Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unveiling Global Discourse Structures: Theoretical Analysis and NLP Applications in Argument Mining

Christopher van Le

TL;DR

The paper addresses the challenge of extracting and representing global discourse structures in argumentative texts through Argument Mining. It surveys ILP-based, transformer-based, and multi-task approaches, and proposes a theoretical, domain-general architecture based on a text-to-text transformer with domain-specific embeddings (e.g., SciBERT) and Longformer-enabled sequences, trained via multi-task pre-training and fine-tuning. Key contributions include a critical assessment of corpora and annotation standards, evidence that multi-task learning enhances cross-domain generalization, and a proposed neural architecture (MT-AM + T2T with biaffine linking) aimed at robust argument structure prediction. The work highlights practical implications for scalable, domain-adaptive argument analysis, while candidly acknowledging data scarcity and annotation heterogeneity as ongoing hurdles.

Abstract

Particularly in the structure of global discourse, coherence plays a pivotal role in human text comprehension and is a hallmark of high-quality text. This is especially true for persuasive texts, where coherent argument structures support claims effectively. This paper discusses and proposes methods for detecting, extracting and representing these global discourse structures in a proccess called Argument(ation) Mining. We begin by defining key terms and processes of discourse structure analysis, then continue to summarize existing research on the matter, and identify shortcomings in current argument component extraction and classification methods. Furthermore, we will outline an architecture for argument mining that focuses on making models more generalisable while overcoming challenges in the current field of research by utilizing novel NLP techniques. This paper reviews current knowledge, summarizes recent works, and outlines our NLP pipeline, aiming to contribute to the theoretical understanding of global discourse structures.

Unveiling Global Discourse Structures: Theoretical Analysis and NLP Applications in Argument Mining

TL;DR

The paper addresses the challenge of extracting and representing global discourse structures in argumentative texts through Argument Mining. It surveys ILP-based, transformer-based, and multi-task approaches, and proposes a theoretical, domain-general architecture based on a text-to-text transformer with domain-specific embeddings (e.g., SciBERT) and Longformer-enabled sequences, trained via multi-task pre-training and fine-tuning. Key contributions include a critical assessment of corpora and annotation standards, evidence that multi-task learning enhances cross-domain generalization, and a proposed neural architecture (MT-AM + T2T with biaffine linking) aimed at robust argument structure prediction. The work highlights practical implications for scalable, domain-adaptive argument analysis, while candidly acknowledging data scarcity and annotation heterogeneity as ongoing hurdles.

Abstract

Particularly in the structure of global discourse, coherence plays a pivotal role in human text comprehension and is a hallmark of high-quality text. This is especially true for persuasive texts, where coherent argument structures support claims effectively. This paper discusses and proposes methods for detecting, extracting and representing these global discourse structures in a proccess called Argument(ation) Mining. We begin by defining key terms and processes of discourse structure analysis, then continue to summarize existing research on the matter, and identify shortcomings in current argument component extraction and classification methods. Furthermore, we will outline an architecture for argument mining that focuses on making models more generalisable while overcoming challenges in the current field of research by utilizing novel NLP techniques. This paper reviews current knowledge, summarizes recent works, and outlines our NLP pipeline, aiming to contribute to the theoretical understanding of global discourse structures.

Paper Structure

This paper contains 15 sections, 1 equation, 8 figures.

Table of Contents

  1. Introduction
  2. Argumentative Structures
  3. Argument Mining
  4. Concepts
  5. Corpora
  6. Automatic Argument Structure Extraction
  7. Related Work
  8. Integer Linear Programming (ILP) Joint Model
  9. Transformer-Based Argument Mining
  10. Multi-Task Argument Mining (MT-AM)
  11. Model discussion
  12. Theoretical Architecture
  13. Requirements
  14. Architecture
  15. Conclusion

Figures (8)

  • Figure 1: Argument structure from an example essay ParsingArgumentationStructures. An argument modeled in this way can be found in the Appendix A
  • Figure 2: Visualization of complexity levels of tasks in argument mining techniques AMSurvey.
  • Figure 3: Multilogue argument structure Ampersand.
  • Figure 4: Distribution of outgoing edges from nodes in CDCP and Essay TowardsNonTree
  • Figure 5: A subset of features used by ParsingArgumentationStructures for component classification
  • ...and 3 more figures