Reproduce “LLM-based Argument Mining meets Argumentation and Description Logics: a Unified Framework for Reasoning about Debates” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Gianvincenzo Alfano, Sergio Greco, Lucio La Cava, Stefano Francesco Monea, Irina Trubitsyna
The paper introduces the Fuzzy Argumentative Knowledge Base (FAKB) framework, which unifies LLM-based argument mining with quantitative argumentation semantics and fuzzy description logics to provide transparent and verifiable reasoning over debate texts. FAKB extracts structured, fuzzy argumentative knowledge, refines argument strengths using LLM internal confidence, and enables expressive, ontology-based query answering, with query processing achievable in polynomial time.
Key insights
- Leveraging LLM's internal confidence (log-probabilities) yields more nuanced and semantically interpretable initial argument and relation strengths than purely prompt-based methods.
- Integrating LLM-extracted fuzzy argumentative knowledge with formal argumentation and description logics provides a transparent, explainable, and formally grounded reasoning framework.
- The FAKB framework allows for efficient and formally verifiable consistency checking, along with expressive querying over combined argumentative and domain-specific knowledge.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.