Query-Chain Focused Summarization

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 913–922, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics Query-Chain Focused Summarization Tal Baumel Dept. of Computer Science Ben-Gurion University Beer-Sheva, Israel talbau@cs.bgu.ac.il Raphael Cohen Dept. of Computer Science Ben-Gurion University Beer-Sheva, Israel cohenrap@cs.bgu.ac.il Michael Elhadad Dept. of Computer Science Ben-Gurion University Beer-Sheva, Israel elhadad@cs.bgu.ac.il Abstract Update summarization is a form of multi- document summarization where a document set must be summarized in the context of other documents assumed to be known. Efficient update summarization must focus on identify- ing new information and avoiding repetition of known information. In Query-focused summa- rization, the task is to produce a summary as an answer to a given query. We introduce a new task, Query-Chain Summarization, which combines aspects of the two previous tasks: starting from a given document set, increas- ingly specific queries are considered, and a new summary is produced at each step. This process models exploratory search: a user ex- plores a new topic by submitting a sequence of queries, inspecting a summary of the result set and phrasing a new query at each step. We present a novel dataset comprising 22 query- chains sessions of length up to 3 with 3 match- ing human summaries each in the consumer- health domain. Our analysis demonstrates that summaries produced in the context of such exploratory process are different from in- formative summaries. We present an algorithm for Query-Chain Summarization based on a new LDA topic model variant. Evaluation in- dicates the algorithm improves on strong base- lines. 1 Introduction In the past 10 years, the general objective of text summarization has been refined into more specific tasks. Such summarization tasks include: (i) Generic Multi Document Summarization: aims at summarizing a cluster of topically related documents, such as the top results of a search engine query; (ii) in Update Summarization, a set of documents is summarized while assuming the user has already read a summary of earlier doc- uments on the same topic; (iii) in Query-Focused Summarization, the summary of a documents set is produced to convey an informative answer in the context of a specific query. The importance of these specialized tasks is that they help us dis- tinguish criteria that lead to the selection of con- tent in a summary: centrality, novelty, relevance, and techniques to avoid redundancy. We present in this paper a variant summariza- tion task which combines the two aspects of up- date and query-focused summarization. The task is related to exploratory search (Marchionini, 2006). In contrast to classical information seek- ing, in exploratory search, the user is uncertain about the information available, and aims at learning and understanding a new topic (White and Roth, 2009). In typical exploratory search behavior, a user posts a series of queries, and based on information gathered at each step, de- cides how to further explore a set of documents. The metaphor of berrypicking introduced in (Bates, 1989) captures this interactive process. At each step, …

Content truncated to 500 words