Multilingual Single-Document Summarization with MUSE

MUltilingual Sentence Extractor (MUSE) is aimed at multilingual single-document summarization. MUSE implements a supervised language-independent summarization approach based on optimization of multiple sentence ranking methods using a Genetic Algorithm. The main advantage of MUSE is its language-independency – it is using statistical sentence features, which can be calculated for sentences in any language. In our previous work, the performance of MUSE was found to be significantly better than the best known state-of-the-art extractive summarization approaches and tools in three different languages: English, Hebrew, and Arabic. Moreover, our experimental results in the cross-lingual domain suggest that MUSE does not need to be retrained on a summarization corpus in each new language, and the same weighting model can be used across several languages (Last and Litvak, 2012). MUSE participated in the MultiLing 2013 single document summarization task on three languages: English, Hebrew and Arabic. Due to a very limited time that was given to the participants to run their systems on the MultiLing 2013 data, the results submitted to evaluation were obtained by summarizing the documents using models pre-trained on different corpora. As such, no training has been performed on the MultiLing 2013 corpus. 1 MUltilingual Sentence Extractor (MUSE): Overview