ISO 24617-2:2012
(Main)Language resource management — Semantic annotation framework (SemAF) — Part 2: Dialogue acts
Language resource management — Semantic annotation framework (SemAF) — Part 2: Dialogue acts
ISO 24617-2:2012 provides a set of empirically and theoretically well-motivated concepts for dialogue annotation, a formal language for expressing dialogue annotations -- the dialogue act markup language (DiAML) -- and a method for segmenting a dialogue into semantic units. This allows the manual or automatic annotation of dialogue segments with information about the communicative actions which the participants perform by their contributions to the dialogue. It supports multidimensional annotation, in which units in dialogue are viewed as having multiple communicative functions. The DiAML language has an XML-based representation format and a formal semantics which makes it possible to apply inference to DiAML representations. ISO 24617-2:2012 specifies data categories for reference sets of communicative functions and dimensions of dialogue analysis and provides principles and guidelines for extending these sets or selecting coherent subsets of them. Additionally, it provides guidelines for annotators and annotated examples. It is applicable to spoken, written and multimodal dialogues involving two or more participants.
Gestion des ressources langagières — Cadre d'annotation sémantique (SemAF) — Partie 2: Actes de dialogue
Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 2. del: Dialogi
Ta del standarda ISO 24617 podaja sklop empirično in teoretično utemeljenih pojmov za označevanje dialoga, uradni jezik za izražanje označevanja dialoga – jezik za označevanje dialoga (DiAML) – in metodo za segmentiranje dialoga v semantične enote. To omogoča ročno ali samodejno označevanje segmentov dialoga z informacijami o komunikacijskih dejanjih, ki jih izvedejo udeleženci, ki sodelujejo v dialogu. Podpira multidimenzionalno označevanje, pri katerem se enote dialoga obravnavajo, kot da imajo več komunikacijskih funkcij. Jezik DiAML ima format predstavitve, ki temelji na XML in formalni semantiki, kar omogoča sklepanje na podlagi predstavitev DiAML. Ta del standarda ISO 24617 določa podatkovne kategorije za referenčne sklope komunikacijskih funkcij in dimenzij analize dialoga ter zagotavlja načela in smernice za razširitev teh sklopov ali izbiro skladnih podsklopov. Zagotavlja tudi smernice za označevalce in primere označevanja. Uporablja se lahko za govorjene, zapisane in multimodalne dialoge, ki vključujejo dva ali več udeležencev.
General Information
Relations
Buy Standard
Standards Content (Sample)
SLOVENSKI STANDARD
SIST ISO 24617-2:2013
01-julij-2013
Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 2.
del: Dialogi
Language resource management -- Semantic annotation framework (SemAF) -- Part 2:
Dialogue acts
Gestion des ressources langagières -- Cadre d'annotation sémantique (SemAF) -- Partie
2: Actes de dialogue
Ta slovenski standard je istoveten z: ISO 24617-2:2012
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 24617-2:2013 en,fr,de
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
SIST ISO 24617-2:2013
---------------------- Page: 2 ----------------------
SIST ISO 24617-2:2013
INTERNATIONAL ISO
STANDARD 24617-2
First edition
2012-09-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 2:
Dialogue acts
Gestion des ressources langagières — Cadre d'annotation sémantique
(SemAF) —
Partie 2: Actes de dialogue
Reference number
ISO 24617-2:2012(E)
©
ISO 2012
---------------------- Page: 3 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2012
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing 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 2012 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
Contents Page
Foreword . iv
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 1
4 Purpose and justification . 5
5 Basic concepts and metamodel . 6
6 Definition of communicative functions . 8
7 Annotation schemes . 9
7.1 Structure of annotation schemes . 9
7.2 Multidimensionality and multifunctionality . 10
7.3 Multidimensionality, clustering and dimensions . 11
7.4 Dimension- specific and general-purpose functions . 11
8 Dialogue segmentation . 13
9 Dimensions . 14
9.1 Task. 15
9.2 Auto-Feedback . 15
9.3 Allo-Feedback . 15
9.4 Turn Management . 15
9.5 Time Management . 16
9.6 Discourse Structuring . 16
9.7 Social Obligations Management . 16
9.8 Own Communication Management . 16
9.9 Partner Communication Management . 16
10 Core dialogue acts . 17
10.1 General-purpose functions . 19
10.2 Dimension-specific functions . 20
10.3 Function qualifiers . 22
11 Dialogue act markup language (DiAML) . 23
11.1 Abstract syntax . 23
11.2 Concrete syntax . 24
12 Principles for extending and restricting the standard . 25
12.1 Main design principles . 25
12.2 Schema extension . 27
12.3 Scheme restriction . 27
Annex A (informative) Annotation guidelines . 29
Annex B (informative) Annotated dialogue examples . 43
Annex C (normative) Formal definition of DiAML . 56
Annex D (normative) DiAML technical schema . 63
Annex E (normative) Data categories for core concepts . 68
Annex F (informative) Examples of possible additional data categories . 88
Annex G (informative) Concepts in existing schemes . 90
Bibliography . 100
© ISO 2012 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(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.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
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.
ISO 24617-2 was prepared by Technical Committee 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:
Part 1: Time and events (SemAF-Time, ISO-TimeML)
Part 2: Dialogue acts
The following parts are under preparation:
Part 3: Named entities (SemAF-NE)
Part 4: Semantic roles (SemAF-SRL)
Part 5: Discourse structure (SemAF-DS)
Part 6: Principles of semantic annotation (SemAF-Basics)
Part 7: Spatial information (ISO-Space)
Part 8: Semantic relations in discourse (SemAF-DRel)
iv © ISO 2012 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24617-2:2013
INTERNATIONAL STANDARD ISO 24617-2:2012(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 2:
Dialogue acts
1 Scope
This part of ISO 24617 provides a set of empirically and theoretically well-motivated concepts for dialogue
annotation, a formal language for expressing dialogue annotations — the dialogue act markup language
(DiAML) — and a method for segmenting a dialogue into semantic units. This allows the manual or automatic
annotation of dialogue segments with information about the communicative actions which the participants
perform by their contributions to the dialogue. It supports multidimensional annotation, in which units in
dialogue are viewed as having multiple communicative functions. The DiAML language has an XML-based
representation format and a formal semantics which makes it possible to apply inference to DiAML
representations.
This part of ISO 24617 specifies data categories for reference sets of communicative functions and
dimensions of dialogue analysis and provides principles and guidelines for extending these sets or selecting
coherent subsets of them. Additionally, it provides guidelines for annotators and annotated examples. It is
applicable to spoken, written and multimodal dialogues involving two or more participants.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 12620:2009, Terminology and other language resources — Specification of data categories and
management of a Data Category Registry for language resources
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO 24612:2011, Language resource management — Linguistic annotation framework
3 Terms and definitions
1)
For the purposes of this document, the following terms and definitions apply.
1) In this document, “he”, “him” and “his” are used in a generic sense, without implying any gender-related distinctions.
© ISO 2012 – All rights reserved 1
---------------------- Page: 7 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.1
addressee
dialogue (3.5) participant (3.13) oriented to by the sender (3.18) in a manner to suggest that his utterances
(3.22) are particularly intended for this participant and that some response is therefore anticipated from this
participant, more so than from the other participants
Note to entry: This definition is a de facto standard in the linguistics literature. It has been slightly modified here, in
replacing “speaker” by “sender” and avoiding the use of ambiguous pronouns. Goffman's original definition says: “dialogue
participant oriented to by the speaker in a manner to suggest that his utterances are particularly intended for him and that
some response is therefore anticipated from him/her, more so than from the other participants”.
[SOURCE: Goffman (1981).]
3.2
allo-feedback act
feedback act (3.8) where the sender (3.18) elicits information about the addressee's (3.1) processing of an
utterance (3.22) that the sender contributed to the dialogue (3.5) or where the sender provides information
about his perceived processing by the addressee of an utterance that the sender contributed to the dialogue
before
EXAMPLE A: Now move up.
B: Slightly northeast you mean?
A: Slightly yeah.
A performs an allo-feedback act signalling that he thinks B understood his first utterance correctly.
3.3
auto-feedback act
feedback act (3.8) where the sender (3.18) provides information about his own processing of an utterance
(3.22) contributed to the dialogue (3.5) by another participant (3.13)
EXAMPLE B's utterance in the example dialogue fragment in (3.2) signals that he is uncertain whether he
understood the previous utterance correctly.
3.4
communicative function
property of certain stretches of communicative behaviour, describing how the behaviour changes the
information state (3.12) of an understander of the behaviour
Note to entry: A communicative function may be “qualified”, i.e. one or more qualifiers (3.14) may be associated with it.
For example, an answer may be qualified as “uncertain” and the acceptance of a request may be “conditional”. See 10.3
for explanation and examples.
3.5
dialogue
exchange of utterances (3.22) between two or more persons or artificial conversational systems
3.6
dialogue act
communicative activity of a dialogue (3.5) participant (3.13), interpreted as having a certain communicative
function (3.4) and semantic content (3.16)
Note to entry: A dialogue act may also have certain functional dependence relations (3.10), rhetorical relations (3.15) and
feedback dependence relations (3.9) with other units in a dialogue (3.5).
3.7
dimension
class of dialogue acts (3.6) that are concerned with a particular aspect of communication, corresponding to a
particular category of semantic content
2 © ISO 2012 – All rights reserved
---------------------- Page: 8 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
EXAMPLE Dialogue acts advancing the task or activity that motivates the dialogue (the Task dimension), dialogue
acts providing and eliciting feedback (the Auto- and Allo-Feedback dimensions) and dialogue acts for allocating the
speaker role (the Turn Management dimension).
Note to entry: See Clauses 5, 7 and 9 for discussion and more examples.
3.8
feedback act
dialogue act (3.6) which provides or elicits information about the sender's (3.18) or the addressee's (3.1)
processing of something that was uttered in the dialogue
Note to entry: Two classes of feedback are distinguished in this part of ISO 24617: allo-feedback acts (3.2) and auto-
feedback acts (3.3).
3.9
feedback dependence relation
relation between a feedback act (3.8) and the stretch of communicative behaviour whose processing the act
provides or elicits information about
EXAMPLE In the example that accompanies definition 3.2, both the allo-feedback act expressed by utterance 3 and
the auto-feedback act expressed by utterance 2 have a feedback dependence relation to utterance 1.
3.10
functional dependence relation
relation between a given dialogue act (3.6) and a preceding dialogue act on which the semantic content of the
given dialogue act depends due to its communicative function (3.4)
EXAMPLE The relation between an answer and the corresponding question, such as between utterance 3 and
utterance 2 in the example accompanying definition 3.2; or the relation between the acceptance of an offer and the
corresponding offer.
Note to entry: A dialogue act, A2, may also depend on another dialogue act, A1, occurring earlier in a dialogue because
of relations between their semantic contents, e.g. because A2 contains a reference to an element occurring in A1. This is
not a functional dependence relation, since it is not due to A2's communicative function.
3.11
functional segment
minimal stretch of communicative behaviour that has one or more communicative functions (3.4)
EXAMPLE The functional segment corresponding to the answer given by S in the following dialogue fragment does
not include the parts “Just a moment please” and “. let me see.” but only the parts “the first train to the airport on
Sunday morning is” and “at 5:45”:
1. U: What time is the first train to the airport on Sunday morning please?
2. S: Just a moment please. the first train to the airport on Sunday morning is . let me see. at 5:45.
Note 1 to entry: A consequence of this definition is that functional segments may be discontinuous, may overlap or be
embedded and may contain parts contributed by different participants.
Note 2 to entry: The condition of being “minimal” ensures that functional segments do not include material that does
not contribute to the expression of a communicative function that identifies the segment.
3.12
information state
context
totality of a dialogue (3.5) participant's (3.13) beliefs, assumptions, expectations, goals, preferences, hopes
and other attitudes that may influence the participant's interpretation and generation of communicative
behaviour
3.13
participant
person or artificial agent involved in the exchange of utterances (3.22)
© ISO 2012 – All rights reserved 3
---------------------- Page: 9 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.14
qualifier
predicate that can be associated with a communicative function (3.4)
EXAMPLE A: Would you like to have some coffee?
B: Only if you have it ready.
B's utterance accepts A's offer under a certain condition; this can be described by qualifying the communicative function
Accept Offer with the predicate “conditional”.
Note to entry: See 10.3 for more examples.
3.15
rhetorical relation
relation between two dialogue acts (3.6), indicating a pragmatic connection between the two or between their
semantic contents (3.16)
EXAMPLE 1 The statement in the second utterance which follows provides a motivation for the question in the first
utterance:
A: Can you tell me what flights there are to Sydney on Saturday? I'd like to attend my mother's 80th birthday.
EXAMPLE 2 A rhetorical relation between the semantic contents of two dialogue act occurs in the following, where the
content of B's statement mentions a cause for the content of A's statement:
A: I can never find these stupid remote controls
B: That's because they don't have a fixed location
Note to entry: Relations such as elaboration, explanation, justification, cause and concession have been studied
extensively in the analysis of (monologue) text, where they are often called “rhetorical relations” or “discourse relations”
and are mostly viewed either as relations between text segments or as relations between events or propositions,
described in text segments. See, for example, Hovy and Maier, 1992, Lascarides & Asher, 2007 or Mann & Thompson,
1988. Many of these relations also occur in dialogue, either as relations between dialogue acts or between the semantic
contents of dialogue acts.
3.16
semantic content
information, situation, action, event or objects that a stretch of communicative behaviour refers to
3.17
semantic content category
semantic content type
kind of information, situation, action, event or objects that form the semantic content (3.16) of a dialogue act
(3.6)
EXAMPLE The various dimensions (3.7) defined in this part of ISO 24617 correspond to categories of semantic
content. In particular, the Task dimension corresponds to the category of task-specific actions and information; the Allo-
and Auto-Feedback dimensions correspond to the categories of information about the processing by the current speaker
or by the addressee, respectively, of something that was said before; the Turn Management dimension corresponds to the
category of information about the allocation of the speaker role and so forth.
3.18
sender
dialogue (3.5) participant (3.13) who produces a dialogue act (3.6)
3.19
speaker
sender (3.18) of a dialogue act (3.6) in the form of speech, possibly combined with nonverbal communicative
behaviour
Note to entry: A dialogue participant may say something while another participant occupies the speaker role (3.20),
therefore the term “speaker” is not synonymous with “participant who occupies the speaker role”.
4 © ISO 2012 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.20
speaker role
role occupied by a dialogue (3.5) participant (3.13) who has temporary control of the dialogue and speaks for
some period of time
[SOURCE: DAMSL Revised Manual.]
3.21
turn unit
stretch of communicative activity produced by one participant (3.13) who occupies the speaker role (3.20),
bounded by periods where another participant occupies the speaker role
3.22
utterance
anything said, written, keyed, gesticulated or otherwise expressed
Note to entry: An utterance is mostly a part of what a sender contributes in a turn unit.
4 Purpose and justification
The notion of a dialogue act plays a key role in the analysis of spoken and multimodal dialogue, as well as in
the design of spoken dialogue systems and embodied conversational agents. These activities all depend on
the availability of dialogue corpora, annotated with dialogue act information.
Over the years a variety of dialogue act annotation schemes have been developed, such as those of the
TRAINS human-computer dialogue project (Allen et al., 1994), the Map Task studies of human-human
dialogue (Carletta et al., 1996) and of the Verbmobil speech translation project (Alexandersson et al., 1998).
These schemes were developed for specific purposes and application domains. They contain overlapping sets
of concepts and make use of often mutually inconsistent terminology, sometimes employing different terms for
the same concept or the same term for different concepts.
The multidimensional DIT scheme (Bunt, 1984) was developed for information-seeking dialogues without
depending on a particular domain. The DAMSL scheme (Dialogue Act Markup using Several Layers, Allen
and Core,1997; Core et al., 1998) constitutes an application-independent multidimensional annotation
++
scheme. The DIT scheme (Bunt, 2006; 2009) combines the DIT scheme with concepts from DAMSL and
other more recent schemes into a comprehensive general-purpose annotation scheme.
In the EU-funded project LIRICS (Linguistic Infrastructure for Interoperable Resources and Systems, Romary
++
et al., 2007) a reference set of dialogue acts, taken from the DIT taxonomy, was defined in the form of data
categories, following ISO 12620. This set of concepts has been tested for its usability and coverage a) in the
manual annotation of spoken dialogues in English, Dutch and Italian and b) in the automatic annotation of
spoken and multimodal dialogue in English and forms a significant part of the background of this part of
ISO 24617.
The main purpose of this part of ISO 24617 is to define a reference set of domain-independent basic concepts
for dialogue act annotation, plus a formal language, based on XML, for representing such annotations.
Guidelines are provided for how to use the defined concepts and the annotation language, supported by
extended examples. This formal language, the Dialogue act markup language (DiAML) has a formal
semantics, which makes it possible to apply techniques for automatic reasoning to DiAML annotations.
Guidelines and principles are also provided for extending the set of concepts defined in this part of ISO 24617,
for example, with domain-specific concepts, as well as for selecting coherent subsets.
© ISO 2012 – All rights reserved 5
---------------------- Page: 11 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
5 Basic concepts and metamodel
The term “dialogue act” is often used rather loosely in the sense of a speech act used in dialogue. Indeed, the
idea of interpreting communicative behaviour in terms of actions, such as questions, promises and requests,
goes back to speech act theory (Austin, 1962; Searle, 1969). But where speech act theory is primarily an
action-based approach to meaning within the philosophy of language, dialogue act theory is an
empirically-based approach to the computational modelling of linguistic and nonverbal communicative
behaviour in dialogue.
Dialogue acts offer a way of characterizing the meaning of communicative behaviour in terms of update
operations, to be applied to the information states of participants in the dialogue; this approach is commonly
known as the “information-state update” or “context-change” approach — see e.g. Bunt (1989; 2000a); Traum
and Larsson (2003). For instance, when an addressee understands the utterance “Do you know what time it
is?” as a question about the time, then the addressee's information state is updated to contain (among other
things) the information that the speaker does not know what time it is and would like to know that. If, by
contrast, it is understood that the speaker is reproaching the addressee for being late, then the addressee's
information state is updated to include (among other things) the information that the speaker does know what
time it is. Distinctions such as that between a question and a reproach concern the communicative function of
a dialogue act, which is one of its two main components. The other main component is its semantic content,
which describes the objects, properties, relations, situations, actions or events that the dialogue act is about.
The communicative function of a dialogue act specifies how an addressee should update his information state
with the information expressed in the semantic content when he understands the dialogue act.
A dialogue act as defined in this part of ISO 24617 (3.6) is a semantic unit of communicative behaviour.
Dialogue act annotation is the marking up of stretches of dialogue with information about the dialogue acts
performed in these stretches and is often limited to assigning communicative function tags. A dialogue act
being a semantic unit in communicative behaviour, the question arises as to which stretches of communicative
behaviour are considered as corresponding to dialogue acts. Spoken dialogues are traditionally segmented
into turns, defined as stretches of communicative behaviour produced by one speaker, bounded by periods of
inactivity of that speaker. Turns in this sense can be quite long and complex and are therefore not very useful
units of behaviour for assigning communicative functions to. Communicative functions can be assigned more
accurately to smaller units, which are called functional segments and which are defined as the minimal
stretches of communicative behaviour that are functionally relevant. See Clause 8 for more details about
dialogue segmentation.
Inherent to the notion of a dialogue act is that there is an agent who produces the dialogue act, called the
“sender” and one or more agents who are addressed, called “addressees”. Dialogue studies often focus on
two-person dialogues, in which case the dialogue acts have only one addressee. Besides sender and
addressee(s), there may be various types of side-participants who are present but do not or only marginally
participate (see Clark, 1996).
Dialogue act annotation is often limited to assigning communicative functions to dialogue segments, which
corresponds intuitively to indicating the type of communicative action that is performed. A semantically more
complete characterization also provides information about the type of semantic content. The DAMSL
annotation scheme distinguishes three categories of semantic content: task, task management and
communication, which indicate whether the semantic content of the dialogue act is concerned with performing
the task which underlies the dialogue or with discussing how to perform the task or with the communication.
++
The DIT scheme distinguishes a number of subcategories of communication-related information, such as
feedback informa
...
INTERNATIONAL ISO
STANDARD 24617-2
First edition
2012-09-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 2:
Dialogue acts
Gestion des ressources langagières — Cadre d'annotation sémantique
(SemAF) —
Partie 2: Actes de dialogue
Reference number
ISO 24617-2:2012(E)
©
ISO 2012
---------------------- Page: 1 ----------------------
ISO 24617-2:2012(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2012
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing 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 2012 – All rights reserved
---------------------- Page: 2 ----------------------
ISO 24617-2:2012(E)
Contents Page
Foreword . iv
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 1
4 Purpose and justification . 5
5 Basic concepts and metamodel . 6
6 Definition of communicative functions . 8
7 Annotation schemes . 9
7.1 Structure of annotation schemes . 9
7.2 Multidimensionality and multifunctionality . 10
7.3 Multidimensionality, clustering and dimensions . 11
7.4 Dimension- specific and general-purpose functions . 11
8 Dialogue segmentation . 13
9 Dimensions . 14
9.1 Task. 15
9.2 Auto-Feedback . 15
9.3 Allo-Feedback . 15
9.4 Turn Management . 15
9.5 Time Management . 16
9.6 Discourse Structuring . 16
9.7 Social Obligations Management . 16
9.8 Own Communication Management . 16
9.9 Partner Communication Management . 16
10 Core dialogue acts . 17
10.1 General-purpose functions . 19
10.2 Dimension-specific functions . 20
10.3 Function qualifiers . 22
11 Dialogue act markup language (DiAML) . 23
11.1 Abstract syntax . 23
11.2 Concrete syntax . 24
12 Principles for extending and restricting the standard . 25
12.1 Main design principles . 25
12.2 Schema extension . 27
12.3 Scheme restriction . 27
Annex A (informative) Annotation guidelines . 29
Annex B (informative) Annotated dialogue examples . 43
Annex C (normative) Formal definition of DiAML . 56
Annex D (normative) DiAML technical schema . 63
Annex E (normative) Data categories for core concepts . 68
Annex F (informative) Examples of possible additional data categories . 88
Annex G (informative) Concepts in existing schemes . 90
Bibliography . 100
© ISO 2012 – All rights reserved iii
---------------------- Page: 3 ----------------------
ISO 24617-2:2012(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.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
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.
ISO 24617-2 was prepared by Technical Committee 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:
Part 1: Time and events (SemAF-Time, ISO-TimeML)
Part 2: Dialogue acts
The following parts are under preparation:
Part 3: Named entities (SemAF-NE)
Part 4: Semantic roles (SemAF-SRL)
Part 5: Discourse structure (SemAF-DS)
Part 6: Principles of semantic annotation (SemAF-Basics)
Part 7: Spatial information (ISO-Space)
Part 8: Semantic relations in discourse (SemAF-DRel)
iv © ISO 2012 – All rights reserved
---------------------- Page: 4 ----------------------
INTERNATIONAL STANDARD ISO 24617-2:2012(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 2:
Dialogue acts
1 Scope
This part of ISO 24617 provides a set of empirically and theoretically well-motivated concepts for dialogue
annotation, a formal language for expressing dialogue annotations — the dialogue act markup language
(DiAML) — and a method for segmenting a dialogue into semantic units. This allows the manual or automatic
annotation of dialogue segments with information about the communicative actions which the participants
perform by their contributions to the dialogue. It supports multidimensional annotation, in which units in
dialogue are viewed as having multiple communicative functions. The DiAML language has an XML-based
representation format and a formal semantics which makes it possible to apply inference to DiAML
representations.
This part of ISO 24617 specifies data categories for reference sets of communicative functions and
dimensions of dialogue analysis and provides principles and guidelines for extending these sets or selecting
coherent subsets of them. Additionally, it provides guidelines for annotators and annotated examples. It is
applicable to spoken, written and multimodal dialogues involving two or more participants.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 12620:2009, Terminology and other language resources — Specification of data categories and
management of a Data Category Registry for language resources
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO 24612:2011, Language resource management — Linguistic annotation framework
3 Terms and definitions
1)
For the purposes of this document, the following terms and definitions apply.
1) In this document, “he”, “him” and “his” are used in a generic sense, without implying any gender-related distinctions.
© ISO 2012 – All rights reserved 1
---------------------- Page: 5 ----------------------
ISO 24617-2:2012(E)
3.1
addressee
dialogue (3.5) participant (3.13) oriented to by the sender (3.18) in a manner to suggest that his utterances
(3.22) are particularly intended for this participant and that some response is therefore anticipated from this
participant, more so than from the other participants
Note to entry: This definition is a de facto standard in the linguistics literature. It has been slightly modified here, in
replacing “speaker” by “sender” and avoiding the use of ambiguous pronouns. Goffman's original definition says: “dialogue
participant oriented to by the speaker in a manner to suggest that his utterances are particularly intended for him and that
some response is therefore anticipated from him/her, more so than from the other participants”.
[SOURCE: Goffman (1981).]
3.2
allo-feedback act
feedback act (3.8) where the sender (3.18) elicits information about the addressee's (3.1) processing of an
utterance (3.22) that the sender contributed to the dialogue (3.5) or where the sender provides information
about his perceived processing by the addressee of an utterance that the sender contributed to the dialogue
before
EXAMPLE A: Now move up.
B: Slightly northeast you mean?
A: Slightly yeah.
A performs an allo-feedback act signalling that he thinks B understood his first utterance correctly.
3.3
auto-feedback act
feedback act (3.8) where the sender (3.18) provides information about his own processing of an utterance
(3.22) contributed to the dialogue (3.5) by another participant (3.13)
EXAMPLE B's utterance in the example dialogue fragment in (3.2) signals that he is uncertain whether he
understood the previous utterance correctly.
3.4
communicative function
property of certain stretches of communicative behaviour, describing how the behaviour changes the
information state (3.12) of an understander of the behaviour
Note to entry: A communicative function may be “qualified”, i.e. one or more qualifiers (3.14) may be associated with it.
For example, an answer may be qualified as “uncertain” and the acceptance of a request may be “conditional”. See 10.3
for explanation and examples.
3.5
dialogue
exchange of utterances (3.22) between two or more persons or artificial conversational systems
3.6
dialogue act
communicative activity of a dialogue (3.5) participant (3.13), interpreted as having a certain communicative
function (3.4) and semantic content (3.16)
Note to entry: A dialogue act may also have certain functional dependence relations (3.10), rhetorical relations (3.15) and
feedback dependence relations (3.9) with other units in a dialogue (3.5).
3.7
dimension
class of dialogue acts (3.6) that are concerned with a particular aspect of communication, corresponding to a
particular category of semantic content
2 © ISO 2012 – All rights reserved
---------------------- Page: 6 ----------------------
ISO 24617-2:2012(E)
EXAMPLE Dialogue acts advancing the task or activity that motivates the dialogue (the Task dimension), dialogue
acts providing and eliciting feedback (the Auto- and Allo-Feedback dimensions) and dialogue acts for allocating the
speaker role (the Turn Management dimension).
Note to entry: See Clauses 5, 7 and 9 for discussion and more examples.
3.8
feedback act
dialogue act (3.6) which provides or elicits information about the sender's (3.18) or the addressee's (3.1)
processing of something that was uttered in the dialogue
Note to entry: Two classes of feedback are distinguished in this part of ISO 24617: allo-feedback acts (3.2) and auto-
feedback acts (3.3).
3.9
feedback dependence relation
relation between a feedback act (3.8) and the stretch of communicative behaviour whose processing the act
provides or elicits information about
EXAMPLE In the example that accompanies definition 3.2, both the allo-feedback act expressed by utterance 3 and
the auto-feedback act expressed by utterance 2 have a feedback dependence relation to utterance 1.
3.10
functional dependence relation
relation between a given dialogue act (3.6) and a preceding dialogue act on which the semantic content of the
given dialogue act depends due to its communicative function (3.4)
EXAMPLE The relation between an answer and the corresponding question, such as between utterance 3 and
utterance 2 in the example accompanying definition 3.2; or the relation between the acceptance of an offer and the
corresponding offer.
Note to entry: A dialogue act, A2, may also depend on another dialogue act, A1, occurring earlier in a dialogue because
of relations between their semantic contents, e.g. because A2 contains a reference to an element occurring in A1. This is
not a functional dependence relation, since it is not due to A2's communicative function.
3.11
functional segment
minimal stretch of communicative behaviour that has one or more communicative functions (3.4)
EXAMPLE The functional segment corresponding to the answer given by S in the following dialogue fragment does
not include the parts “Just a moment please” and “. let me see.” but only the parts “the first train to the airport on
Sunday morning is” and “at 5:45”:
1. U: What time is the first train to the airport on Sunday morning please?
2. S: Just a moment please. the first train to the airport on Sunday morning is . let me see. at 5:45.
Note 1 to entry: A consequence of this definition is that functional segments may be discontinuous, may overlap or be
embedded and may contain parts contributed by different participants.
Note 2 to entry: The condition of being “minimal” ensures that functional segments do not include material that does
not contribute to the expression of a communicative function that identifies the segment.
3.12
information state
context
totality of a dialogue (3.5) participant's (3.13) beliefs, assumptions, expectations, goals, preferences, hopes
and other attitudes that may influence the participant's interpretation and generation of communicative
behaviour
3.13
participant
person or artificial agent involved in the exchange of utterances (3.22)
© ISO 2012 – All rights reserved 3
---------------------- Page: 7 ----------------------
ISO 24617-2:2012(E)
3.14
qualifier
predicate that can be associated with a communicative function (3.4)
EXAMPLE A: Would you like to have some coffee?
B: Only if you have it ready.
B's utterance accepts A's offer under a certain condition; this can be described by qualifying the communicative function
Accept Offer with the predicate “conditional”.
Note to entry: See 10.3 for more examples.
3.15
rhetorical relation
relation between two dialogue acts (3.6), indicating a pragmatic connection between the two or between their
semantic contents (3.16)
EXAMPLE 1 The statement in the second utterance which follows provides a motivation for the question in the first
utterance:
A: Can you tell me what flights there are to Sydney on Saturday? I'd like to attend my mother's 80th birthday.
EXAMPLE 2 A rhetorical relation between the semantic contents of two dialogue act occurs in the following, where the
content of B's statement mentions a cause for the content of A's statement:
A: I can never find these stupid remote controls
B: That's because they don't have a fixed location
Note to entry: Relations such as elaboration, explanation, justification, cause and concession have been studied
extensively in the analysis of (monologue) text, where they are often called “rhetorical relations” or “discourse relations”
and are mostly viewed either as relations between text segments or as relations between events or propositions,
described in text segments. See, for example, Hovy and Maier, 1992, Lascarides & Asher, 2007 or Mann & Thompson,
1988. Many of these relations also occur in dialogue, either as relations between dialogue acts or between the semantic
contents of dialogue acts.
3.16
semantic content
information, situation, action, event or objects that a stretch of communicative behaviour refers to
3.17
semantic content category
semantic content type
kind of information, situation, action, event or objects that form the semantic content (3.16) of a dialogue act
(3.6)
EXAMPLE The various dimensions (3.7) defined in this part of ISO 24617 correspond to categories of semantic
content. In particular, the Task dimension corresponds to the category of task-specific actions and information; the Allo-
and Auto-Feedback dimensions correspond to the categories of information about the processing by the current speaker
or by the addressee, respectively, of something that was said before; the Turn Management dimension corresponds to the
category of information about the allocation of the speaker role and so forth.
3.18
sender
dialogue (3.5) participant (3.13) who produces a dialogue act (3.6)
3.19
speaker
sender (3.18) of a dialogue act (3.6) in the form of speech, possibly combined with nonverbal communicative
behaviour
Note to entry: A dialogue participant may say something while another participant occupies the speaker role (3.20),
therefore the term “speaker” is not synonymous with “participant who occupies the speaker role”.
4 © ISO 2012 – All rights reserved
---------------------- Page: 8 ----------------------
ISO 24617-2:2012(E)
3.20
speaker role
role occupied by a dialogue (3.5) participant (3.13) who has temporary control of the dialogue and speaks for
some period of time
[SOURCE: DAMSL Revised Manual.]
3.21
turn unit
stretch of communicative activity produced by one participant (3.13) who occupies the speaker role (3.20),
bounded by periods where another participant occupies the speaker role
3.22
utterance
anything said, written, keyed, gesticulated or otherwise expressed
Note to entry: An utterance is mostly a part of what a sender contributes in a turn unit.
4 Purpose and justification
The notion of a dialogue act plays a key role in the analysis of spoken and multimodal dialogue, as well as in
the design of spoken dialogue systems and embodied conversational agents. These activities all depend on
the availability of dialogue corpora, annotated with dialogue act information.
Over the years a variety of dialogue act annotation schemes have been developed, such as those of the
TRAINS human-computer dialogue project (Allen et al., 1994), the Map Task studies of human-human
dialogue (Carletta et al., 1996) and of the Verbmobil speech translation project (Alexandersson et al., 1998).
These schemes were developed for specific purposes and application domains. They contain overlapping sets
of concepts and make use of often mutually inconsistent terminology, sometimes employing different terms for
the same concept or the same term for different concepts.
The multidimensional DIT scheme (Bunt, 1984) was developed for information-seeking dialogues without
depending on a particular domain. The DAMSL scheme (Dialogue Act Markup using Several Layers, Allen
and Core,1997; Core et al., 1998) constitutes an application-independent multidimensional annotation
++
scheme. The DIT scheme (Bunt, 2006; 2009) combines the DIT scheme with concepts from DAMSL and
other more recent schemes into a comprehensive general-purpose annotation scheme.
In the EU-funded project LIRICS (Linguistic Infrastructure for Interoperable Resources and Systems, Romary
++
et al., 2007) a reference set of dialogue acts, taken from the DIT taxonomy, was defined in the form of data
categories, following ISO 12620. This set of concepts has been tested for its usability and coverage a) in the
manual annotation of spoken dialogues in English, Dutch and Italian and b) in the automatic annotation of
spoken and multimodal dialogue in English and forms a significant part of the background of this part of
ISO 24617.
The main purpose of this part of ISO 24617 is to define a reference set of domain-independent basic concepts
for dialogue act annotation, plus a formal language, based on XML, for representing such annotations.
Guidelines are provided for how to use the defined concepts and the annotation language, supported by
extended examples. This formal language, the Dialogue act markup language (DiAML) has a formal
semantics, which makes it possible to apply techniques for automatic reasoning to DiAML annotations.
Guidelines and principles are also provided for extending the set of concepts defined in this part of ISO 24617,
for example, with domain-specific concepts, as well as for selecting coherent subsets.
© ISO 2012 – All rights reserved 5
---------------------- Page: 9 ----------------------
ISO 24617-2:2012(E)
5 Basic concepts and metamodel
The term “dialogue act” is often used rather loosely in the sense of a speech act used in dialogue. Indeed, the
idea of interpreting communicative behaviour in terms of actions, such as questions, promises and requests,
goes back to speech act theory (Austin, 1962; Searle, 1969). But where speech act theory is primarily an
action-based approach to meaning within the philosophy of language, dialogue act theory is an
empirically-based approach to the computational modelling of linguistic and nonverbal communicative
behaviour in dialogue.
Dialogue acts offer a way of characterizing the meaning of communicative behaviour in terms of update
operations, to be applied to the information states of participants in the dialogue; this approach is commonly
known as the “information-state update” or “context-change” approach — see e.g. Bunt (1989; 2000a); Traum
and Larsson (2003). For instance, when an addressee understands the utterance “Do you know what time it
is?” as a question about the time, then the addressee's information state is updated to contain (among other
things) the information that the speaker does not know what time it is and would like to know that. If, by
contrast, it is understood that the speaker is reproaching the addressee for being late, then the addressee's
information state is updated to include (among other things) the information that the speaker does know what
time it is. Distinctions such as that between a question and a reproach concern the communicative function of
a dialogue act, which is one of its two main components. The other main component is its semantic content,
which describes the objects, properties, relations, situations, actions or events that the dialogue act is about.
The communicative function of a dialogue act specifies how an addressee should update his information state
with the information expressed in the semantic content when he understands the dialogue act.
A dialogue act as defined in this part of ISO 24617 (3.6) is a semantic unit of communicative behaviour.
Dialogue act annotation is the marking up of stretches of dialogue with information about the dialogue acts
performed in these stretches and is often limited to assigning communicative function tags. A dialogue act
being a semantic unit in communicative behaviour, the question arises as to which stretches of communicative
behaviour are considered as corresponding to dialogue acts. Spoken dialogues are traditionally segmented
into turns, defined as stretches of communicative behaviour produced by one speaker, bounded by periods of
inactivity of that speaker. Turns in this sense can be quite long and complex and are therefore not very useful
units of behaviour for assigning communicative functions to. Communicative functions can be assigned more
accurately to smaller units, which are called functional segments and which are defined as the minimal
stretches of communicative behaviour that are functionally relevant. See Clause 8 for more details about
dialogue segmentation.
Inherent to the notion of a dialogue act is that there is an agent who produces the dialogue act, called the
“sender” and one or more agents who are addressed, called “addressees”. Dialogue studies often focus on
two-person dialogues, in which case the dialogue acts have only one addressee. Besides sender and
addressee(s), there may be various types of side-participants who are present but do not or only marginally
participate (see Clark, 1996).
Dialogue act annotation is often limited to assigning communicative functions to dialogue segments, which
corresponds intuitively to indicating the type of communicative action that is performed. A semantically more
complete characterization also provides information about the type of semantic content. The DAMSL
annotation scheme distinguishes three categories of semantic content: task, task management and
communication, which indicate whether the semantic content of the dialogue act is concerned with performing
the task which underlies the dialogue or with discussing how to perform the task or with the communication.
++
The DIT scheme distinguishes a number of subcategories of communication-related information, such as
feedback information, turn allocation information and topic progression information. The various categories of
semantic content are also called “dimensions” and are discussed in more detail in Clause 7.
Some types of dialogue acts are inherently dependent for their full meaning on one or more dialogue acts that
occurred earlier in the dialogue. This is, for example, the case for answers, whose meaning is partly
determined by the question being answered and for the acceptance or rejection of offers, suggestions,
requests and apologies. The following example illustrates this, where the meaning of (1.3) clearly depends
very much on whether it is an answer to the question (1.1) or to the question (1.2).
EXAMPLE 1
(1.1) B: Do you know who's coming tonight?
6 © ISO 2012 – All rights reserved
---------------------- Page: 10 ----------------------
ISO 24617-2:2012(E)
(1.2) B: Which of the project members d'you think will be there?
(1.3) A: I'm expecting Jan, Alex, Claudia and David, and maybe Olga and Andrei.
As an answer to (1.1), it says that nobody else is expected to come than the people that are mentioned, but as
an answer to (1.2) it leaves open the possibility that other peop
...
SLOVENSKI STANDARD
SIST ISO 24617-2:2013
01-julij-2013
Upravljanje z jezikovnimi viri - Ogrodje za semantično označevanje (SemAF) - 2.
del: Dialogi
Language resource management -- Semantic annotation framework (SemAF) -- Part 2:
Dialogue acts
Gestion des ressources langagières -- Cadre d'annotation sémantique (SemAF) -- Partie
2: Actes de dialogue
Ta slovenski standard je istoveten z: ISO 24617-2:2012
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 24617-2:2013 en,fr,de
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
SIST ISO 24617-2:2013
---------------------- Page: 2 ----------------------
SIST ISO 24617-2:2013
INTERNATIONAL ISO
STANDARD 24617-2
First edition
2012-09-01
Language resource management —
Semantic annotation framework
(SemAF) —
Part 2:
Dialogue acts
Gestion des ressources langagières — Cadre d'annotation sémantique
(SemAF) —
Partie 2: Actes de dialogue
Reference number
ISO 24617-2:2012(E)
©
ISO 2012
---------------------- Page: 3 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2012
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing 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 2012 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
Contents Page
Foreword . iv
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 1
4 Purpose and justification . 5
5 Basic concepts and metamodel . 6
6 Definition of communicative functions . 8
7 Annotation schemes . 9
7.1 Structure of annotation schemes . 9
7.2 Multidimensionality and multifunctionality . 10
7.3 Multidimensionality, clustering and dimensions . 11
7.4 Dimension- specific and general-purpose functions . 11
8 Dialogue segmentation . 13
9 Dimensions . 14
9.1 Task. 15
9.2 Auto-Feedback . 15
9.3 Allo-Feedback . 15
9.4 Turn Management . 15
9.5 Time Management . 16
9.6 Discourse Structuring . 16
9.7 Social Obligations Management . 16
9.8 Own Communication Management . 16
9.9 Partner Communication Management . 16
10 Core dialogue acts . 17
10.1 General-purpose functions . 19
10.2 Dimension-specific functions . 20
10.3 Function qualifiers . 22
11 Dialogue act markup language (DiAML) . 23
11.1 Abstract syntax . 23
11.2 Concrete syntax . 24
12 Principles for extending and restricting the standard . 25
12.1 Main design principles . 25
12.2 Schema extension . 27
12.3 Scheme restriction . 27
Annex A (informative) Annotation guidelines . 29
Annex B (informative) Annotated dialogue examples . 43
Annex C (normative) Formal definition of DiAML . 56
Annex D (normative) DiAML technical schema . 63
Annex E (normative) Data categories for core concepts . 68
Annex F (informative) Examples of possible additional data categories . 88
Annex G (informative) Concepts in existing schemes . 90
Bibliography . 100
© ISO 2012 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(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.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
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.
ISO 24617-2 was prepared by Technical Committee 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:
Part 1: Time and events (SemAF-Time, ISO-TimeML)
Part 2: Dialogue acts
The following parts are under preparation:
Part 3: Named entities (SemAF-NE)
Part 4: Semantic roles (SemAF-SRL)
Part 5: Discourse structure (SemAF-DS)
Part 6: Principles of semantic annotation (SemAF-Basics)
Part 7: Spatial information (ISO-Space)
Part 8: Semantic relations in discourse (SemAF-DRel)
iv © ISO 2012 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24617-2:2013
INTERNATIONAL STANDARD ISO 24617-2:2012(E)
Language resource management — Semantic annotation
framework (SemAF) —
Part 2:
Dialogue acts
1 Scope
This part of ISO 24617 provides a set of empirically and theoretically well-motivated concepts for dialogue
annotation, a formal language for expressing dialogue annotations — the dialogue act markup language
(DiAML) — and a method for segmenting a dialogue into semantic units. This allows the manual or automatic
annotation of dialogue segments with information about the communicative actions which the participants
perform by their contributions to the dialogue. It supports multidimensional annotation, in which units in
dialogue are viewed as having multiple communicative functions. The DiAML language has an XML-based
representation format and a formal semantics which makes it possible to apply inference to DiAML
representations.
This part of ISO 24617 specifies data categories for reference sets of communicative functions and
dimensions of dialogue analysis and provides principles and guidelines for extending these sets or selecting
coherent subsets of them. Additionally, it provides guidelines for annotators and annotated examples. It is
applicable to spoken, written and multimodal dialogues involving two or more participants.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 12620:2009, Terminology and other language resources — Specification of data categories and
management of a Data Category Registry for language resources
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO 24612:2011, Language resource management — Linguistic annotation framework
3 Terms and definitions
1)
For the purposes of this document, the following terms and definitions apply.
1) In this document, “he”, “him” and “his” are used in a generic sense, without implying any gender-related distinctions.
© ISO 2012 – All rights reserved 1
---------------------- Page: 7 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.1
addressee
dialogue (3.5) participant (3.13) oriented to by the sender (3.18) in a manner to suggest that his utterances
(3.22) are particularly intended for this participant and that some response is therefore anticipated from this
participant, more so than from the other participants
Note to entry: This definition is a de facto standard in the linguistics literature. It has been slightly modified here, in
replacing “speaker” by “sender” and avoiding the use of ambiguous pronouns. Goffman's original definition says: “dialogue
participant oriented to by the speaker in a manner to suggest that his utterances are particularly intended for him and that
some response is therefore anticipated from him/her, more so than from the other participants”.
[SOURCE: Goffman (1981).]
3.2
allo-feedback act
feedback act (3.8) where the sender (3.18) elicits information about the addressee's (3.1) processing of an
utterance (3.22) that the sender contributed to the dialogue (3.5) or where the sender provides information
about his perceived processing by the addressee of an utterance that the sender contributed to the dialogue
before
EXAMPLE A: Now move up.
B: Slightly northeast you mean?
A: Slightly yeah.
A performs an allo-feedback act signalling that he thinks B understood his first utterance correctly.
3.3
auto-feedback act
feedback act (3.8) where the sender (3.18) provides information about his own processing of an utterance
(3.22) contributed to the dialogue (3.5) by another participant (3.13)
EXAMPLE B's utterance in the example dialogue fragment in (3.2) signals that he is uncertain whether he
understood the previous utterance correctly.
3.4
communicative function
property of certain stretches of communicative behaviour, describing how the behaviour changes the
information state (3.12) of an understander of the behaviour
Note to entry: A communicative function may be “qualified”, i.e. one or more qualifiers (3.14) may be associated with it.
For example, an answer may be qualified as “uncertain” and the acceptance of a request may be “conditional”. See 10.3
for explanation and examples.
3.5
dialogue
exchange of utterances (3.22) between two or more persons or artificial conversational systems
3.6
dialogue act
communicative activity of a dialogue (3.5) participant (3.13), interpreted as having a certain communicative
function (3.4) and semantic content (3.16)
Note to entry: A dialogue act may also have certain functional dependence relations (3.10), rhetorical relations (3.15) and
feedback dependence relations (3.9) with other units in a dialogue (3.5).
3.7
dimension
class of dialogue acts (3.6) that are concerned with a particular aspect of communication, corresponding to a
particular category of semantic content
2 © ISO 2012 – All rights reserved
---------------------- Page: 8 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
EXAMPLE Dialogue acts advancing the task or activity that motivates the dialogue (the Task dimension), dialogue
acts providing and eliciting feedback (the Auto- and Allo-Feedback dimensions) and dialogue acts for allocating the
speaker role (the Turn Management dimension).
Note to entry: See Clauses 5, 7 and 9 for discussion and more examples.
3.8
feedback act
dialogue act (3.6) which provides or elicits information about the sender's (3.18) or the addressee's (3.1)
processing of something that was uttered in the dialogue
Note to entry: Two classes of feedback are distinguished in this part of ISO 24617: allo-feedback acts (3.2) and auto-
feedback acts (3.3).
3.9
feedback dependence relation
relation between a feedback act (3.8) and the stretch of communicative behaviour whose processing the act
provides or elicits information about
EXAMPLE In the example that accompanies definition 3.2, both the allo-feedback act expressed by utterance 3 and
the auto-feedback act expressed by utterance 2 have a feedback dependence relation to utterance 1.
3.10
functional dependence relation
relation between a given dialogue act (3.6) and a preceding dialogue act on which the semantic content of the
given dialogue act depends due to its communicative function (3.4)
EXAMPLE The relation between an answer and the corresponding question, such as between utterance 3 and
utterance 2 in the example accompanying definition 3.2; or the relation between the acceptance of an offer and the
corresponding offer.
Note to entry: A dialogue act, A2, may also depend on another dialogue act, A1, occurring earlier in a dialogue because
of relations between their semantic contents, e.g. because A2 contains a reference to an element occurring in A1. This is
not a functional dependence relation, since it is not due to A2's communicative function.
3.11
functional segment
minimal stretch of communicative behaviour that has one or more communicative functions (3.4)
EXAMPLE The functional segment corresponding to the answer given by S in the following dialogue fragment does
not include the parts “Just a moment please” and “. let me see.” but only the parts “the first train to the airport on
Sunday morning is” and “at 5:45”:
1. U: What time is the first train to the airport on Sunday morning please?
2. S: Just a moment please. the first train to the airport on Sunday morning is . let me see. at 5:45.
Note 1 to entry: A consequence of this definition is that functional segments may be discontinuous, may overlap or be
embedded and may contain parts contributed by different participants.
Note 2 to entry: The condition of being “minimal” ensures that functional segments do not include material that does
not contribute to the expression of a communicative function that identifies the segment.
3.12
information state
context
totality of a dialogue (3.5) participant's (3.13) beliefs, assumptions, expectations, goals, preferences, hopes
and other attitudes that may influence the participant's interpretation and generation of communicative
behaviour
3.13
participant
person or artificial agent involved in the exchange of utterances (3.22)
© ISO 2012 – All rights reserved 3
---------------------- Page: 9 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.14
qualifier
predicate that can be associated with a communicative function (3.4)
EXAMPLE A: Would you like to have some coffee?
B: Only if you have it ready.
B's utterance accepts A's offer under a certain condition; this can be described by qualifying the communicative function
Accept Offer with the predicate “conditional”.
Note to entry: See 10.3 for more examples.
3.15
rhetorical relation
relation between two dialogue acts (3.6), indicating a pragmatic connection between the two or between their
semantic contents (3.16)
EXAMPLE 1 The statement in the second utterance which follows provides a motivation for the question in the first
utterance:
A: Can you tell me what flights there are to Sydney on Saturday? I'd like to attend my mother's 80th birthday.
EXAMPLE 2 A rhetorical relation between the semantic contents of two dialogue act occurs in the following, where the
content of B's statement mentions a cause for the content of A's statement:
A: I can never find these stupid remote controls
B: That's because they don't have a fixed location
Note to entry: Relations such as elaboration, explanation, justification, cause and concession have been studied
extensively in the analysis of (monologue) text, where they are often called “rhetorical relations” or “discourse relations”
and are mostly viewed either as relations between text segments or as relations between events or propositions,
described in text segments. See, for example, Hovy and Maier, 1992, Lascarides & Asher, 2007 or Mann & Thompson,
1988. Many of these relations also occur in dialogue, either as relations between dialogue acts or between the semantic
contents of dialogue acts.
3.16
semantic content
information, situation, action, event or objects that a stretch of communicative behaviour refers to
3.17
semantic content category
semantic content type
kind of information, situation, action, event or objects that form the semantic content (3.16) of a dialogue act
(3.6)
EXAMPLE The various dimensions (3.7) defined in this part of ISO 24617 correspond to categories of semantic
content. In particular, the Task dimension corresponds to the category of task-specific actions and information; the Allo-
and Auto-Feedback dimensions correspond to the categories of information about the processing by the current speaker
or by the addressee, respectively, of something that was said before; the Turn Management dimension corresponds to the
category of information about the allocation of the speaker role and so forth.
3.18
sender
dialogue (3.5) participant (3.13) who produces a dialogue act (3.6)
3.19
speaker
sender (3.18) of a dialogue act (3.6) in the form of speech, possibly combined with nonverbal communicative
behaviour
Note to entry: A dialogue participant may say something while another participant occupies the speaker role (3.20),
therefore the term “speaker” is not synonymous with “participant who occupies the speaker role”.
4 © ISO 2012 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
3.20
speaker role
role occupied by a dialogue (3.5) participant (3.13) who has temporary control of the dialogue and speaks for
some period of time
[SOURCE: DAMSL Revised Manual.]
3.21
turn unit
stretch of communicative activity produced by one participant (3.13) who occupies the speaker role (3.20),
bounded by periods where another participant occupies the speaker role
3.22
utterance
anything said, written, keyed, gesticulated or otherwise expressed
Note to entry: An utterance is mostly a part of what a sender contributes in a turn unit.
4 Purpose and justification
The notion of a dialogue act plays a key role in the analysis of spoken and multimodal dialogue, as well as in
the design of spoken dialogue systems and embodied conversational agents. These activities all depend on
the availability of dialogue corpora, annotated with dialogue act information.
Over the years a variety of dialogue act annotation schemes have been developed, such as those of the
TRAINS human-computer dialogue project (Allen et al., 1994), the Map Task studies of human-human
dialogue (Carletta et al., 1996) and of the Verbmobil speech translation project (Alexandersson et al., 1998).
These schemes were developed for specific purposes and application domains. They contain overlapping sets
of concepts and make use of often mutually inconsistent terminology, sometimes employing different terms for
the same concept or the same term for different concepts.
The multidimensional DIT scheme (Bunt, 1984) was developed for information-seeking dialogues without
depending on a particular domain. The DAMSL scheme (Dialogue Act Markup using Several Layers, Allen
and Core,1997; Core et al., 1998) constitutes an application-independent multidimensional annotation
++
scheme. The DIT scheme (Bunt, 2006; 2009) combines the DIT scheme with concepts from DAMSL and
other more recent schemes into a comprehensive general-purpose annotation scheme.
In the EU-funded project LIRICS (Linguistic Infrastructure for Interoperable Resources and Systems, Romary
++
et al., 2007) a reference set of dialogue acts, taken from the DIT taxonomy, was defined in the form of data
categories, following ISO 12620. This set of concepts has been tested for its usability and coverage a) in the
manual annotation of spoken dialogues in English, Dutch and Italian and b) in the automatic annotation of
spoken and multimodal dialogue in English and forms a significant part of the background of this part of
ISO 24617.
The main purpose of this part of ISO 24617 is to define a reference set of domain-independent basic concepts
for dialogue act annotation, plus a formal language, based on XML, for representing such annotations.
Guidelines are provided for how to use the defined concepts and the annotation language, supported by
extended examples. This formal language, the Dialogue act markup language (DiAML) has a formal
semantics, which makes it possible to apply techniques for automatic reasoning to DiAML annotations.
Guidelines and principles are also provided for extending the set of concepts defined in this part of ISO 24617,
for example, with domain-specific concepts, as well as for selecting coherent subsets.
© ISO 2012 – All rights reserved 5
---------------------- Page: 11 ----------------------
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
5 Basic concepts and metamodel
The term “dialogue act” is often used rather loosely in the sense of a speech act used in dialogue. Indeed, the
idea of interpreting communicative behaviour in terms of actions, such as questions, promises and requests,
goes back to speech act theory (Austin, 1962; Searle, 1969). But where speech act theory is primarily an
action-based approach to meaning within the philosophy of language, dialogue act theory is an
empirically-based approach to the computational modelling of linguistic and nonverbal communicative
behaviour in dialogue.
Dialogue acts offer a way of characterizing the meaning of communicative behaviour in terms of update
operations, to be applied to the information states of participants in the dialogue; this approach is commonly
known as the “information-state update” or “context-change” approach — see e.g. Bunt (1989; 2000a); Traum
and Larsson (2003). For instance, when an addressee understands the utterance “Do you know what time it
is?” as a question about the time, then the addressee's information state is updated to contain (among other
things) the information that the speaker does not know what time it is and would like to know that. If, by
contrast, it is understood that the speaker is reproaching the addressee for being late, then the addressee's
information state is updated to include (among other things) the information that the speaker does know what
time it is. Distinctions such as that between a question and a reproach concern the communicative function of
a dialogue act, which is one of its two main components. The other main component is its semantic content,
which describes the objects, properties, relations, situations, actions or events that the dialogue act is about.
The communicative function of a dialogue act specifies how an addressee should update his information state
with the information expressed in the semantic content when he understands the dialogue act.
A dialogue act as defined in this part of ISO 24617 (3.6) is a semantic unit of communicative behaviour.
Dialogue act annotation is the marking up of stretches of dialogue with information about the dialogue acts
performed in these stretches and is often limited to assigning communicative function tags. A dialogue act
being a semantic unit in communicative behaviour, the question arises as to which stretches of communicative
behaviour are considered as corresponding to dialogue acts. Spoken dialogues are traditionally segmented
into turns, defined as stretches of communicative behaviour produced by one speaker, bounded by periods of
inactivity of that speaker. Turns in this sense can be quite long and complex and are therefore not very useful
units of behaviour for assigning communicative functions to. Communicative functions can be assigned more
accurately to smaller units, which are called functional segments and which are defined as the minimal
stretches of communicative behaviour that are functionally relevant. See Clause 8 for more details about
dialogue segmentation.
Inherent to the notion of a dialogue act is that there is an agent who produces the dialogue act, called the
“sender” and one or more agents who are addressed, called “addressees”. Dialogue studies often focus on
two-person dialogues, in which case the dialogue acts have only one addressee. Besides sender and
addressee(s), there may be various types of side-participants who are present but do not or only marginally
participate (see Clark, 1996).
Dialogue act annotation is often limited to assigning communicative functions to dialogue segments, which
corresponds intuitively to indicating the type of communicative action that is performed. A semantically more
complete characterization also provides information about the type of semantic content. The DAMSL
annotation scheme distinguishes three categories of semantic content: task, task management and
communication, which indicate whether the semantic content of the dialogue act is concerned with performing
the task which underlies the dialogue or with discussing how to perform the task or with the communication.
++
The DIT scheme distinguishes a number of subcategories of communication-related information, such as
feedback information, turn allocation information and topic progre
...
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.Gestion des ressources langagières -- Cadre d'annotation sémantique (SemAF) -- Partie 2: Actes de dialogueLanguage resource management -- Semantic annotation framework (SemAF) -- Part 2: Dialogue acts01.020NRRUGLQDFLMDTerminology (principles and coordination)ICS:Ta slovenski standard je istoveten z:ISO 24617-2:2012SIST ISO 24617-2:2013en,fr,de01-julij-2013SIST ISO 24617-2:2013SLOVENSKI
STANDARD
SIST ISO 24617-2:2013
Reference numberISO 24617-2:2012(E)© ISO 2012
INTERNATIONAL STANDARD ISO24617-2First edition2012-09-01Language resource management — Semantic annotation framework (SemAF) — Part 2: Dialogue acts Gestion des ressources langagières — Cadre d'annotation sémantique (SemAF) — Partie 2: Actes de dialogue
SIST ISO 24617-2:2013
ISO 24617-2:2012(E)
COPYRIGHT PROTECTED DOCUMENT
©
ISO 2012 All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing 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 2012 – All rights reserved
SIST ISO 24617-2:2013
ISO 24617-2:2012(E) © ISO 2012 – All rights reserved
iii Contents Page Foreword . iv 1 Scope . 1 2 Normative references . 1 3 Terms and definitions . 1 4 Purpose and justification . 5 5 Basic concepts and metamodel . 6 6 Definition of communicative functions . 8 7 Annotation schemes . 9 7.1 Structure of annotation schemes . 9 7.2 Multidimensionality and multifunctionality . 10 7.3 Multidimensionality, clustering and dimensions . 11 7.4 Dimension- specific and general-purpose functions . 11 8 Dialogue segmentation . 13 9 Dimensions . 14 9.1 Task. 15 9.2 Auto-Feedback . 15 9.3 Allo-Feedback . 15 9.4 Turn Management . 15 9.5 Time Management . 16 9.6 Discourse Structuring . 16 9.7 Social Obligations Management . 16 9.8 Own Communication Management . 16 9.9 Partner Communication Management . 16 10 Core dialogue acts . 17 10.1 General-purpose functions . 19 10.2 Dimension-specific functions . 20 10.3 Function qualifiers . 22 11 Dialogue act markup language (DiAML) . 23 11.1 Abstract syntax . 23 11.2 Concrete syntax . 24 12 Principles for extending and restricting the standard . 25 12.1 Main design principles . 25 12.2 Schema extension . 27 12.3 Scheme restriction . 27 Annex A (informative)
Annotation guidelines . 29 Annex B (informative)
Annotated dialogue examples . 43 Annex C (normative)
Formal definition of DiAML . 56 Annex D (normative)
DiAML technical schema . 63 Annex E (normative)
Data categories for core concepts . 68 Annex F (informative)
Examples of possible additional data categories . 88 Annex G (informative)
Concepts in existing schemes . 90 Bibliography . 100 SIST ISO 24617-2:2013
ISO 24617-2:2012(E) iv
© ISO 2012 – All rights reserved 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. International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2. The main task of technical committees is to prepare International Standards. Draft International Standards adopted by the technical committees are circulated to the member bodies for voting. Publication as an International Standard requires approval by at least 75 % of the member bodies casting a vote. 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. ISO 24617-2 was prepared by Technical Committee 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: Part 1: Time and events (SemAF-Time, ISO-TimeML) Part 2: Dialogue acts The following parts are under preparation: Part 3: Named entities (SemAF-NE) Part 4: Semantic roles (SemAF-SRL) Part 5: Discourse structure (SemAF-DS) Part 6: Principles of semantic annotation (SemAF-Basics) Part 7: Spatial information (ISO-Space) Part 8: Semantic relations in discourse (SemAF-DRel)
SIST ISO 24617-2:2013
INTERNATIONAL STANDARD ISO 24617-2:2012(E) © ISO 2012 – All rights reserved 1
Language resource management — Semantic annotation framework (SemAF) — Part 2: Dialogue acts 1 Scope This part of ISO 24617 provides a set of empirically and theoretically well-motivated concepts for dialogue annotation, a formal language for expressing dialogue annotations — the dialogue act markup language (DiAML) — and a method for segmenting a dialogue into semantic units. This allows the manual or automatic annotation of dialogue segments with information about the communicative actions which the participants perform by their contributions to the dialogue. It supports multidimensional annotation, in which units in dialogue are viewed as having multiple communicative functions. The DiAML language has an XML-based representation format and a formal semantics which makes it possible to apply inference to DiAML representations.
This part of ISO 24617 specifies data categories for reference sets of communicative functions and dimensions of dialogue analysis and provides principles and guidelines for extending these sets or selecting coherent subsets of them. Additionally, it provides guidelines for annotators and annotated examples. It is applicable to spoken, written and multimodal dialogues involving two or more participants. 2 Normative references The following referenced documents are indispensable for the application of this document. For dated references, only the edition cited applies. For undated references, the latest edition of the referenced document (including any amendments) applies. ISO 12620:2009, Terminology and other language resources — Specification of data categories and management of a Data Category Registry for language resources ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure representation ISO 24612:2011, Language resource management — Linguistic annotation framework 3 Terms and definitions For the purposes of this document, the following terms and definitions apply.1) 1) In this document, “he”, “him” and “his” are used in a generic sense, without implying any gender-related distinctions. SIST ISO 24617-2:2013
ISO 24617-2:2012(E) 2
© ISO 2012 – All rights reserved 3.1 addressee dialogue (3.5) participant (3.13) oriented to by the sender (3.18) in a manner to suggest that his utterances (3.22) are particularly intended for this participant and that some response is therefore anticipated from this participant, more so than from the other participants Note to entry: This definition is a de facto standard in the linguistics literature. It has been slightly modified here, in replacing “speaker” by “sender” and avoiding the use of ambiguous pronouns. Goffman's original definition says: “dialogue participant oriented to by the speaker in a manner to suggest that his utterances are particularly intended for him and that some response is therefore anticipated from him/her, more so than from the other participants”. [SOURCE: Goffman (1981).]
3.2 allo-feedback act feedback act (3.8) where the sender (3.18) elicits information about the addressee's (3.1) processing of an utterance (3.22) that the sender contributed to the dialogue (3.5) or where the sender provides information about his perceived processing by the addressee of an utterance that the sender contributed to the dialogue before EXAMPLE A: Now move up.
B: Slightly northeast you mean?
A: Slightly yeah.
A performs an allo-feedback act signalling that he thinks B understood his first utterance correctly. 3.3 auto-feedback act feedback act (3.8) where the sender (3.18) provides information about his own processing of an utterance (3.22) contributed to the dialogue (3.5) by another participant (3.13) EXAMPLE B's utterance in the example dialogue fragment in (3.2) signals that he is uncertain whether he understood the previous utterance correctly. 3.4 communicative function property of certain stretches of communicative behaviour, describing how the behaviour changes the information state (3.12) of an understander of the behaviour Note to entry: A communicative function may be “qualified”, i.e. one or more qualifiers (3.14) may be associated with it. For example, an answer may be qualified as “uncertain” and the acceptance of a request may be “conditional”. See 10.3 for explanation and examples. 3.5 dialogue exchange of utterances (3.22) between two or more persons or artificial conversational systems 3.6 dialogue act communicative activity of a dialogue (3.5) participant (3.13), interpreted as having a certain communicative function (3.4) and semantic content (3.16) Note to entry: A dialogue act may also have certain functional dependence relations (3.10), rhetorical relations (3.15) and feedback dependence relations (3.9) with other units in a dialogue (3.5). 3.7 dimension class of dialogue acts (3.6) that are concerned with a particular aspect of communication, corresponding to a particular category of semantic content SIST ISO 24617-2:2013
ISO 24617-2:2012(E) © ISO 2012 – All rights reserved 3 EXAMPLE Dialogue acts advancing the task or activity that motivates the dialogue (the Task dimension), dialogue acts providing and eliciting feedback (the Auto- and Allo-Feedback dimensions) and dialogue acts for allocating the speaker role (the Turn Management dimension). Note to entry: See Clauses 5, 7 and 9 for discussion and more examples. 3.8 feedback act dialogue act (3.6) which provides or elicits information about the sender's (3.18) or the addressee's (3.1) processing of something that was uttered in the dialogue Note to entry: Two classes of feedback are distinguished in this part of ISO 24617: allo-feedback acts (3.2) and auto-feedback acts (3.3). 3.9 feedback dependence relation relation between a feedback act (3.8) and the stretch of communicative behaviour whose processing the act provides or elicits information about EXAMPLE In the example that accompanies definition 3.2, both the allo-feedback act expressed by utterance 3 and the auto-feedback act expressed by utterance 2 have a feedback dependence relation to utterance 1. 3.10 functional dependence relation relation between a given dialogue act (3.6) and a preceding dialogue act on which the semantic content of the given dialogue act depends due to its communicative function (3.4) EXAMPLE The relation between an answer and the corresponding question, such as between utterance 3 and utterance 2 in the example accompanying definition 3.2; or the relation between the acceptance of an offer and the corresponding offer. Note to entry: A dialogue act, A2, may also depend on another dialogue act, A1, occurring earlier in a dialogue because of relations between their semantic contents, e.g. because A2 contains a reference to an element occurring in A1. This is not a functional dependence relation, since it is not due to A2's communicative function. 3.11 functional segment minimal stretch of communicative behaviour that has one or more communicative functions (3.4) EXAMPLE The functional segment corresponding to the answer given by S in the following dialogue fragment does not include the parts “Just a moment please” and
“. let me see.” but only the parts “the first train to the airport on Sunday morning is” and “at 5:45”: 1. U: What time is the first train to the airport on Sunday morning please? 2. S: Just a moment please. the first train to the airport on Sunday morning is . let me see. at 5:45. Note 1 to entry:
A consequence of this definition is that functional segments may be discontinuous, may overlap or be embedded and may contain parts contributed by different participants. Note 2 to entry: The condition of being “minimal” ensures that functional segments do not include material that does not contribute to the expression of a communicative function that identifies the segment. 3.12 information state context totality of a dialogue (3.5) participant's (3.13) beliefs, assumptions, expectations, goals, preferences, hopes and other attitudes that may influence the participant's interpretation and generation of communicative behaviour 3.13 participant person or artificial agent involved in the exchange of utterances (3.22) SIST ISO 24617-2:2013
ISO 24617-2:2012(E) 4
© ISO 2012 – All rights reserved 3.14 qualifier predicate that can be associated with a communicative function (3.4) EXAMPLE A:
Would you like to have some coffee? B:
Only if you have it ready. B's utterance accepts A's offer under a certain condition; this can be described by qualifying the communicative function Accept Offer with the predicate “conditional”.
Note to entry: See 10.3 for more examples. 3.15 rhetorical relation relation between two dialogue acts (3.6), indicating a pragmatic connection between the two or between their semantic contents (3.16) EXAMPLE 1 The statement in the second utterance which follows provides a motivation for the question in the first utterance: A:
Can you tell me what flights there are to Sydney on Saturday? I'd like to attend my mother's 80th birthday. EXAMPLE 2 A rhetorical relation between the semantic contents of two dialogue act occurs in the following, where the content of B's statement mentions a cause for the content of A's statement: A:
I can never find these stupid remote controls
B:
That's because they don't have a fixed location Note to entry: Relations such as elaboration, explanation, justification, cause and concession have been studied extensively in the analysis of (monologue) text, where they are often called “rhetorical relations” or “discourse relations” and are mostly viewed either as relations between text segments or as relations between events or propositions, described in text segments. See, for example, Hovy and Maier, 1992, Lascarides & Asher, 2007 or Mann & Thompson, 1988. Many of these relations also occur in dialogue, either as relations between dialogue acts or between the semantic contents of dialogue acts. 3.16 semantic content information, situation, action, event or objects that a stretch of communicative behaviour refers to 3.17 semantic content category semantic content type kind of information, situation, action, event or objects that form the semantic content (3.16) of a dialogue act (3.6) EXAMPLE The various dimensions (3.7) defined in this part of ISO 24617 correspond to categories of semantic content. In particular, the Task dimension corresponds to the category of task-specific actions and information; the Allo- and Auto-Feedback dimensions correspond to the categories of information about the processing by the current speaker or by the addressee, respectively, of something that was said before; the Turn Management dimension corresponds to the category of information about the allocation of the speaker role and so forth. 3.18 sender dialogue (3.5) participant (3.13) who produces a dialogue act (3.6) 3.19 speaker sender (3.18) of a dialogue act (3.6) in the form of speech, possibly combined with nonverbal communicative behaviour Note to entry: A dialogue participant may say something while another participant occupies the speaker role (3.20), therefore the term “speaker” is not synonymous with “participant who occupies the speaker role”. SIST ISO 24617-2:2013
ISO 24617-2:2012(E) © ISO 2012 – All rights reserved 5 3.20 speaker role role occupied by a dialogue (3.5) participant (3.13) who has temporary control of the dialogue and speaks for some period of time [SOURCE: DAMSL Revised Manual.] 3.21 turn unit stretch of communicative activity produced by one participant (3.13) who occupies the speaker role (3.20), bounded by periods where another participant occupies the speaker role 3.22 utterance anything said, written, keyed, gesticulated or otherwise expressed Note to entry: An utterance is mostly a part of what a sender contributes in a turn unit. 4 Purpose and justification The notion of a dialogue act plays a key role in the analysis of spoken and multimodal dialogue, as well as in the design of spoken dialogue systems and embodied conversational agents. These activities all depend on the availability of dialogue corpora, annotated with dialogue act information. Over the years a variety of dialogue act annotation schemes have been developed, such as those of the TRAINS human-computer dialogue project (Allen et al., 1994), the Map Task studies of human-human dialogue (Carletta et al., 1996) and of the Verbmobil speech translation project (Alexandersson et al., 1998). These schemes were developed for specific purposes and application domains. They contain overlapping sets of concepts and make use of often mutually inconsistent terminology, sometimes employing different terms for the same concept or the same term for different concepts. The multidimensional DIT scheme (Bunt, 1984) was developed for information-seeking dialogues without depending on a particular domain. The DAMSL scheme (Dialogue Act Markup using Several Layers, Allen and Core,1997; Core et al., 1998) constitutes an application-independent multidimensional annotation scheme. The DIT++ scheme (Bunt, 2006; 2009) combines the DIT scheme with concepts from DAMSL and other more recent schemes into a comprehensive general-purpose annotation scheme. In the EU-funded project LIRICS (Linguistic Infrastructure for Interoperable Resources and Systems, Romary et al., 2007) a reference set of dialogue acts, taken from the DIT++ taxonomy, was defined in the form of data categories, following ISO 12620. This set of concepts has been tested for its usability and coverage a) in the manual annotation of spoken dialogues in English, Dutch and Italian and b) in the automatic annotation of spoken and multimodal dialogue in English and forms a significant part of the background of this part of ISO 24617. The main purpose of this part of ISO 24617 is to define a reference set of domain-independent basic concepts for dialogue act annotation, plus a formal language, based on XML, for representing such annotations. Guidelines are provided for how to use the defined concepts and the annotation language, supported by extended examples. This formal language, the Dialogue act markup language (DiAML) has a formal semantics, which makes it possible to apply techniques for automatic reasoning to DiAML annotations. Guidelines and principles are also provided for extending the set of concepts defined in this part of ISO 24617, for example, with domain-specific concepts, as well as for selecting coherent subsets. SIST ISO 24617-2:2013
ISO 24617-2:2012(E) 6
© ISO 2012 – All rights reserved 5 Basic concepts and metamodel The term “dialogue act” is often used rather loosely in the sense of a speech act used in dialogue. Indeed, the idea of interpreting communicative behaviour in terms of actions, such as questions, promises and requests, goes back to speech act theory (Austin, 1962; Searle, 1969). But where speech act theory is primarily an action-based approach to meaning within the philosophy of language, dialogue act theory is an empirically-based approach to the computational modelling of linguistic and nonverbal communicative behaviour in dialogue. Dialogue acts offer a way of characterizing the meaning of communicative behaviour in terms of update operations, to be applied to the information states of participants in the dialogue; this approach is commonly known as the “information-state update” or “context-change” approach — see e.g. Bunt (1989; 2000a); Traum and Larsson (2003). For instance, when an addressee understands the utterance “Do you know what time it is?” as a question about the time, then the addressee's information state is updated to contain (among other things) the information that the speaker does not know what time it is and would like to know that. If, by contrast, it is understood that the speaker is reproaching the addressee for being late, then the addressee's information state is updated to include (among other things) the information that the speaker does know what time it is. Distinctions such as that between a question and a reproach concern the communicative function of a dialogue act, which is one of its two main components. The other main component is its semantic content, which describes the objects, properties, relations, situations, actions or events that the dialogue act is about. The communicative function of a dialogue act specifies how an addressee should update his information state with the information expressed in the semantic content when he understands the dialogue act. A dialogue act as defined in this part of ISO 24617 (3.6) is a semantic unit of communicative behaviour. Dialogue act annotation is the marking up of stretches of dialogue with information about the dialogue acts performed in these stretches and is often limited to assigning communicative function tags. A dialogue act being a semantic unit in communicative behaviour, the question arises as to which stretches of communicative behaviour are considered as corresponding to dialogue acts. Spoken dialogues are traditionally segmented into turns, defined as stretches of communicative behaviour produced by one speaker, bounded by periods of inactivity of that speaker. Turns in this sense can be quite long and complex and are therefore not very useful units of behaviour for assigning communicative functions to. Communicative functions can be assigned more accurately to smaller units, which are called functional segments and which are defined as the minimal stretches of communicative behaviour that are functionally relevant. See Clause 8 for more details about dialogue segmentation. Inherent to the notion of a dialogue act is that there is an agent who produces the dialogue act, called the “sender” and one or more agents who are addressed, called “addressees”. Dialogue studies often focus on two-person dialogues, in which case the dialogue acts have only one addressee. Besides sender and addressee(s), there may be various types of side-participants who are present but do not or only marginally participate (see Clark, 1996). Dialogue act annotation is often limited to assigning communicative functions to dialogue segments, which corresponds intuitively to indicating the type of communicative action that is performed. A semantically more complete characterization also provides information about the type of semantic content. The DAMSL annotation scheme distinguishes three categories of semantic content: task, task management and communication, which indicate whether the semantic content of the dialogue act is concerned with performing the task which underlies the dialogue or with discussing how to perform the task or with the communication. The DIT++ scheme distinguishes a number of subcategories of communication-related information, such as feedback information, turn allocation information and topic progression information. The various categories of semantic content are also called “dimensions” and are discussed in more detail in Clause 7. Some types of dialogue acts are inherently dependent for their full meaning on one or more dialogue acts that occurred earlier in the dialogue. This is, for example, the case for answers, whose meaning is partly determined by the question being answered and for the acceptance or rejection of offers, suggestions, requests and apologies. The following example illustrates this, where the meaning of (1.3) clearly depends very much on whether it is an answer to the question (1.1) or to the question (1.2). EXAMPLE 1
(1.1) B: Do you know who's coming tonight? SIST ISO 24617-2:2013
ISO 24617-2:2012(E) © ISO 2012 – All rights reserved 7
(1.2) B: Which of the project members d'you think will be there? (1.3) A: I'm expecting Jan, Alex, Claudia and David, and maybe Olga and Andrei. As an answer to (1.1), it says that nobody else is expected to come than the people that are mentioned, but as an answer to (1.2) it leaves open the possibility that other people will come, who are not members of “the project”. For dialogue acts which have such a dependence on other dialogue acts, due to their responsive character, the marking up of the links to these “antecedent” dialogue acts allows the annotation not just to express e.g. that the utter
...
Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.