A New Approach to Improving Multilingual Summarization Using a Genetic Algorithm

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 927–936, Uppsala, Sweden, 11-16 July 2010. c⃝2010 Association for Computational Linguistics A new Approach to Improving Multilingual Summarization using a Genetic Algorithm Marina Litvak Ben-Gurion University of the Negev Beer Sheva, Israel litvakm@bgu.ac.il Mark Last Ben-Gurion University of the Negev Beer Sheva, Israel mlast@bgu.ac.il Menahem Friedman Ben-Gurion University of the Negev Beer Sheva, Israel fmenahem@bgu.ac.il Abstract Automated summarization methods can be defined as “language-independent,” if they are not based on any language- specific knowledge. Such methods can be used for multilingual summarization defined by Mani (2001) as “processing several languages, with summary in the same language as input.” In this pa- per, we introduce MUSE, a language- independent approach for extractive sum- marization based on the linear optimiza- tion of several sentence ranking measures using a genetic algorithm. We tested our methodology on two languages—English and Hebrew—and evaluated its perfor- mance with ROUGE-1 Recall vs. state- of-the-art extractive summarization ap- proaches. Our results show that MUSE performs better than the best known multi- lingual approach (TextRank1) in both lan- guages. Moreover, our experimental re- sults on a bilingual (English and Hebrew) document collection suggest that MUSE does not need to be retrained on each lan- guage and the same model can be used across at least two different languages. 1 Introduction Document summaries should use a minimum number of words to express a document’s main ideas. As such, high quality summaries can sig- nificantly reduce the information overload many professionals in a variety of fields must contend 1We evaluated several summarizers—SUMMA, MEAD, Microsoft Word Autosummarize and TextRank—on the DUC 2002 corpus. Our results show that TextRank performed best. In addition, TextRank can be considered language- independent as long as it does not perform any morphological analysis. with on a daily basis (Filippova et al., 2009), as- sist in the automated classification and filtering of documents, and increase search engines precision. Automated summarization methods can use different levels of linguistic analysis: morphological, syntactic, semantic and dis- course/pragmatic (Mani, 2001). Although the summary quality is expected to improve when a summarization technique includes language specific knowledge, the inclusion of that knowl- edge impedes the use of the summarizer on multiple languages. Only systems that perform equally well on different languages without language-specific knowledge (including linguistic analysis) can be considered language-independent summarizers. The publication of information on the Internet in an ever-increasing variety of languages 2 dic- tates the importance of developing multilingual summarization approaches. There is a particu- lar need for language-independent statistical tech- niques that can be readily applied to text in any language without depending on language-specific linguistic tools. In the absence of such techniques, the only alternative to language-independent sum- marization would be the labor-intensive transla- tion of the entire document into a common lan- guage. Here we introduce MUSE (MUltilingual Sen- tence Extractor), a new approach to multilingual single-document extractive summarization where summarization is considered as an optimization or a search problem. We use a …

Content truncated to 500 words