ThreadSumm: Summarization of Nested Discourse Threads Using Tree of Thoughts

2026-04-19

AI Summary

This work addresses the challenge of accurately summarizing deeply nested discussions, where interleaved replies, quotations, and overlapping topics hinder existing large language model–based summarization approaches. To overcome this, the authors propose a hierarchical thread-aware summarization method that leverages explicit discourse structure to guide content planning. The approach first extracts atomic content units and constructs multi-perspective, thread-aware sequences by integrating sentence ordering. It then employs Tree-of-Thoughts search to generate and score multiple paragraph candidates, jointly optimizing coherence and coverage. Experimental results demonstrate that the proposed method significantly outperforms current baselines in preserving logical structure, topic coverage, and salient viewpoints, thereby substantially improving the quality of summaries for nested discussions.

Technology Category

• Natural Language Processing: Summarization
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Data Visualization & Summarization

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

Summarizing deeply nested discussion threads requires handling interleaved replies, quotes, and overlapping topics, which standard LLM summarizers struggle to capture reliably. We introduce ThreadSumm, a multi-stage LLM framework that treats thread summarization as a hierarchical reasoning problem over explicit aspect and content unit representations. Our method first performs content planning via LLM-based extraction of discourse aspects and Atomic Content Units, then applies sentence ordering to construct thread-aware sequences that surface multiple viewpoints rather than a single linear strand. On top of these interpretable units, ThreadSumm employs a Tree of Thoughts search that generates and scores multiple paragraph candidates, jointly optimizing coherence and coverage within a unified search space. With this multi-proposal and iterative refinement design, we show improved performance in generating logically structured summaries compared to existing baselines, while achieving higher aspect retention and opinion coverage in nested discussions.

Problem

• nested discourse threads
• summarization
• interleaved replies
• overlapping topics
• aspect retention

Innovation

• ThreadSumm
• Tree of Thoughts
• nested discourse summarization
• Atomic Content Units
• multi-viewpoint summarization