Unsupervised Query-Focused Multi-document Summarization Using uSIF Sentence Embedding Model and Maximal Marginal Relevance Criterion

Extractive query-focused multi-document summarization provides a substitute summary for a collection of documents, by ranking sentences according to their relevance to a pre-given query. In the current study, we propose a novel unsupervised method for query-focused multi-document summarization based on uSIF sentence embedding model and maximal marginal relevance (MMR) criterion. uSIF model is exploited to represent the documents' sentences and users' queries into dense vectors that capture the semantic relationships among multiple words and phrases. MMR criterion is used to re-rank sentences by maintaining query relevance and minimizing redundancy. The proposed method is simple, efficient, and requires no labeled training data. Experiments on the three DUC'2005-2007 benchmarks assess and confirm the effectiveness of the proposed method. The obtained results using ROUGE metrics show that the proposed method outperforms several state-of-the-art systems, including complex deep learning-based systems.