ISO 24617-2:2020

Overview

ISO 24617-2:2020 (ISO 24617-2:2020) defines a standardized framework for annotating the communicative actions participants perform in dialogues. Part of the SemAF (Semantic Annotation Framework) series, it specifies an empirically grounded set of dialogue act concepts, a formal annotation language - the Dialogue Act Markup Language (DiAML) - and a method for segmenting dialogues into semantic units. The standard supports multidimensional and multifunctional annotation of spoken, written and multimodal dialogues and provides an XML-based representation with formal semantics to enable automated inference.

Key topics and technical requirements

• Dialogue Act Markup Language (DiAML): formal language with an XML concrete syntax and formal semantics to represent annotated dialogue acts and enable machine reasoning.
• Segmentation method: procedures to split dialogues into semantic units suitable for annotation.
• Multidimensional annotation: dialogue units can carry multiple communicative functions across dimensions (e.g., task, feedback, turn management).
• Multifunctionality: support for multiple communicative functions per unit.
• Data categories and vocabularies: prescriptive sets for dimensions, communicative functions, qualifiers, and relations between acts.
• Relations and qualifiers: formalization of dependence relations, rhetorical relations and qualifiers to capture interaction structure and feedback.
• Extension & customization mechanisms: plug-in architecture (including triple-layered plug-ins) for application- or domain-specific concepts, semantic content, and emotion annotations (e.g., import of W3C EmotionML).
• Interoperability artifacts: normative annexes such as the DiAML formal specification, DiAML-XML technical schema, and data category listings to ensure consistent implementation.
• Backward compatibility: annotations compliant with earlier editions remain valid.

Practical applications and who should use it

• NLP researchers & corpus developers: annotate corpora for dialogue act labeling, discourse analysis, and multilingual studies.
• Conversational AI & chatbot developers: improve dialogue understanding, intent classification, and response generation by adopting a standard annotation scheme.
• Dialogue system designers: use DiAML for modeling turn-taking, feedback, task management, and social obligations in interactive systems.
• Speech and multimodal interaction engineers: annotate multimodal contributions (speech + gestures) to train recognition and generation components.
• Evaluation and interoperability: create standardized test sets and exchange annotated data across projects and tools.

Benefits include improved annotation consistency, machine-readable semantics for inference, and easy customization for domain-specific dialogue behavior (customer service, healthcare, virtual assistants).

Related standards and references

• Part of the ISO 24617 SemAF series (see ISO catalogue for other parts).
• Integrates concepts from ISO 24617-8:2016 (DR-core) for rhetorical relations.
• Supports enrichment via W3C EmotionML for affective annotation.

Keywords: ISO 24617-2, SemAF, dialogue acts, DiAML, semantic annotation framework, dialogue segmentation, multidimensional annotation, conversational AI, dialogue systems, NLP.