MUSE – A Multilingual Sentence Extractor

MUltilingual Sentence Extractor (MUSE) is aimed at multilingual single-document summarization. MUSE implements the supervised language-independent summarization approach based on optimization of multiple statistical sentence ranking methods. The MUSE tool consists of two main modules: the training module activated in the offline mode, and the on-line summarization module. The training module can be provided with a corpus of summarized texts in any language. Then, it learns the best linear combination of user specified sentence ranking measures applying a Genetic Algorithm to the given training data. The summarization module performs real-time sentence extraction by computing sentence rankings according to the weighted model induced in the training phase. The main advantage of MUSE is its language-independency – it can be applied to any language given a gold standard summaries in that language. The performance of MUSE in our previous works was found to be significantly better than the best known state-of-theart extractive summarization approaches and tools in the three different languages: English, Hebrew, and Arabic. Moreover, our experimental results in a cross-lingual domain suggest that MUSE does not need to be retrained on each new language, and the same weighting model can be used across several languages.