Experiments in Extractive Summarization: Integer Linear Programming, Term/Sentence Scoring, and Title-driven Models
This work addresses the problem of generating concise summaries from single documents, which is incremental as it builds upon and combines existing techniques in extractive summarization.
The paper tackled unsupervised single-document summarization by developing a flexible framework called NewsSumm that integrates integer linear programming, parameterized normalization, and title-driven approaches, resulting in performance improvements for both unsupervised and supervised methods.
In this paper, we revisit the challenging problem of unsupervised single-document summarization and study the following aspects: Integer linear programming (ILP) based algorithms, Parameterized normalization of term and sentence scores, and Title-driven approaches for summarization. We describe a new framework, NewsSumm, that includes many existing and new approaches for summarization including ILP and title-driven approaches. NewsSumm's flexibility allows to combine different algorithms and sentence scoring schemes seamlessly. Our results combining sentence scoring with ILP and normalization are in contrast to previous work on this topic, showing the importance of a broader search for optimal parameters. We also show that the new title-driven reduction idea leads to improvement in performance for both unsupervised and supervised approaches considered.