Document-Based Question Answering Improves Query-Focused Multi-document Summarization

Abstract

Due to the lack of large scale datasets, it remains difficult to train neural Query-focused Multi-Document Summarization (QMDS) models. Several large size datasets on the Document-based Question Answering (DQA) have been released and numerous neural network models achieve good performance. These two tasks above are similar in that they all select sentences from a document to answer a given query/question. We therefore propose a novel adaptation method to improve QMDS by using the relatively large datasets from DQA. Specifically, we first design a neural network model to model both tasks. The model, which consists of a sentence encoder, a query filter and a document encoder, can model the sentence salience and query relevance well. Then we train this model on both the QMDS and DQA datasets with several different strategies. Experimental results on three benchmark DUC datasets demonstrate that our approach outperforms a variety of baselines by a wide margin and achieves comparable results with state-of-the-art methods.