Overview
ISO/TS 24617-5:2014 - SemAF-DS (Discourse structure) is a Technical Specification within the ISO Semantic Annotation Framework (SemAF) for language resource management. It defines how a discourse is modeled in two complementary layers:
- Segment structure - the formal/presentational organization of discourse (words, phrases, clauses, sentences, paragraphs).
- Content structure - the semantic and pragmatic entities and relations those segments refer to.
The standard shows how these dual structures can be represented as labelled directed graphs (or trees where appropriate) and how mappings from segments to content nodes are annotated. The current edition focuses on text-based annotation but is explicitly extensible to other modalities (e.g., audio, video).
Key topics and technical requirements
- Definition and scope of discourse-related terms (segment, discourse relation, entity, circumstance).
- Segment structure: represents syntactic organization; nodes are discourse segments; supports continuous and discontinuous segments; distinguishes governing vs non-governing segments.
- Content structure: nodes represent semantic/pragmatic entities; edges represent discourse relations (semantic/pragmatic links between circumstances/objects).
- Mapping: explicit annotations that link segment structure elements to content structure nodes, supporting interoperability between presentation and meaning.
- Representation conventions: compatibility with graph-based encodings (LAF) and neutrality between inline and stand-off annotation methods.
- Interoperability guidance: designed to interoperate with related ISO frameworks (SynAF for syntax, LAF for graph encodings) and existing discourse annotation practices (e.g., RST, SDRT).
Applications and users
ISO/TS 24617-5:2014 is practical for:
- NLP researchers and corpus linguists building discourse-annotated corpora and gold standards.
- Annotation tool developers and language resource engineers needing a standard graph-based encoding for discourse.
- Machine translation and natural language generation teams that require consistent mappings from presentation to semantic content (the standard even supports use cases such as automated multilingual subtitle generation).
- Digital humanities, dialogue-system developers, and multimodal annotation projects that want an extensible, language-neutral pivot for interoperation across formats.
Practical benefits include improved corpus interoperability, clearer semantics-to-form mappings for downstream NLP tasks, and reduced cost for multilingual generation workflows.
Related standards
- ISO 24612:2012 - Linguistic Annotation Framework (LAF) - graph encoding.
- ISO 24615:2010 - Syntactic Annotation Framework (SynAF).
- Other SemAF parts: ISO/TS 24617-1 (Time & events), 24617-2 (Dialogue acts), 24617-4 (Semantic roles), plus parts on principles and spatial/semantic relations.
Keywords: ISO/TS 24617-5:2014, SemAF-DS, discourse structure, semantic annotation, language resource management, segment structure, content structure, discourse relations, LAF, SynAF.