Timeline Generation through Evolutionary Trans-Temporal Summarization

Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 433–443, Edinburgh, Scotland, UK, July 27–31, 2011. c⃝2011 Association for Computational Linguistics Timeline Generation through Evolutionary Trans-Temporal Summarization Rui Yan†, Liang Kong† , Congrui Huang†, Xiaojun Wan‡, Xiaoming Li♮, Yan Zhang†∗ †School of Electronics Engineering and Computer Science, Peking University, China ‡Institute of Computer Science and Technology, Peking University, China ♮State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, China {r.yan,kongliang,hcr,lxm}@pku.edu.cn, wanxiaojun@icst.pku.edu.cn,zhy@cis.pku.edu.cn Abstract We investigate an important and challeng- ing problem in summary generation, i.e., Evolutionary Trans-Temporal Summarization (ETTS), which generates news timelines from massive data on the Internet. ETTS greatly facilitates fast news browsing and knowl- edge comprehension, and hence is a neces- sity. Given the collection of time-stamped web documents related to the evolving news, ETTS aims to return news evolution along the time- line, consisting of individual but correlated summaries on each date. Existing summariza- tion algorithms fail to utilize trans-temporal characteristics among these component sum- maries. We propose to model trans-temporal correlations among component summaries for timelines, using inter-date and intra-date sen- tence dependencies, and present a novel com- bination. We develop experimental systems to compare 5 rival algorithms on 6 instinctively different datasets which amount to 10251 doc- uments. Evaluation results in ROUGE metrics indicate the effectiveness of the proposed ap- proach based on trans-temporal information. 1 Introduction Along with the rapid growth of the World Wide Web, document floods spread throughout the Inter- net. Given a large document collection related to a news subject (for example, BP Oil Spill), readers get lost in the sea of articles, feeling confused and powerless. General search engines can rank these ∗Corresponding author. news webpages by relevance to a user specified as- pect, i.e., a query such as “first relief effort for BP Oil Spill”, but search engines are not quite capable of ranking documents given the whole news subject without particular aspects. Faced with thousands of news documents, people usually have a myriad of in- terest aspects about the beginning, the development or the latest situation. However, traditional infor- mation retrieval techniques can only rank webpages according to their understanding of relevance, which is obviously insufficient (Jin et al., 2010). Even if the ranked documents could be in a satis- fying order to help users understand news evolution, readers prefer to monitor the evolutionary trajecto- ries by simply browsing rather than navigate every document in the overwhelming collection. Summa- rization is an ideal solution to provide an abbrevi- ated, informative reorganization for faster and bet- ter representation of news documents. Particularly, a timeline (see Table 1) can summarize evolutionary news as a series of individual but correlated com- ponent summaries (items in Table 1) and offer an option to understand the big picture of evolution. With unique characteristics, summarizing time- lines is significantly different from traditional sum- marization methods which are awkward in such sce- narios. We first study a manual timeline of BP Oil Spill in Mexico Gulf in Table 1 …

Content truncated to 500 words