Explanatory Summarization with Discourse-Driven Planning

2025-04-27

AI Summary

Existing automatic summarization methods for scientific documents lack explicit modeling of explanatory content, resulting in imbalanced explanation ratios in generated summaries and diminished comprehensibility. To address this, we propose a discourse-structure-driven two-stage planning framework tailored for science communication summarization. Our approach decouples planning from generation by treating discourse planning either as an input condition or as a prefix to the output—marking the first such synergistic, disentangled design. It integrates the PDTB/DRS discourse analysis framework, prompt-augmented sequence generation, controllable decoding, and multi-stage planning. Evaluated on three science communication summarization datasets, our method consistently outperforms state-of-the-art baselines. Both automatic and human evaluations confirm significant improvements in explanatory appropriateness, factual consistency, and controllability, alongside effective hallucination suppression.

Technology Category

• Natural Language Processing: Summarization
• Planning, Routing, and Scheduling: Planning with Language Models
• Data Mining & Knowledge Management: Data Visualization & Summarization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Interdisciplinary science discovery with web data mining

Abstract

Lay summaries for scientific documents typically include explanations to help readers grasp sophisticated concepts or arguments. However, current automatic summarization methods do not explicitly model explanations, which makes it difficult to align the proportion of explanatory content with human-written summaries. In this paper, we present a plan-based approach that leverages discourse frameworks to organize summary generation and guide explanatory sentences by prompting responses to the plan. Specifically, we propose two discourse-driven planning strategies, where the plan is conditioned as part of the input or part of the output prefix, respectively. Empirical experiments on three lay summarization datasets show that our approach outperforms existing state-of-the-art methods in terms of summary quality, and it enhances model robustness, controllability, and mitigates hallucination.

Problem

• Aligning explanatory content proportion with human summaries
• Improving summary quality via discourse-driven planning
• Enhancing model robustness and reducing hallucination

Innovation

• Plan-based approach using discourse frameworks
• Two discourse-driven planning strategies
• Enhances summary quality and model robustness