A Query-Based Summarization Service from Multiple News Sources

Abstract

It can be time consuming to search Internet news, due to multiple sources reporting repetitive information. Given a query and a set of relevant text articles, query-focused multi-document summarization (QMDS) aims to generate a fluent, well-organized, and compact summary that answers the query. While QMDS helps to summarize search results, most top-performing systems for this purpose remain largely extractive. Extractive summarization extracts a group of sentences and concatenates them. In this paper, we propose a summarization service based on abstractive QMDS using multi-sentence compression (MSC). Our proposed service generates a novel summary representing the gist of the content of the source document(s). Experiments using popular summarization benchmark datasets demonstrate the effectiveness of the proposed service.