Summarizing Definition from Wikipedia

Proceedings of the 47th Annual Meeting of the ACL and the 4th IJCNLP of the AFNLP, pages 199–207, Suntec, Singapore, 2-7 August 2009. c⃝2009 ACL and AFNLP Summarizing Definition from Wikipedia Shiren Ye and Tat-Seng Chua and Jie Lu Lab of Media Search National University of Singapore {yesr|chuats|luj}@comp.nus.edu.sg Abstract Wikipedia provides a wealth of knowl- edge, where the first sentence, infobox (and relevant sentences), and even the en- tire document of a wiki article could be considered as diverse versions of sum- maries (definitions) of the target topic. We explore how to generate a series of summaries with various lengths based on them. To obtain more reliable associations between sentences, we introduce wiki con- cepts according to the internal links in Wikipedia. In addition, we develop an extended document concept lattice model to combine wiki concepts and non-textual features such as the outline and infobox. The model can concatenate representative sentences from non-overlapping salient lo- cal topics for summary generation. We test our model based on our annotated wiki ar- ticles which topics come from TREC-QA 2004-2006 evaluations. The results show that the model is effective in summariza- tion and definition QA. 1 Introduction Nowadays, ‘ask Wikipedia’ has become as pop- ular as ‘Google it’ during Internet surfing, as Wikipedia is able to provide reliable information about the concept (entity) that the users want. As the largest online encyclopedia, Wikipedia assem- bles immense human knowledge from thousands of volunteer editors, and exhibits significant contribu- tions to NLP problems such as semantic related- ness, word sense disambiguation and question an- swering (QA). For a given definition query, many search en- gines (e.g., specified by ‘define:’ in Google) often place the first sentence of the corresponding wiki1 article at the top of the returned list. The use of 1 For readability, we follow the upper/lower case rule on web (say, ‘web pages’ and ‘on the Web’), and utilize one-sentence snippets provides a brief and concise description of the query. However, users often need more information beyond such a one-sentence de- finition, while feeling that the corresponding wiki article is too long. Thus, there is a strong demand to summarize wiki articles as definitions with vari- ous lengths to suite different user needs. The initial motivation of this investigation is to find better definition answer for TREC-QA task using Wikipedia (Kor and Chua, 2007). Accord- ing to past results on TREC-QA (Voorhees, 2004; Voorhees and Dang, 2005), definition queries are usually recognized as being more difficult than fac- toid and list queries. Wikipedia could help to improve the quality of answer finding and even provide the answers directly. Its results are bet- ter than other external resources such as WordNet, Gazetteers and Google’s define operator, especially for definition QA (Lita et al., 2004). Different from the free text used in QA and sum- marization, a wiki article usually contains valuable information like infobox and wiki link. Infobox tabulates the key properties about the target, such as birth place/date and spouse for a person …

Content truncated to 500 words