Discourse Relations and Connectives in Higher Text Structure

2026-04-19 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, quick

Summary

This article explores the capabilities and constraints of local discourse coherence analysis in relation to global coherence and higher text structures. Researchers analyzed Czech and English corpora annotated with local discourse relations to identify indicators of higher discourse structure. The study classified patterns of sequential and overlapping local relation pairs, as well as hierarchies of nested local relations, observing up to five levels in both languages. Specific attention was given to relations spanning paragraph boundaries, their semantic types, and discourse connectives initiating paragraphs. The analysis also examined instances where annotators marked large arguments (exceeding one sentence) despite a minimality principle, revealing a prevalence of large left-sided arguments in locally annotated data. This work provides new linguistic insights into coherence signals and reflections on annotation methodologies.

Key takeaway

For NLP Engineers developing discourse parsers, understanding the interplay between local and global coherence is crucial. Your models should account for nested discourse relation hierarchies and the distinct semantic types of relations crossing paragraph boundaries to improve overall text understanding. Consider how large arguments are handled in annotation schemes to refine parsing accuracy.

Key insights

Local discourse analysis can reveal clues about higher text structure and global coherence.

Principles

• Discourse relations form hierarchies up to 5 levels.
• Semantic types differ across paragraph boundaries.

Method

The study classified local relation patterns, analyzed cross-paragraph relations and paragraph-initial connectives, and examined large argument annotation tendencies in Czech and English corpora.

In practice

• Identify Czech connectives typical for higher structures.
• Note prevalence of large left-sided arguments.

Topics

• Discourse Relations
• Discourse Coherence
• Higher Text Structure
• Discourse Connectives
• Corpus Annotation

Best for: NLP Engineer, AI Scientist, Research Scientist

Related on AIssential

• A Cross-Genre Analysis of Discourse Relation Signaling in the GUM Corpus — The paper analyzes discourse relation signaling across diverse genres in the GUM Corpus.
• Beyond semantics: the challenges of annotating pragmatic and discourse phenomena — Annotating abstract semantic and pragmatic information is challenging due to its implicit, context-dependent nature.
• Cohesion-6K: An Arabic Dataset for Analyzing Social Cohesion and Conflict in Online Discourse — Divisive online discourse consistently attracts disproportionately higher user engagement compared to unifying narratives.
• The Cosine Similarity Trap: Why Embeddings Can’t Distinguish “War” from “Union” — Language models often cannot distinguish complementarity from conflict due to how embeddings encode relatedness, not relationship.
• Presupposition and Reasoning in Conditionals: A Theory-Based Study of Humans and LLMs — LLMs often match human linguistic judgments via surface patterns, not deep pragmatic reasoning.
• How NLP Communication Agents Actually Understand You (Spoiler: It’s a Lot Weirder Than You’d Think) — NLP systems process language through five interdependent layers, from tokenization to pragmatic intent.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.