Optimized Event Storyline Generation based on Mixture-Event-Aspect Model

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 726–735, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Optimized Event Storyline Generation based on Mixture-Event-Aspect Model Lifu Huang Shenzhen Key Lab for Cloud Computing Technology and Applications, Peking University Shenzhen Graduate School, Shenzhen, Guangdong 518055, P.R.China warrior.fu@gmail.com Lian’en Huang∗ Shenzhen Key Lab for Cloud Computing Technology and Applications, Peking University Shenzhen Graduate School, Shenzhen, Guangdong 518055, P.R.China hle@net.pku.edu.cn Abstract Recently, much research focuses on event s- toryline generation, which aims to produce a concise, global and temporal event summary from a collection of articles. Generally, each event contains multiple sub-events and the sto- ryline should be composed by the componen- t summaries of all the sub-events. However, different sub-events have different part-whole relationship with the major event, which is important to correspond to users’ interests but seldom considered in previous work. To distinguish different types of sub-events, we propose a mixture-event-aspect model which models different sub-events into local and global aspects. Combining these local/global aspects with summarization requirements to- gether, we utilize an optimization method to generate the component summaries along the timeline. We develop experimental systems on 6 distinctively different datasets. Evaluation and comparison results indicate the effective- ness of our proposed method. 1 Introduction With the rapid growth of the World Wide Web, in- formation explosion has become an important issue to modern people. Those who search for informa- tion from the Internet often get lost and confused by the overwhelmingly large collection of web doc- uments. So how to get a concise and global pic- ture for a given event subject is an urgent prob- lem to be solved. Although many document un- derstanding systems have been proposed, such as ∗Corresponding author multi-document summarization systems, to gener- ate a compressed summary by extracting the ma- jor information from the collection of documents, they ignored the dynamic development information of an event. Intuitively, each event is long-running and contains multiple sub-events, including related events. Users are likely to prefer a summary of all occurrences of all the sub-events along the timeline of the event. This motivates us to study the task of generating event storyline from a collection of web documents related to an event subject. The research of event storyline summarization is popular in recent years. Its task is to summarize a collection of web documents by extracting represen- tative information based on all the sub-events and generate a global summary. Generally, generating such a global storyline is quite interesting for the fol- lowing main reasons: (1) It can help people catch the whole incident based on an overall temporal struc- tured summary for a given subject, and understand the cause, climax, development process and result of an event. (2) It can also make people know what other events are related, or the effect of this incident to subsequent events, which can present the evolu- tion of an event along a timeline. Though several methods of generating event sto- ryline …

Content truncated to 500 words