DFKI-LT - Language Generation for Cross-Lingual Document Summarisation

Stephan Busemann
Language Generation for Cross-Lingual Document Summarisation
1 Proceedings of the International Workshop on Innovative Language Technology and Chinese Information Processing (ILT&CIP '01), Science Press, Shanghai, 2001
User-adaptive summaries of longer texts in the user's language are a major prerequisite for successful and efficient navigation in the results offered by WWW search engines and information retrieval systems. Current summarisation systems are either monolingual or use existing low-performance machine translation technology to target the user's desired language. What is more, their results canot be tailored to the user's needs. The present contribution describes generation techniques required for cross-lingual summarisation, both with respect to the content of the summary and to user-oriented meta-information about the original document. These techniques have been implemented in the MUSI system that summarises scientific medical texts originally written in Italian or English in German or French. MUSI uses deep linguistic processing on selected parts of the document. By using query-based selection and presenting additional information abˇut the documents upon request, MUSI generates summaries tailored towards the user's needs.
Files: BibTeX, musinlg.pdf