Advances in Natural Language Processing for Discourse and Argumentation

The field of natural language processing is moving towards more nuanced and context-aware models for discourse and argumentation. Recent research has focused on developing models that can better capture the complexities of human language, including the use of implicit and explicit cues to convey meaning. One key area of development is in the use of large language models to improve performance on tasks such as argument mining, entailment detection, and readability assessment. These models have shown significant promise in their ability to capture long-range dependencies and contextual relationships, but still face challenges in terms of transparency and interpretability. Noteworthy papers in this area include: JUDGEBERT, which introduces a novel evaluation metric for legal meaning preservation in French legal text simplification, demonstrating superior correlation with human judgment. LongReasonArena, which presents a benchmark for assessing the long reasoning capabilities of large language models, highlighting significant challenges for current models. ArgCMV, which introduces a new argument key point extraction dataset comprising around 12K arguments from actual online human debates, setting the stage for the next generation of LLM-driven summarization research.

Sources

What makes an entity salient in discourse?

JUDGEBERT: Assessing Legal Meaning Preservation Between Sentences

A Straightforward Pipeline for Targeted Entailment and Contradiction Detection

German4All - A Dataset and Model for Readability-Controlled Paraphrasing in German

AMELIA: A Family of Multi-task End-to-end Language Models for Argumentation

Evaluating the Evaluators: Are readability metrics good measures of readability?

LongReasonArena: A Long Reasoning Benchmark for Large Language Models

Towards a Holistic and Automated Evaluation Framework for Multi-Level Comprehension of LLMs in Book-Length Contexts

ArgCMV: An Argument Summarization Benchmark for the LLM-era

Joint Enhancement of Relational Reasoning for Long-Context LLMs

Multi-Lingual Implicit Discourse Relation Recognition with Multi-Label Hierarchical Learning

GDLLM: A Global Distance-aware Modeling Approach Based on Large Language Models for Event Temporal Relation Extraction

Built with on top of