An integer linear programming model for multi document summarization of learning materials using phrase embedding technique

Abstract

Automatic text summarization (ATS) plays a vital role in condensing original text documents while preserving the most crucial information. Its benefits extend to various domains, including e-Learning systems, where educational content can be summarized to facilitate easier access and comprehension. Multi-document summarization (MDS) techniques enable the creation of concise summaries from groups of related text documents. Leveraging MDS for summarizing learning materials opens new avenues, offering students and teachers reference summaries for enhanced learning experiences. This paper introduces a concept-based Integer Linear Programming model for summarizing learning materials, leveraging a phrase embedding technique. Phrases are treated as fundamental and significant semantic building blocks of sentences, facilitating the comprehension and summarization of documents. Embedding techniques are employed to semantically identify related phrases, eliminate redundancy, and enhance coherence through vector representations. Summaries are generated using the ILP technique, selecting key sentences and reducing redundancy with phrase vectors. The paper proposes sentence reordering techniques based on phrases and sentences to further enhance coherence. The resulting summaries are automatically evaluated using ROUGE metrics, demonstrating the superior performance of the proposed approach compared to various benchmark and baseline methods on both the DUC 2004 benchmark dataset and the newly created educational dataset, EduSumm.