was melted by the flame, the passive provides

syntactic evidence that the wax is playing the same role (Patient) that it plays in example (1). Since the

particular pair of syntactic variations illustrated by melt does not occur with every transitive verb [see

example (5)], it is not easily predictable. Other transitive verbs can also occur in an intransitive form

while maintaining the same semantic role for the subject as the transitive, as in the following example,

where soprano is the Agent of sing in both sentences (the aria is the Theme):
(3) The soprano sang an aria.
(4) The soprano sang.
The verb slice can also move the Patient (the bread) to subject position, as in
(5) John sliced the bread easily./ This bread slices easily.
although other transitive verbs, such as eat, cannot:
(6) John ate the apple.
(7) John ate.
(8) *The apple ate crunchily in the background.
The last sentence is starred (*) to indicate its ungrammaticality.

Accurate interpretation of the semantic roles of the verb arguments (i.e. ‘Who did what to whom?’) is a

crucial goal for natural language processing systems. Our ability to do this automatically has improved

enormously in recent years and has been largely based on the availability of annotated corpora. In

fact, there are corpora, such as FrameNet and PropBank, available with quite different semantic role

annotations, and this prompts questions about the nature and number of semantic roles. This part of

ISO 24617 attempts to provide definitions and examples clarifying their definition.

For semantic roles to maximize the benefit to the information processing community, it is desirable that

the definitions of the semantic roles should, as far as possible, have the following properties:

— consistently recognizable;
— able to clarify sense distinctions;
— generalizable;
© ISO 2014 – All rights reserved 5
---------------------- Page: 13 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
— machine learnable;
— able to provide an appropriate foundation for inferencing.

The purpose of the specifications for semantic role definitions in this part of ISO 24617 is to provide

these attributes.
5.2 Typical examples of semantic roles

A list of the best-known roles and the properties usually associated with them is given below. They are

taken from the EAGLES discussion on Standardizing Subcategorization (see Reference [61]). Comments

in parentheses have been added for clarification purposes and include comparisons with LIRICS

(Linguistic Infrastructure for Interoperable Resources and Systems), and occasionally with VerbNet.

NOTE These role definitions, and the ones for LIRICS, are quite general and can cover a wide range of

participant types. The specific preposition associated with an individual semantic role quite often adds nuances

of meaning in addition to that conveyed by the semantic role itself. For instance, in “Eat the fish with caution,” the

phrase “with caution” would typically be labelled as MANNER. Another example of a MANNER role could be the

phrase “in three bites” from “Eat the fish in three bites.” Clearly the interpretation of these two phrases will be

quite different, as will be their impact on the representation of the eating eventuality. Semantic role labels cannot

be expected to clarify all such types of subtle differences in meaning, and additional research on the definitions

[48]

of prepositions and their interactions with semantic roles, such as Srikumar and Roth, 2013 , is needed.

Agent

A participant that the meaning of the verb specifies as doing or causing something, possibly intentionally;

for example, as the subject of kill, eat, hit, smash, kick, and watch. (LIRICS has a similar Agent, which acts

intentionally or consciously.)
Patient

A participant that the verb characterizes as having something happen to it, and as being affected by

what happens to it; for example, as the object of kill, eat, and smash, but not of watch, hear, and love. (If

someone watches television, the television is not affected by the watching, so it would be a Theme rather

than a Patient. LIRICS has a similar Patient.)
Experiencer

A participant that is characterized as aware of something; for example, as the subject of love or as the

object of annoy. (LIRICS uses an Agent for these verbs and has no Experiencer; VerbNet has the same

role, but only when the Experiencer is affected by the event.)
Theme

A participant that is characterized as changing its position or condition, or as being in a state or position;

for example, as the object of give and hand or as the subjects of walk (in line with the policy of labelling

the object in motion as a Theme) and of die. (According to the EAGLES definition, people who die of

old age would not be considered t
...

INTERNATIONAL ISO
STANDARD 24617-4
First edition
2014-08-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 4:
Semantic roles (SemAF-SR)
Gestion de ressources linguistiques — Cadre d’annotation sémantique
(SemAF) —
Partie 4: Rôles sémantiques (SemAF-SR)
Reference number
ISO 24617-4:2014(E)
ISO 2014
---------------------- Page: 1 ----------------------
ISO 24617-4:2014(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2014

All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form

or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior

written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of

the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2014 – All rights reserved
---------------------- Page: 2 ----------------------
ISO 24617-4:2014(E)
Contents Page

Foreword ........................................................................................................................................................................................................................................iv

Introduction ..................................................................................................................................................................................................................................v

1 Scope ................................................................................................................................................................................................................................. 1

2 Terms and definitions ..................................................................................................................................................................................... 1

2.1 Formal semantic units ...................................................................................................................................................................... 1

2.2 Conceptual semantic units and relations ......................................................................................................................... 2

2.3 General linguistic units .................................................................................................................................................................... 3

3 Abbreviated terms .............................................................................................................................................................................................. 4

4 Purpose and justification ............................................................................................................................................................................ 4

5 The nature of semantic roles ................................................................................................................................................................... 4

5.1 General ........................................................................................................................................................................................................... 4

5.2 Typical examples of semantic roles ...................................................................................................................................... 6

6 Metamodel .................................................................................................................................................................................................................. 7

6.1 Key concepts ............................................................................................................................................................................................. 7

6.2 Introduction to a metamodel for semantic role annotation .........................................................................11

6.3 Criteria for distinguishing semantic roles ....................................................................................................................12

6.4 Defining eventuality frames......................................................................................................................................................13

7 Interactions .............................................................................................................................................................................................................14

7.1 Semantic roles and semantic types ....................................................................................................................................14

7.2 Complexities ...........................................................................................................................................................................................14

7.3 The role of context in semantic role assignment ....................................................................................................14

7.4 Fuzzy boundaries between roles ..........................................................................................................................................15

7.5 Multiple classification ....................................................................................................................................................................15

7.6 Inheritance relations between semantic roles .........................................................................................................15

8 Guidelines for developing new semantic role frameworks for languages and/

or domains ...............................................................................................................................................................................................................16

8.1 General ........................................................................................................................................................................................................16

8.2 Mapping from VerbNet to LIRICS .........................................................................................................................................17

Annex A (informative) Specification of ISO semantic roles ........................................................................................................19

Annex B (informative) Review of existing frameworks ..................................................................................................................28

Annex C (informative) Specification of the annotation language ........................................................................................37

Bibliography .............................................................................................................................................................................................................................41

© ISO 2014 – All rights reserved iii
---------------------- Page: 3 ----------------------
ISO 24617-4:2014(E)
Foreword

ISO (the International Organization for Standardization) is a worldwide federation of national standards

bodies (ISO member bodies). The work of preparing International Standards is normally carried out

through ISO technical committees. Each member body interested in a subject for which a technical

committee has been established has the right to be represented on that committee. International

organizations, governmental and non-governmental, in liaison with ISO, also take part in the work.

ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of

electrotechnical standardization.

The procedures used to develop this document and those intended for its further maintenance are

described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the

different types of ISO documents should be noted. This document was drafted in accordance with the

editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).

Attention is drawn to the possibility that some of the elements of this document may be the subject of

patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of

any patent rights identified during the development of the document will be in the Introduction and/or

on the ISO list of patent declarations received (see www.iso.org/patents).

Any trade name used in this document is information given for the convenience of users and does not

constitute an endorsement.

For an explanation on the meaning of ISO specific terms and expressions related to conformity

assessment, as well as information about ISO’s adherence to the WTO principles in the Technical Barriers

to Trade (TBT) see the following URL: Foreword - Supplementary information

The committee responsible for this document is ISO/TC 37, Terminology and other language and content

resources, Subcommittee SC 4, Language resource management.

ISO 24617 consists of the following parts, under the general title Language resource management —

Semantic annotation framework (SemAF):
— Part 1: Time and events (SemAF-Time, ISO-TimeML)
— Part 2: Dialogue acts
— Part 4: Semantic roles (SemAF-SR)
— Part 5: Discourse structure (SemAF-DS)
— Part 7: Spatial information (ISO-Space)
The following parts are under preparation:
— Part 8: Semantic relations in discourse (SemAF-DRel)

Principles of semantic annotation (SemAF-Basics) will form the subject of future Part 6.

iv © ISO 2014 – All rights reserved
---------------------- Page: 4 ----------------------
ISO 24617-4:2014(E)
Introduction

This part of ISO 24617 aims to specify criteria for defining semantic roles (SRs), and is the outcome of

an agreement that the various semantic role frameworks being used to support data annotation (e.g.

FrameNet, VerbNet, PropBank, EngVallex, and LIRICS, to name only a few examples for English) have

strong underlying compatibilities. The goal is to provide both an explanation of these compatibilities and

a loose mapping between definitions of individual semantic roles, as listed in the different frameworks,

that will benefit the community as a whole.

The current specification has been developed under the aegis of the ISO Semantic Annotation Framework

(SemAF), where it is known as SemAF-SR.
The main parts of ISO 24617-4 consist of the following:
— Scope;
— Normative references;
— Terms and definitions;
— motivation and requirements;
— basic concepts and metamodel specifications;
— examples of mapping existing frameworks to the metamodel.

This part of ISO 24617 contains three informative annexes. In Annex A, the ISO semantic roles are

specified. In Annex B, information is provided both on past and current activities in semantic role

annotation and on tools and frame files. Annex C contains the abstract and concrete syntax for the

metamodel.
© ISO 2014 – All rights reserved v
---------------------- Page: 5 ----------------------
INTERNATIONAL STANDARD ISO 24617-4:2014(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 4:
Semantic roles (SemAF-SR)
1 Scope

The aim of this part of ISO 24617 is to propose a consensual annotation scheme for semantic roles; that

is to say, a scheme that indicates the role that a participant plays in an event or state, as described mostly

by a verb, and typically providing answers to questions such as “‘who’ did ‘what’ to ‘whom’”, and ‘when’,

‘where’, ‘why’, and ‘how’. This includes not only the semantic relations between a verb and its arguments

but also those relations that are relevant for other predicative elements such as nominalizations, nouns,

adjectives, and predicate modifiers; the predicating role of adverbs and the use of coercion fall outside

the scope of this part of ISO 24617.

NOTE In linguistics, coercion occurs when the grammatical context causes the language-user to reinterpret

[60]

all or parts of the semantic and/or formal features of a lexeme that appear in that context.

2 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
2.1 Formal semantic units
2.1.1
argument

formal semantic unit that is an essential element of a predicate argument structure (2.1.3) and can have

variable instantiations depending on the utterance

Note 1 to entry: An argument corresponds to a participant (2.2.5) of an eventuality (2.2.2) described by the

predicate argument structure (2.1.3).

Note 2 to entry: Arguments typically satisfy certain argument positions and can be described as being syntactico-

semantic notions, whereas participants (2.2.5) are semantico-conceptual. The standard view is that subsets of the

participants associated with an eventuality (2.2.2) are selected as arguments by the verb (or nominal or adjective)

expressing the eventuality (2.2.2). Other participants (2.2.5) are either incorporated or realized as eventuality

modifiers (2.2.4).

Note 3 to entry: Natural language predicates typically have one, two, or three arguments, although they can have

more.
2.1.2
predicate

formal semantic unit that represents a semantic relation between one or more arguments (2.1.1) in a

predicate argument structure (2.1.3)

Note 1 to entry: Predicates are indicated by predicative linguistic elements such as verbs, nouns, and adjectives.

© ISO 2014 – All rights reserved 1
---------------------- Page: 6 ----------------------
ISO 24617-4:2014(E)
2.1.3
predicate argument structure

formal representation of the core semantic content of an utterance, consisting of a predicate (2.1.2)

constant, and its arguments (2.1.1)

Note 1 to entry: In classical logic-based semantics, this corresponds to predicate argument structures in first-

order predicate logic.

Note 2 to entry: One of the arguments (2.1.1) can be a variable uniquely identifying the instance of the predicate

argument structure to allow references to it in other predicate argument structures.

Note 3 to entry: The representation of event semantics is subject to many variations; some of them, such as

in Reference [41], can have separate predicates (2.1.2) for each semantic role (2.2.6) relation. In this case, the

predicate argument structure of an utterance is the sum of the individual predicate semantic role (2.2.6) assertions

representing the semantic content of the utterance.
2.2 Conceptual semantic units and relations
2.2.1
entity
conceptual semantic unit that typically functions as a participant (2.2.5)

Note 1 to entry: An entity is an individual such as a person, organization, physical object, or logical entity, as

well as, on occasion, a number, quantity, dimension, or a reification of an eventuality, a property, or a quality, e.g.

emotion (anger, love), the value of a colour, etc.
2.2.2
eventuality

event, state, process, or action which can have participants (2.2.5) and which is being referred to by a

verbal, adjectival, or nominal description in an utterance

Note 1 to entry: The formal representation of an eventuality is a predicate argument structure (2.1.3).

Note 2 to entry: See ISO 24617-1. An eventuality can also be described as ‘something that can be said to obtain or

hold true, to happen or to occur’, as in ISO 24617-1. As such, they can be actual, hypothetical, or generic, covering

situations such as “You should go home,” or “He might be John’s brother.”
2.2.3
eventuality frame

generalized abstract specification of the word sense (2.3.6) associated with an eventuality (2.2.2) in an

utterance

Note 1 to entry: The frame consists of the specification of (a) a predicate (2.1.2) that can participate in a class

hierarchy if such a hierarchy is specified, and (b) the arguments (2.1.1) that this predicate (2.1.2) expects along

with their semantic roles (2.2.6).
2.2.4
eventuality modifier

particular type of participant (2.2.5) that completes the description of an eventuality (2.2.2) but is

optional and not essential

Note 1 to entry: Eventuality modifiers are distinct from other types of participants in that they are used in

supplying information that is typically more peripheral and more general, for example, situating the eventuality

in time or space.

Note 2 to entry: In FrameNet, these would be peripheral frame elements and in PropBank, ArgM’s.

Note 3 to entry: Eventuality modifiers typically correspond to syntactic adjuncts.

2 © ISO 2014 – All rights reserved
---------------------- Page: 7 ----------------------
ISO 24617-4:2014(E)
2.2.5
participant

conceptual semantic unit referred to by one or more lexical items in an utterance, which is or can be

involved in an eventuality (2.2.2)

Note 1 to entry: Both entities (2.2.1) and eventualities (2.2.2) can function as participants.

2.2.6
semantic role
mode of involvement of a participant (2.2.5) in an eventuality (2.2.2)

Note 1 to entry: Semantic roles for specific eventualities are often associated with prototypical semantic relations,

e.g. if John causes a breaking event, he is the Agent; if he uses a hammer, it is the Instrument; and someone who

receives something is a Recipient (see Clause 5 for descriptions).
2.3 General linguistic units
2.3.1
lemma
lemmatized form
conventional form chosen to represent a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.2
lexeme

fundamental unit, generally associated to a set of word forms sharing a common meaning

Note 1 to entry: See ISO 24611.
2.3.3
lexical entry

container for managing a set of word forms and possibly one or several meanings [word senses (2.3.6)]

to describe a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.4
lexicon
resource comprising a collection of lexical entries (2.3.3) for a language
Note 1 to entry: See ISO 24611.
2.3.5
utterance

stretch of speech about which no assumptions have been made in terms of linguistic theory

Note 1 to entry: See Reference [12].
2.3.6
word sense
meaning associated with a lexeme (2.3.2) in a context

Note 1 to entry: The ‘river bank’ sense of bank and the ‘financial institution’ sense of bank are considered to be

two different word senses, or lexical units, with the same word form, or lexeme (2.3.2). I called him on the radio

and Call me a taxi are associated to different word senses of the lexeme (2.3.2)call. Unrelated senses, as in bank,

are called homonyms. Senses of the same word form or lexeme which are clearly related (and can be difficult to

distinguish) are called polysemes, e.g. Coins with an image of the king, preoccupied with body image, evokes a strong

mental image.
© ISO 2014 – All rights reserved 3
---------------------- Page: 8 ----------------------
ISO 24617-4:2014(E)
3 Abbreviated terms
EngVallex English Valency Lexicon
LIRICS Linguistic Infrastructure for Interoperable Resources and Systems
PropBank Proposition Bank
SR semantic roles
SRL semantic role labelling
WSD word sense disambiguation
4 Purpose and justification

Semantic roles are arousing increasing interest in the information-processing community because they

make explicit the key conceptual relations of participation between a verb and its arguments; that is to

say, they specify ‘who’ did ‘what’ to ‘whom’, and ‘when’, ‘where’, ‘why’, and ‘how’. For English alone, there

are already several different semantic role frameworks, including FrameNet, VerbNet, LIRICS, EngVallex,

and PropBank. Although these have been developed independently, there are strong underlying

compatibilities between them, and they share a central definition of what a semantic role is, and what

its span is, within an individual sentence. In addition to defining key concepts, this part of ISO 24617

aims to clarify and specify these underlying compatibilities and provide, where possible, a mapping

between similar semantic roles across different frameworks. This mapping is intended to serve as an

illustration of how different semantic role definitions can be linked to each other across frameworks,

and presupposes a specification of clearly defined criteria for distinguishing semantic roles.

The specification will be used in two different situations:
— in annotations where the semantic roles are recorded in annotated corpora;

— as a dynamic structure produced by automatic systems, a process typically known as semantic role

labelling (SRL).
The objectives of this specification are to provide

— a reference set of data categories that define a structured collection of semantic roles with an

explicit semantics,

— a pivot representation based on a framework for defining semantic roles that can facilitate

mapping between different formalisms (alternative semantic role representations/syntactic

theories/eventually different languages) and, in the future, between different languages, and

— guidelines for creating new resources that will be immediately interoperable with pre-existing

resources.
5 The nature of semantic roles
5.1 General

For computers to make effective use of information encoded in text, they must be able to detect the

eventualities that are being described and the eventuality participants. The processing of a sentence like

John threw a ball to Mary in the park should result in the identification of a throwing event involving John

as the Agent of the event, Mary as the Recipient, and the ball as the item being thrown; the location of the

throwing event, or where it took place, is the park. This description of the event specifies the conceptual

relations of participation that the referents of the noun phrases play with respect to the event. The

semantic notions being specified are the roles of the participants in an eventuality (i.e. semantic roles).

4 © ISO 2014 – All rights reserved
---------------------- Page: 9 ----------------------
ISO 24617-4:2014(E)

This part of ISO 24617 establishes LIRICS (see Annex A) as a reference set of semantic roles with precise

definitions. Researchers are free to define their own sets of semantic roles, but explicit information on

how they can be mapped to the reference set will make resources more interoperable. Many resources

currently map to PropBank, VerbNet, or FrameNet. Since this part of ISO 24617 includes mappings of

these resources to LIRICS, such mappings already qualify as meeting the requirement of interoperability.

Our throw example seems fairly straightforward, but complexities quickly arise. English, for instance,

allows not only several different syntactic constituents to present the same semantic role, but also

several different semantic roles to be presented by the same syntactic constituent. For decades, a central

concern of linguists has been the elucidation of the process of mapping back and forth between the

syntactic analysis of the sentence and the conceptual structure and relations in the event described. For

example, in the following two sentences,
(1) The flame melted the wax.
(2) The wax melted.

a standard syntactic parser represents the wax as the verb’s direct object in the first sentence and its

subject in the second. There is nothing overt to indicate that it has the same conceptual relation in

both cases despite the fact that it is expressed syntactically in a different way. We can capture this by

annotating the wax as having the same semantic role (or conceptual relation) in both sentences. It would

typically be labelled the Patient, the participant undergoing a change of state. Note that both sentences

are in the active voice, not the passive voice. In The wax was melted by the flame, the passive provides

syntactic evidence that the wax is playing the same role (Patient) that it plays in example (1). Since the

particular pair of syntactic variations illustrated by melt does not occur with every transitive verb [see

example (5)], it is not easily predictable. Other transitive verbs can also occur in an intransitive form

while maintaining the same semantic role for the subject as the transitive, as in the following example,

where soprano is the Agent of sing in both sentences (the aria is the Theme):
(3) The soprano sang an aria.
(4) The soprano sang.
The verb slice can also move the Patient (the bread) to subject position, as in
(5) John sliced the bread easily./ This bread slices easily.
although other transitive verbs, such as eat, cannot:
(6) John ate the apple.
(7) John ate.
(8) *The apple ate crunchily in the background.
The last sentence is starred (*) to indicate its ungrammaticality.

Accurate interpretation of the semantic roles of the verb arguments (i.e. ‘Who did what to whom?’) is a

crucial goal for natural language processing systems. Our ability to do this automatically has improved

enormously in recent years and has been largely based on the availability of annotated corpora. In

fact, there are corpora, such as FrameNet and PropBank, available with quite different semantic role

annotations, and this prompts questions about the nature and number of semantic roles. This part of

ISO 24617 attempts to provide definitions and examples clarifying their definition.

For semantic roles to maximize the benefit to the information processing community, it is desirable that

the definitions of the semantic roles should, as far as possible, have the following properties:

— consistently recognizable;
— able to clarify sense distinctions;
— generalizable;
© ISO 2014 – All rights reserved 5
---------------------- Page: 10 ----------------------
ISO 24617-4:2014(E)
— machine learnable;
— able to provide an appropriate foundation for inferencing.

The purpose of the specifications for semantic role definitions in this part of ISO 24617 is to provide

these attributes.
5.2 Typical examples of semantic roles

A list of the best-known roles and the properties usually associated with them is given below. They are

taken from the EAGLES discussion on Standardizing Subcategorization (see Reference [61]). Comments

in parentheses have been added for clarification purposes and include comparisons with LIRICS

(Linguistic Infrastructure for Interoperable Resources and Systems), and occasionally with VerbNet.

NOTE These role definitions, and the ones for LIRICS, are quite general and can cover a wide range of

participant types. The specific preposition associated with an individual semantic role quite often adds nuances

of meaning in addition to that conveyed by the semantic role itself. For instance, in “Eat the fish with caution,” the

phrase “with caution” would typically be labelled as MANNER. Another example of a MANNER role could be the

phrase “in three bites” from “Eat the fish in three bites.” Clearly the interpretation of these two phrases will be

quite different, as will be their impact on the representation of the eating eventuality. Semantic role labels cannot

be expected to clarify all such types of subtle differences in meaning, and additional research on the definitions

[48]

of prepositions and their interactions with semantic roles, such as Srikumar and Roth, 2013 , is needed.

Agent

A participant that the meaning of the verb specifies as doing or causing something, possibly intentionally;

for example, as the subject of kill, eat, hit, smash, kick, and watch. (LIRICS has a similar Agent, which acts

intentionally or consciously.)
Patient

A participant that the verb characterizes as having something happen to it, and as being affected by

what happens to it; for example, as the object of kill, eat, and smash, but not of watch, hear, and love. (If

someone watches television, the television is not affected by the watching, so it would be a Theme rather

than a Patient. LIRICS has a similar Patient.)
Experiencer

A participant that is characterized as aware of something; for example, as the subject of love or as the

object of annoy. (LIRICS uses an Agent for these verbs and has no Experiencer; VerbNet has the same

role, but only when the Experiencer is affected by the event.)
Theme

A participant that is characterized as changing its position or condition, or as being in a state or position;

for example, as the object of give and hand or as the subjects of walk (in line with the policy of labelling

the object in motion as a Theme) and of die. (According to the EAGLES definition, people who die of

old age would not be considered to be Patients because an Agent has not acted upon them. EngVallex

stays more syntactically oriented and marks such participants as Actors or Patients, depending on their

syntactic position.)
Location

The thematic role associated with the Noun Phrase expressing the location in a sentence with a verb

of location [perhaps in a Prepositional Phrase (PP)]; for example, the subjects of keep, own, retain, and

know, and locative PPs. (For EAGLES, the location of the thing kept or owned is considered to be with the

keeper or the owner, and so on. LIRICS has a Location role, but uses it in a more restrictive way, and uses

Agent or Pivot for several of these verbs instead of Location, as does VerbNet.)
6 © ISO 2014 – All rights reserved
---------------------- Page: 11 ----------------------
ISO 24617-4:2014(E)
Source

Object from which motion proceeds; for example, as the subject of sell and promise or as the objects

of deprive, free, and cure. Note that the motion can be abstract. (LIRICS uses Source for these abstract

examples, but uses Initial-Location instead for physical motion verbs.)
Goal

Object to which motion proceeds (e.g. the Path-prepositional phrase for caused-motion verbs like throw,

and for metaphorical motion events), the subject of receive and buy and the dative object of tell and give.

[71]
(Adapted
...

SLOVENSKI STANDARD
SIST ISO 24617-4:2018
01-september-2018
Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 4.
del: Semantične vloge (SemAF-SR)
Language resource management -- Semantic annotation framework (SemAF) -- Part 4:
Semantic roles (SemAF-SR)

Gestion de ressources linguistiques -- Cadre d'annotation sémantique (SemAF) -- Partie

4: Rôles sémantiques (SemAF-SR)
Ta slovenski standard je istoveten z: ISO 24617-4:2014
ICS:
01.020 Terminologija (načela in Terminology (principles and
koordinacija) coordination)
35.240.30 Uporabniške rešitve IT v IT applications in information,
informatiki, dokumentiranju in documentation and
založništvu publishing
SIST ISO 24617-4:2018 en,fr,de

2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

---------------------- Page: 1 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 2 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL ISO
STANDARD 24617-4
First edition
2014-08-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 4:
Semantic roles (SemAF-SR)
Gestion de ressources linguistiques — Cadre d’annotation sémantique
(SemAF) —
Partie 4: Rôles sémantiques (SemAF-SR)
Reference number
ISO 24617-4:2014(E)
ISO 2014
---------------------- Page: 3 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2014

All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form

or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior

written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of

the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2014 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Contents Page

Foreword ........................................................................................................................................................................................................................................iv

Introduction ..................................................................................................................................................................................................................................v

1 Scope ................................................................................................................................................................................................................................. 1

2 Terms and definitions ..................................................................................................................................................................................... 1

2.1 Formal semantic units ...................................................................................................................................................................... 1

2.2 Conceptual semantic units and relations ......................................................................................................................... 2

2.3 General linguistic units .................................................................................................................................................................... 3

3 Abbreviated terms .............................................................................................................................................................................................. 4

4 Purpose and justification ............................................................................................................................................................................ 4

5 The nature of semantic roles ................................................................................................................................................................... 4

5.1 General ........................................................................................................................................................................................................... 4

5.2 Typical examples of semantic roles ...................................................................................................................................... 6

6 Metamodel .................................................................................................................................................................................................................. 7

6.1 Key concepts ............................................................................................................................................................................................. 7

6.2 Introduction to a metamodel for semantic role annotation .........................................................................11

6.3 Criteria for distinguishing semantic roles ....................................................................................................................12

6.4 Defining eventuality frames......................................................................................................................................................13

7 Interactions .............................................................................................................................................................................................................14

7.1 Semantic roles and semantic types ....................................................................................................................................14

7.2 Complexities ...........................................................................................................................................................................................14

7.3 The role of context in semantic role assignment ....................................................................................................14

7.4 Fuzzy boundaries between roles ..........................................................................................................................................15

7.5 Multiple classification ....................................................................................................................................................................15

7.6 Inheritance relations between semantic roles .........................................................................................................15

8 Guidelines for developing new semantic role frameworks for languages and/

or domains ...............................................................................................................................................................................................................16

8.1 General ........................................................................................................................................................................................................16

8.2 Mapping from VerbNet to LIRICS .........................................................................................................................................17

Annex A (informative) Specification of ISO semantic roles ........................................................................................................19

Annex B (informative) Review of existing frameworks ..................................................................................................................28

Annex C (informative) Specification of the annotation language ........................................................................................37

Bibliography .............................................................................................................................................................................................................................41

© ISO 2014 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Foreword

ISO (the International Organization for Standardization) is a worldwide federation of national standards

bodies (ISO member bodies). The work of preparing International Standards is normally carried out

through ISO technical committees. Each member body interested in a subject for which a technical

committee has been established has the right to be represented on that committee. International

organizations, governmental and non-governmental, in liaison with ISO, also take part in the work.

ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of

electrotechnical standardization.

The procedures used to develop this document and those intended for its further maintenance are

described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the

different types of ISO documents should be noted. This document was drafted in accordance with the

editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).

Attention is drawn to the possibility that some of the elements of this document may be the subject of

patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of

any patent rights identified during the development of the document will be in the Introduction and/or

on the ISO list of patent declarations received (see www.iso.org/patents).

Any trade name used in this document is information given for the convenience of users and does not

constitute an endorsement.

For an explanation on the meaning of ISO specific terms and expressions related to conformity

assessment, as well as information about ISO’s adherence to the WTO principles in the Technical Barriers

to Trade (TBT) see the following URL: Foreword - Supplementary information

The committee responsible for this document is ISO/TC 37, Terminology and other language and content

resources, Subcommittee SC 4, Language resource management.

ISO 24617 consists of the following parts, under the general title Language resource management —

Semantic annotation framework (SemAF):
— Part 1: Time and events (SemAF-Time, ISO-TimeML)
— Part 2: Dialogue acts
— Part 4: Semantic roles (SemAF-SR)
— Part 5: Discourse structure (SemAF-DS)
— Part 7: Spatial information (ISO-Space)
The following parts are under preparation:
— Part 8: Semantic relations in discourse (SemAF-DRel)

Principles of semantic annotation (SemAF-Basics) will form the subject of future Part 6.

iv © ISO 2014 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Introduction

This part of ISO 24617 aims to specify criteria for defining semantic roles (SRs), and is the outcome of

an agreement that the various semantic role frameworks being used to support data annotation (e.g.

FrameNet, VerbNet, PropBank, EngVallex, and LIRICS, to name only a few examples for English) have

strong underlying compatibilities. The goal is to provide both an explanation of these compatibilities and

a loose mapping between definitions of individual semantic roles, as listed in the different frameworks,

that will benefit the community as a whole.

The current specification has been developed under the aegis of the ISO Semantic Annotation Framework

(SemAF), where it is known as SemAF-SR.
The main parts of ISO 24617-4 consist of the following:
— Scope;
— Normative references;
— Terms and definitions;
— motivation and requirements;
— basic concepts and metamodel specifications;
— examples of mapping existing frameworks to the metamodel.

This part of ISO 24617 contains three informative annexes. In Annex A, the ISO semantic roles are

specified. In Annex B, information is provided both on past and current activities in semantic role

annotation and on tools and frame files. Annex C contains the abstract and concrete syntax for the

metamodel.
© ISO 2014 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 8 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL STANDARD ISO 24617-4:2014(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 4:
Semantic roles (SemAF-SR)
1 Scope

The aim of this part of ISO 24617 is to propose a consensual annotation scheme for semantic roles; that

is to say, a scheme that indicates the role that a participant plays in an event or state, as described mostly

by a verb, and typically providing answers to questions such as “‘who’ did ‘what’ to ‘whom’”, and ‘when’,

‘where’, ‘why’, and ‘how’. This includes not only the semantic relations between a verb and its arguments

but also those relations that are relevant for other predicative elements such as nominalizations, nouns,

adjectives, and predicate modifiers; the predicating role of adverbs and the use of coercion fall outside

the scope of this part of ISO 24617.

NOTE In linguistics, coercion occurs when the grammatical context causes the language-user to reinterpret

[60]

all or parts of the semantic and/or formal features of a lexeme that appear in that context.

2 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
2.1 Formal semantic units
2.1.1
argument

formal semantic unit that is an essential element of a predicate argument structure (2.1.3) and can have

variable instantiations depending on the utterance

Note 1 to entry: An argument corresponds to a participant (2.2.5) of an eventuality (2.2.2) described by the

predicate argument structure (2.1.3).

Note 2 to entry: Arguments typically satisfy certain argument positions and can be described as being syntactico-

semantic notions, whereas participants (2.2.5) are semantico-conceptual. The standard view is that subsets of the

participants associated with an eventuality (2.2.2) are selected as arguments by the verb (or nominal or adjective)

expressing the eventuality (2.2.2). Other participants (2.2.5) are either incorporated or realized as eventuality

modifiers (2.2.4).

Note 3 to entry: Natural language predicates typically have one, two, or three arguments, although they can have

more.
2.1.2
predicate

formal semantic unit that represents a semantic relation between one or more arguments (2.1.1) in a

predicate argument structure (2.1.3)

Note 1 to entry: Predicates are indicated by predicative linguistic elements such as verbs, nouns, and adjectives.

© ISO 2014 – All rights reserved 1
---------------------- Page: 9 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.1.3
predicate argument structure

formal representation of the core semantic content of an utterance, consisting of a predicate (2.1.2)

constant, and its arguments (2.1.1)

Note 1 to entry: In classical logic-based semantics, this corresponds to predicate argument structures in first-

order predicate logic.

Note 2 to entry: One of the arguments (2.1.1) can be a variable uniquely identifying the instance of the predicate

argument structure to allow references to it in other predicate argument structures.

Note 3 to entry: The representation of event semantics is subject to many variations; some of them, such as

in Reference [41], can have separate predicates (2.1.2) for each semantic role (2.2.6) relation. In this case, the

predicate argument structure of an utterance is the sum of the individual predicate semantic role (2.2.6) assertions

representing the semantic content of the utterance.
2.2 Conceptual semantic units and relations
2.2.1
entity
conceptual semantic unit that typically functions as a participant (2.2.5)

Note 1 to entry: An entity is an individual such as a person, organization, physical object, or logical entity, as

well as, on occasion, a number, quantity, dimension, or a reification of an eventuality, a property, or a quality, e.g.

emotion (anger, love), the value of a colour, etc.
2.2.2
eventuality

event, state, process, or action which can have participants (2.2.5) and which is being referred to by a

verbal, adjectival, or nominal description in an utterance

Note 1 to entry: The formal representation of an eventuality is a predicate argument structure (2.1.3).

Note 2 to entry: See ISO 24617-1. An eventuality can also be described as ‘something that can be said to obtain or

hold true, to happen or to occur’, as in ISO 24617-1. As such, they can be actual, hypothetical, or generic, covering

situations such as “You should go home,” or “He might be John’s brother.”
2.2.3
eventuality frame

generalized abstract specification of the word sense (2.3.6) associated with an eventuality (2.2.2) in an

utterance

Note 1 to entry: The frame consists of the specification of (a) a predicate (2.1.2) that can participate in a class

hierarchy if such a hierarchy is specified, and (b) the arguments (2.1.1) that this predicate (2.1.2) expects along

with their semantic roles (2.2.6).
2.2.4
eventuality modifier

particular type of participant (2.2.5) that completes the description of an eventuality (2.2.2) but is

optional and not essential

Note 1 to entry: Eventuality modifiers are distinct from other types of participants in that they are used in

supplying information that is typically more peripheral and more general, for example, situating the eventuality

in time or space.

Note 2 to entry: In FrameNet, these would be peripheral frame elements and in PropBank, ArgM’s.

Note 3 to entry: Eventuality modifiers typically correspond to syntactic adjuncts.

2 © ISO 2014 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.2.5
participant

conceptual semantic unit referred to by one or more lexical items in an utterance, which is or can be

involved in an eventuality (2.2.2)

Note 1 to entry: Both entities (2.2.1) and eventualities (2.2.2) can function as participants.

2.2.6
semantic role
mode of involvement of a participant (2.2.5) in an eventuality (2.2.2)

Note 1 to entry: Semantic roles for specific eventualities are often associated with prototypical semantic relations,

e.g. if John causes a breaking event, he is the Agent; if he uses a hammer, it is the Instrument; and someone who

receives something is a Recipient (see Clause 5 for descriptions).
2.3 General linguistic units
2.3.1
lemma
lemmatized form
conventional form chosen to represent a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.2
lexeme

fundamental unit, generally associated to a set of word forms sharing a common meaning

Note 1 to entry: See ISO 24611.
2.3.3
lexical entry

container for managing a set of word forms and possibly one or several meanings [word senses (2.3.6)]

to describe a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.4
lexicon
resource comprising a collection of lexical entries (2.3.3) for a language
Note 1 to entry: See ISO 24611.
2.3.5
utterance

stretch of speech about which no assumptions have been made in terms of linguistic theory

Note 1 to entry: See Reference [12].
2.3.6
word sense
meaning associated with a lexeme (2.3.2) in a context

Note 1 to entry: The ‘river bank’ sense of bank and the ‘financial institution’ sense of bank are considered to be

two different word senses, or lexical units, with the same word form, or lexeme (2.3.2). I called him on the radio

and Call me a taxi are associated to different word senses of the lexeme (2.3.2)call. Unrelated senses, as in bank,

are called homonyms. Senses of the same word form or lexeme which are clearly related (and can be difficult to

distinguish) are called polysemes, e.g. Coins with an image of the king, preoccupied with body image, evokes a strong

mental image.
© ISO 2014 – All rights reserved 3
---------------------- Page: 11 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
3 Abbreviated terms
EngVallex English Valency Lexicon
LIRICS Linguistic Infrastructure for Interoperable Resources and Systems
PropBank Proposition Bank
SR semantic roles
SRL semantic role labelling
WSD word sense disambiguation
4 Purpose and justification

Semantic roles are arousing increasing interest in the information-processing community because they

make explicit the key conceptual relations of participation between a verb and its arguments; that is to

say, they specify ‘who’ did ‘what’ to ‘whom’, and ‘when’, ‘where’, ‘why’, and ‘how’. For English alone, there

are already several different semantic role frameworks, including FrameNet, VerbNet, LIRICS, EngVallex,

and PropBank. Although these have been developed independently, there are strong underlying

compatibilities between them, and they share a central definition of what a semantic role is, and what

its span is, within an individual sentence. In addition to defining key concepts, this part of ISO 24617

aims to clarify and specify these underlying compatibilities and provide, where possible, a mapping

between similar semantic roles across different frameworks. This mapping is intended to serve as an

illustration of how different semantic role definitions can be linked to each other across frameworks,

and presupposes a specification of clearly defined criteria for distinguishing semantic roles.

The specification will be used in two different situations:
— in annotations where the semantic roles are recorded in annotated corpora;

— as a dynamic structure produced by automatic systems, a process typically known as semantic role

labelling (SRL).
The objectives of this specification are to provide

— a reference set of data categories that define a structured collection of semantic roles with an

explicit semantics,

— a pivot representation based on a framework for defining semantic roles that can facilitate

mapping between different formalisms (alternative semantic role representations/syntactic

theories/eventually different languages) and, in the future, between different languages, and

— guidelines for creating new resources that will be immediately interoperable with pre-existing

resources.
5 The nature of semantic roles
5.1 General

For computers to make effective use of information encoded in text, they must be able to detect the

eventualities that are being described and the eventuality participants. The processing of a sentence like

John threw a ball to Mary in the park should result in the identification of a throwing event involving John

as the Agent of the event, Mary as the Recipient, and the ball as the item being thrown; the location of the

throwing event, or where it took place, is the park. This description of the event specifies the conceptual

relations of participation that the referents of the noun phrases play with respect to the event. The

semantic notions being specified are the roles of the participants in an eventuality (i.e. semantic roles).

4 © ISO 2014 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)

This part of ISO 24617 establishes LIRICS (see Annex A) as a reference set of semantic roles with precise

definitions. Researchers are free to define their own sets of semantic roles, but explicit information on

how they can be mapped to the reference set will make resources more interoperable. Many resources

currently map to PropBank, VerbNet, or FrameNet. Since this part of ISO 24617 includes mappings of

these resources to LIRICS, such mappings already qualify as meeting the requirement of interoperability.

Our throw example seems fairly straightforward, but complexities quickly arise. English, for instance,

allows not only several different syntactic constituents to present the same semantic role, but also

several different semantic roles to be presented by the same syntactic constituent. For decades, a central

concern of linguists has been the elucidation of the process of mapping back and forth between the

syntactic analysis of the sentence and the conceptual structure and relations in the event described. For

example, in the following two sentences,
(1) The flame melted the wax.
(2) The wax melted.

a standard syntactic parser represents the wax as the verb’s direct object in the first sentence and its

subject in the second. There is nothing overt to indicate that it has the same conceptual relation in

both cases despite the fact that it is expressed syntactically in a different way. We can capture this by

annotating the wax as having the same semantic role (or conceptual relation) in both sentences. It would

typically be labelled the Patient, the participant undergoing a change of state. Note that both sentences

are in the active voice, not the passive voice. In The wax was melted by the flame, the passive provides

syntactic evidence that the wax is playing the same role (Patient) that it plays in example (1). Since the

particular pair of syntactic variations illustrated by melt does not occur with every transitive verb [see

example (5)], it is not easily predictable. Other transitive verbs can also occur in an intransitive form

while maintaining the same semantic role for the subject as the transitive, as in the following example,

where soprano is the Agent of sing in both sentences (the aria is the Theme):
(3) The soprano sang an aria.
(4) The soprano sang.
The verb slice can also move the Patient (the bread) to subject position, as in
(5) John sliced the bread easily./ This bread slices easily.
although other transitive verbs, such as eat, cannot:
(6) John ate the apple.
(7) John ate.
(8) *The apple ate crunchily in the background.
The last sentence is starred (*) to indicate its ungrammaticality.

Accurate interpretation of the semantic roles of the verb arguments (i.e. ‘Who did what to whom?’) is a

crucial goal for natural language processing systems. Our ability to do this automatically has improved

enormously in recent years and has been largely based on the availability of annotated corpora. In

fact, there are corpora, such as FrameNet and PropBank, available with quite different semantic role

annotations, and this prompts questions about the nature and number of semantic roles. This part of

ISO 24617 attempts to provide definitions and examples clarifying their definition.

For semantic roles to maximize the benefit to the information processing community, it is desirable that

the definitions of the semantic roles should, as far as possible, have the following properties:

— consistently recognizable;
— able to clarify sense distinctions;
— generalizable;
© ISO 2014 – All rights reserved 5
---------------------- Page: 13 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
— machine learnable;
— able to provide an appropriate foundation for inferencing.

The purpose of the specifications for semantic role definitions in this part of ISO 24617 is to provide

these attributes.
5.2 Typical examples of semantic roles

A list of the best-known roles and the properties usually associated with them is given below. They are

taken from the EAGLES discussion on Standardizing Subcategorization (see Reference [61]). Comments

in parentheses have been added for clarification purposes and include comparisons with LIRICS

(Linguistic Infrastructure for Interoperable Resources and Systems), and occasionally with VerbNet.

NOTE These role definitions, and the ones for LIRICS, are quite general and can cover a wide range of

participant types. The specific preposition associated with an individual semantic role quite often adds nuances

of meaning in addition to that conveyed by the semantic role itself. For instance, in “Eat the fish with caution,” the

phrase “with caution” would typically be labelled as MANNER. Another example of a MANNER role could be the

phrase “in three bites” from “Eat the fish in three bites.” Clearly the interpretation of these two phrases will be

quite different, as will be their impact on the representation of the eating eventuality. Semantic role labels cannot

be expected to clarify all such types of subtle differences in meaning, and additional research on the definitions

[48]

of prepositions and their interactions with semantic roles, such as Srikumar and Roth, 2013 , is needed.

Agent

A participant that the meaning of the verb specifies as doing or causing something, possibly intentionally;

for example, as the subject of kill, eat, hit, smash, kick, and watch. (LIRICS has a similar Agent, which acts

intentionally or consciously.)
Patient

A participant that the verb characterizes as having something happen to it, and as being affected by

what happens to it; for example, as the object of kill, eat, and smash, but not of watch, hear, and love. (If

someone watches television, the television is not affected by the watching, so it would be a Theme rather

than a Patient. LIRICS has a similar Patient.)
Experiencer

A participant that is characterized as aware of something; for example, as the subject of love or as the

object of annoy. (LIRICS uses an Agent for these verbs and has no Experiencer; VerbNet has the same

role, but only when the Experiencer is affected by the event.)
Theme

A participant that is characterized as changing its position or condition, or as being in a state or position;

for example, as the object of give and hand or as the subjects of walk (in line with the policy of labelling

the object in motion as a Theme) and of die. (According to the EAGLES definition, people who die of

old age would not be considered to be Patients because an Agent has not acted upon

...

SLOVENSKI STANDARD
SIST ISO 24617-4:2018
01-september-2018
8SUDYOMDQMH]MH]LNRYQLPLYLUL2JURGMH]DVHPDQWLþQRR]QDþHYDQMH 6HP$) 
GHO6HPDQWLþQHYORJH 6HP$)65
Language resource management -- Semantic annotation framework (SemAF) -- Part 4:
Semantic roles (SemAF-SR)

Gestion de ressources linguistiques -- Cadre d'annotation sémantique (SemAF) -- Partie

4: Rôles sémantiques (SemAF-SR)
Ta slovenski standard je istoveten z: ISO 24617-4:2014
ICS:
01.020 7HUPLQRORJLMD QDþHODLQ Terminology (principles and
NRRUGLQDFLMD coordination)
35.060 Jeziki, ki se uporabljajo v Languages used in
informacijski tehniki in information technology
tehnologiji
SIST ISO 24617-4:2018 en,fr,de

2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

---------------------- Page: 1 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 2 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL ISO
STANDARD 24617-4
First edition
2014-08-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 4:
Semantic roles (SemAF-SR)
Gestion de ressources linguistiques — Cadre d’annotation sémantique
(SemAF) —
Partie 4: Rôles sémantiques (SemAF-SR)
Reference number
ISO 24617-4:2014(E)
ISO 2014
---------------------- Page: 3 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2014

All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form

or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior

written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of

the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2014 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Contents Page

Foreword ........................................................................................................................................................................................................................................iv

Introduction ..................................................................................................................................................................................................................................v

1 Scope ................................................................................................................................................................................................................................. 1

2 Terms and definitions ..................................................................................................................................................................................... 1

2.1 Formal semantic units ...................................................................................................................................................................... 1

2.2 Conceptual semantic units and relations ......................................................................................................................... 2

2.3 General linguistic units .................................................................................................................................................................... 3

3 Abbreviated terms .............................................................................................................................................................................................. 4

4 Purpose and justification ............................................................................................................................................................................ 4

5 The nature of semantic roles ................................................................................................................................................................... 4

5.1 General ........................................................................................................................................................................................................... 4

5.2 Typical examples of semantic roles ...................................................................................................................................... 6

6 Metamodel .................................................................................................................................................................................................................. 7

6.1 Key concepts ............................................................................................................................................................................................. 7

6.2 Introduction to a metamodel for semantic role annotation .........................................................................11

6.3 Criteria for distinguishing semantic roles ....................................................................................................................12

6.4 Defining eventuality frames......................................................................................................................................................13

7 Interactions .............................................................................................................................................................................................................14

7.1 Semantic roles and semantic types ....................................................................................................................................14

7.2 Complexities ...........................................................................................................................................................................................14

7.3 The role of context in semantic role assignment ....................................................................................................14

7.4 Fuzzy boundaries between roles ..........................................................................................................................................15

7.5 Multiple classification ....................................................................................................................................................................15

7.6 Inheritance relations between semantic roles .........................................................................................................15

8 Guidelines for developing new semantic role frameworks for languages and/

or domains ...............................................................................................................................................................................................................16

8.1 General ........................................................................................................................................................................................................16

8.2 Mapping from VerbNet to LIRICS .........................................................................................................................................17

Annex A (informative) Specification of ISO semantic roles ........................................................................................................19

Annex B (informative) Review of existing frameworks ..................................................................................................................28

Annex C (informative) Specification of the annotation language ........................................................................................37

Bibliography .............................................................................................................................................................................................................................41

© ISO 2014 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Foreword

ISO (the International Organization for Standardization) is a worldwide federation of national standards

bodies (ISO member bodies). The work of preparing International Standards is normally carried out

through ISO technical committees. Each member body interested in a subject for which a technical

committee has been established has the right to be represented on that committee. International

organizations, governmental and non-governmental, in liaison with ISO, also take part in the work.

ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of

electrotechnical standardization.

The procedures used to develop this document and those intended for its further maintenance are

described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the

different types of ISO documents should be noted. This document was drafted in accordance with the

editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).

Attention is drawn to the possibility that some of the elements of this document may be the subject of

patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of

any patent rights identified during the development of the document will be in the Introduction and/or

on the ISO list of patent declarations received (see www.iso.org/patents).

Any trade name used in this document is information given for the convenience of users and does not

constitute an endorsement.

For an explanation on the meaning of ISO specific terms and expressions related to conformity

assessment, as well as information about ISO’s adherence to the WTO principles in the Technical Barriers

to Trade (TBT) see the following URL: Foreword - Supplementary information

The committee responsible for this document is ISO/TC 37, Terminology and other language and content

resources, Subcommittee SC 4, Language resource management.

ISO 24617 consists of the following parts, under the general title Language resource management —

Semantic annotation framework (SemAF):
— Part 1: Time and events (SemAF-Time, ISO-TimeML)
— Part 2: Dialogue acts
— Part 4: Semantic roles (SemAF-SR)
— Part 5: Discourse structure (SemAF-DS)
— Part 7: Spatial information (ISO-Space)
The following parts are under preparation:
— Part 8: Semantic relations in discourse (SemAF-DRel)

Principles of semantic annotation (SemAF-Basics) will form the subject of future Part 6.

iv © ISO 2014 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
Introduction

This part of ISO 24617 aims to specify criteria for defining semantic roles (SRs), and is the outcome of

an agreement that the various semantic role frameworks being used to support data annotation (e.g.

FrameNet, VerbNet, PropBank, EngVallex, and LIRICS, to name only a few examples for English) have

strong underlying compatibilities. The goal is to provide both an explanation of these compatibilities and

a loose mapping between definitions of individual semantic roles, as listed in the different frameworks,

that will benefit the community as a whole.

The current specification has been developed under the aegis of the ISO Semantic Annotation Framework

(SemAF), where it is known as SemAF-SR.
The main parts of ISO 24617-4 consist of the following:
— Scope;
— Normative references;
— Terms and definitions;
— motivation and requirements;
— basic concepts and metamodel specifications;
— examples of mapping existing frameworks to the metamodel.

This part of ISO 24617 contains three informative annexes. In Annex A, the ISO semantic roles are

specified. In Annex B, information is provided both on past and current activities in semantic role

annotation and on tools and frame files. Annex C contains the abstract and concrete syntax for the

metamodel.
© ISO 2014 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 24617-4:2018
---------------------- Page: 8 ----------------------
SIST ISO 24617-4:2018
INTERNATIONAL STANDARD ISO 24617-4:2014(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 4:
Semantic roles (SemAF-SR)
1 Scope

The aim of this part of ISO 24617 is to propose a consensual annotation scheme for semantic roles; that

is to say, a scheme that indicates the role that a participant plays in an event or state, as described mostly

by a verb, and typically providing answers to questions such as “‘who’ did ‘what’ to ‘whom’”, and ‘when’,

‘where’, ‘why’, and ‘how’. This includes not only the semantic relations between a verb and its arguments

but also those relations that are relevant for other predicative elements such as nominalizations, nouns,

adjectives, and predicate modifiers; the predicating role of adverbs and the use of coercion fall outside

the scope of this part of ISO 24617.

NOTE In linguistics, coercion occurs when the grammatical context causes the language-user to reinterpret

[60]

all or parts of the semantic and/or formal features of a lexeme that appear in that context.

2 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
2.1 Formal semantic units
2.1.1
argument

formal semantic unit that is an essential element of a predicate argument structure (2.1.3) and can have

variable instantiations depending on the utterance

Note 1 to entry: An argument corresponds to a participant (2.2.5) of an eventuality (2.2.2) described by the

predicate argument structure (2.1.3).

Note 2 to entry: Arguments typically satisfy certain argument positions and can be described as being syntactico-

semantic notions, whereas participants (2.2.5) are semantico-conceptual. The standard view is that subsets of the

participants associated with an eventuality (2.2.2) are selected as arguments by the verb (or nominal or adjective)

expressing the eventuality (2.2.2). Other participants (2.2.5) are either incorporated or realized as eventuality

modifiers (2.2.4).

Note 3 to entry: Natural language predicates typically have one, two, or three arguments, although they can have

more.
2.1.2
predicate

formal semantic unit that represents a semantic relation between one or more arguments (2.1.1) in a

predicate argument structure (2.1.3)

Note 1 to entry: Predicates are indicated by predicative linguistic elements such as verbs, nouns, and adjectives.

© ISO 2014 – All rights reserved 1
---------------------- Page: 9 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.1.3
predicate argument structure

formal representation of the core semantic content of an utterance, consisting of a predicate (2.1.2)

constant, and its arguments (2.1.1)

Note 1 to entry: In classical logic-based semantics, this corresponds to predicate argument structures in first-

order predicate logic.

Note 2 to entry: One of the arguments (2.1.1) can be a variable uniquely identifying the instance of the predicate

argument structure to allow references to it in other predicate argument structures.

Note 3 to entry: The representation of event semantics is subject to many variations; some of them, such as

in Reference [41], can have separate predicates (2.1.2) for each semantic role (2.2.6) relation. In this case, the

predicate argument structure of an utterance is the sum of the individual predicate semantic role (2.2.6) assertions

representing the semantic content of the utterance.
2.2 Conceptual semantic units and relations
2.2.1
entity
conceptual semantic unit that typically functions as a participant (2.2.5)

Note 1 to entry: An entity is an individual such as a person, organization, physical object, or logical entity, as

well as, on occasion, a number, quantity, dimension, or a reification of an eventuality, a property, or a quality, e.g.

emotion (anger, love), the value of a colour, etc.
2.2.2
eventuality

event, state, process, or action which can have participants (2.2.5) and which is being referred to by a

verbal, adjectival, or nominal description in an utterance

Note 1 to entry: The formal representation of an eventuality is a predicate argument structure (2.1.3).

Note 2 to entry: See ISO 24617-1. An eventuality can also be described as ‘something that can be said to obtain or

hold true, to happen or to occur’, as in ISO 24617-1. As such, they can be actual, hypothetical, or generic, covering

situations such as “You should go home,” or “He might be John’s brother.”
2.2.3
eventuality frame

generalized abstract specification of the word sense (2.3.6) associated with an eventuality (2.2.2) in an

utterance

Note 1 to entry: The frame consists of the specification of (a) a predicate (2.1.2) that can participate in a class

hierarchy if such a hierarchy is specified, and (b) the arguments (2.1.1) that this predicate (2.1.2) expects along

with their semantic roles (2.2.6).
2.2.4
eventuality modifier

particular type of participant (2.2.5) that completes the description of an eventuality (2.2.2) but is

optional and not essential

Note 1 to entry: Eventuality modifiers are distinct from other types of participants in that they are used in

supplying information that is typically more peripheral and more general, for example, situating the eventuality

in time or space.

Note 2 to entry: In FrameNet, these would be peripheral frame elements and in PropBank, ArgM’s.

Note 3 to entry: Eventuality modifiers typically correspond to syntactic adjuncts.

2 © ISO 2014 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
2.2.5
participant

conceptual semantic unit referred to by one or more lexical items in an utterance, which is or can be

involved in an eventuality (2.2.2)

Note 1 to entry: Both entities (2.2.1) and eventualities (2.2.2) can function as participants.

2.2.6
semantic role
mode of involvement of a participant (2.2.5) in an eventuality (2.2.2)

Note 1 to entry: Semantic roles for specific eventualities are often associated with prototypical semantic relations,

e.g. if John causes a breaking event, he is the Agent; if he uses a hammer, it is the Instrument; and someone who

receives something is a Recipient (see Clause 5 for descriptions).
2.3 General linguistic units
2.3.1
lemma
lemmatized form
conventional form chosen to represent a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.2
lexeme

fundamental unit, generally associated to a set of word forms sharing a common meaning

Note 1 to entry: See ISO 24611.
2.3.3
lexical entry

container for managing a set of word forms and possibly one or several meanings [word senses (2.3.6)]

to describe a lexeme (2.3.2)
Note 1 to entry: See ISO 24611.
2.3.4
lexicon
resource comprising a collection of lexical entries (2.3.3) for a language
Note 1 to entry: See ISO 24611.
2.3.5
utterance

stretch of speech about which no assumptions have been made in terms of linguistic theory

Note 1 to entry: See Reference [12].
2.3.6
word sense
meaning associated with a lexeme (2.3.2) in a context

Note 1 to entry: The ‘river bank’ sense of bank and the ‘financial institution’ sense of bank are considered to be

two different word senses, or lexical units, with the same word form, or lexeme (2.3.2). I called him on the radio

and Call me a taxi are associated to different word senses of the lexeme (2.3.2)call. Unrelated senses, as in bank,

are called homonyms. Senses of the same word form or lexeme which are clearly related (and can be difficult to

distinguish) are called polysemes, e.g. Coins with an image of the king, preoccupied with body image, evokes a strong

mental image.
© ISO 2014 – All rights reserved 3
---------------------- Page: 11 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
3 Abbreviated terms
EngVallex English Valency Lexicon
LIRICS Linguistic Infrastructure for Interoperable Resources and Systems
PropBank Proposition Bank
SR semantic roles
SRL semantic role labelling
WSD word sense disambiguation
4 Purpose and justification

Semantic roles are arousing increasing interest in the information-processing community because they

make explicit the key conceptual relations of participation between a verb and its arguments; that is to

say, they specify ‘who’ did ‘what’ to ‘whom’, and ‘when’, ‘where’, ‘why’, and ‘how’. For English alone, there

are already several different semantic role frameworks, including FrameNet, VerbNet, LIRICS, EngVallex,

and PropBank. Although these have been developed independently, there are strong underlying

compatibilities between them, and they share a central definition of what a semantic role is, and what

its span is, within an individual sentence. In addition to defining key concepts, this part of ISO 24617

aims to clarify and specify these underlying compatibilities and provide, where possible, a mapping

between similar semantic roles across different frameworks. This mapping is intended to serve as an

illustration of how different semantic role definitions can be linked to each other across frameworks,

and presupposes a specification of clearly defined criteria for distinguishing semantic roles.

The specification will be used in two different situations:
— in annotations where the semantic roles are recorded in annotated corpora;

— as a dynamic structure produced by automatic systems, a process typically known as semantic role

labelling (SRL).
The objectives of this specification are to provide

— a reference set of data categories that define a structured collection of semantic roles with an

explicit semantics,

— a pivot representation based on a framework for defining semantic roles that can facilitate

mapping between different formalisms (alternative semantic role representations/syntactic

theories/eventually different languages) and, in the future, between different languages, and

— guidelines for creating new resources that will be immediately interoperable with pre-existing

resources.
5 The nature of semantic roles
5.1 General

For computers to make effective use of information encoded in text, they must be able to detect the

eventualities that are being described and the eventuality participants. The processing of a sentence like

John threw a ball to Mary in the park should result in the identification of a throwing event involving John

as the Agent of the event, Mary as the Recipient, and the ball as the item being thrown; the location of the

throwing event, or where it took place, is the park. This description of the event specifies the conceptual

relations of participation that the referents of the noun phrases play with respect to the event. The

semantic notions being specified are the roles of the participants in an eventuality (i.e. semantic roles).

4 © ISO 2014 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)

This part of ISO 24617 establishes LIRICS (see Annex A) as a reference set of semantic roles with precise

definitions. Researchers are free to define their own sets of semantic roles, but explicit information on

how they can be mapped to the reference set will make resources more interoperable. Many resources

currently map to PropBank, VerbNet, or FrameNet. Since this part of ISO 24617 includes mappings of

these resources to LIRICS, such mappings already qualify as meeting the requirement of interoperability.

Our throw example seems fairly straightforward, but complexities quickly arise. English, for instance,

allows not only several different syntactic constituents to present the same semantic role, but also

several different semantic roles to be presented by the same syntactic constituent. For decades, a central

concern of linguists has been the elucidation of the process of mapping back and forth between the

syntactic analysis of the sentence and the conceptual structure and relations in the event described. For

example, in the following two sentences,
(1) The flame melted the wax.
(2) The wax melted.

a standard syntactic parser represents the wax as the verb’s direct object in the first sentence and its

subject in the second. There is nothing overt to indicate that it has the same conceptual relation in

both cases despite the fact that it is expressed syntactically in a different way. We can capture this by

annotating the wax as having the same semantic role (or conceptual relation) in both sentences. It would

typically be labelled the Patient, the participant undergoing a change of state. Note that both sentences

are in the active voice, not the passive voice. In The wax was melted by the flame, the passive provides

syntactic evidence that the wax is playing the same role (Patient) that it plays in example (1). Since the

particular pair of syntactic variations illustrated by melt does not occur with every transitive verb [see

example (5)], it is not easily predictable. Other transitive verbs can also occur in an intransitive form

while maintaining the same semantic role for the subject as the transitive, as in the following example,

where soprano is the Agent of sing in both sentences (the aria is the Theme):
(3) The soprano sang an aria.
(4) The soprano sang.
The verb slice can also move the Patient (the bread) to subject position, as in
(5) John sliced the bread easily./ This bread slices easily.
although other transitive verbs, such as eat, cannot:
(6) John ate the apple.
(7) John ate.
(8) *The apple ate crunchily in the background.
The last sentence is starred (*) to indicate its ungrammaticality.

Accurate interpretation of the semantic roles of the verb arguments (i.e. ‘Who did what to whom?’) is a

crucial goal for natural language processing systems. Our ability to do this automatically has improved

enormously in recent years and has been largely based on the availability of annotated corpora. In

fact, there are corpora, such as FrameNet and PropBank, available with quite different semantic role

annotations, and this prompts questions about the nature and number of semantic roles. This part of

ISO 24617 attempts to provide definitions and examples clarifying their definition.

For semantic roles to maximize the benefit to the information processing community, it is desirable that

the definitions of the semantic roles should, as far as possible, have the following properties:

— consistently recognizable;
— able to clarify sense distinctions;
— generalizable;
© ISO 2014 – All rights reserved 5
---------------------- Page: 13 ----------------------
SIST ISO 24617-4:2018
ISO 24617-4:2014(E)
— machine learnable;
— able to provide an appropriate foundation for inferencing.

The purpose of the specifications for semantic role definitions in this part of ISO 24617 is to provide

these attributes.
5.2 Typical examples of semantic roles

A list of the best-known roles and the properties usually associated with them is given below. They are

taken from the EAGLES discussion on Standardizing Subcategorization (see Reference [61]). Comments

in parentheses have been added for clarification purposes and include comparisons with LIRICS

(Linguistic Infrastructure for Interoperable Resources and Systems), and occasionally with VerbNet.

NOTE These role definitions, and the ones for LIRICS, are quite general and can cover a wide range of

participant types. The specific preposition associated with an individual semantic role quite often adds nuances

of meaning in addition to that conveyed by the semantic role itself. For instance, in “Eat the fish with caution,” the

phrase “with caution” would typically be labelled as MANNER. Another example of a MANNER role could be the

phrase “in three bites” from “Eat the fish in three bites.” Clearly the interpretation of these two phrases will be

quite different, as will be their impact on the representation of the eating eventuality. Semantic role labels cannot

be expected to clarify all such types of subtle differences in meaning, and additional research on the definitions

[48]

of prepositions and their interactions with semantic roles, such as Srikumar and Roth, 2013 , is needed.

Agent

A participant that the meaning of the verb specifies as doing or causing something, possibly intentionally;

for example, as the subject of kill, eat, hit, smash, kick, and watch. (LIRICS has a similar Agent, which acts

intentionally or consciously.)
Patient

A participant that the verb characterizes as having something happen to it, and as being affected by

what happens to it; for example, as the object of kill, eat, and smash, but not of watch, hear, and love. (If

someone watches television, the television is not affected by the watching, so it would be a Theme rather

than a Patient. LIRICS has a similar Patient.)
Experiencer

A participant that is characterized as aware of something; for example, as the subject of love or as the

object of annoy. (LIRICS uses an Agent for these verbs and has no Experiencer; VerbNet has the same

role, but only when the Experiencer is affected by the event.)
Theme

A participant that is characterized as changing its position or condition, or as being in a state or position;

for example, as the object of give and hand or as the subjects of walk (in line with the policy of labelling

the object in motion as a Theme) and of die. (According to the EAGLES definition, people who die of

old age would not be considered to be Patients because an Agent has not acted upon the

...

Questions, Comments and Discussion

Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.