Discourse Structure: Theory, Practice and Use

Bonnie Webber, Markus Egg, Valia KordoniACL 2010

TL;DR

This tutorial provides a comprehensive overview of discourse structure, a crucial but often overlooked aspect of natural language. It bridges the gap between theoretical linguistics and practical applications in language technology, outlining methods for recognizing and generating discourse structure and highlighting its utility in various NLP tasks.

What Problem Does This Paper Address?

Natural language processing (NLP) systems often struggle to understand the deeper meaning and coherence of text beyond individual sentences. While much research focuses on sentence-level semantics and syntax, the way sentences connect and build upon each other to form a coherent discourse is a complex challenge. This tutorial addresses the need for a structured understanding of discourse, explaining its theoretical underpinnings, practical computational approaches, and its relevance to advancing language technologies. It aims to equip attendees with the knowledge to leverage discourse structure for more sophisticated NLP applications.

Methods

This tutorial is structured into four parts, outlining a comprehensive approach to discourse structure:

  • Part I: General Overview: Introduces the fundamental bases of discourse structuring, including monologic, dialogic, and multiparty discourse. It delves into the aspects of discourse structure that are particularly relevant to language technology and discusses the evidence that supports its existence and importance.
  • Part II: Computational Recognition and Generation: Explores computational methods for automatically identifying and creating discourse structure. This includes techniques like discourse chunking and parsing, recognizing the arguments and senses of discourse connectives (e.g., "because," "however"), and modeling entity-based discourse structure. Dialogue parsing is also covered.
  • Part III: Applications and Resources: Details how discourse structure can be applied to various language technology tasks. It also surveys available discourse structure resources, covering both monolingual and multilingual datasets.
  • Part IV: Future Developments: Discusses potential future directions and advancements in the field.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Discourse structure is key to deeper language understanding: Moving beyond sentence-level analysis, understanding how sentences connect is vital for tasks like summarization, question answering, and machine translation.
  • Bridging theory and practice: The tutorial aims to make theoretical linguistic concepts of discourse structure accessible and applicable to computational linguists and NLP practitioners.
  • Computational approaches are maturing: Significant progress has been made in automatically recognizing and generating discourse structure, paving the way for more sophisticated NLP systems.
  • Resources are growing: The availability of annotated datasets and tools is crucial for advancing research and development in discourse structure.

Key Results

Not discussed in the paper.

Limitations and Future Work

The tutorial acknowledges that while significant progress has been made, there are still many avenues for future research. These include developing more robust computational models for discourse parsing and generation, expanding the availability of multilingual discourse resources, and exploring novel applications of discourse structure in emerging NLP areas. The tutorial implicitly suggests that further work is needed to fully integrate discourse structure understanding into mainstream NLP systems.

Summary updated

BibTeX
@inproceedings{webber2010discourse,
  author = {Bonnie Webber and Markus Egg and Valia Kordoni},
  title = {{Discourse Structure: Theory, Practice and Use}},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  year = {2010},
}
Figure