MUSI addresses the problem of multilingual summarisation to facilitate access to information retrieval systems and, in particular, access to the content of electronic documents over the Internet. As a first phase of a more global service, MUSI will create French and German summaries from English and Italian texts. While most current approaches to summarisation rely in fact on statistics-based sentence extraction and condensation, MUSI aims at conceptual summarisation: First, language analysis techniques are used to represent the content of the texts in an abstract way. Summarisation is then carried out on this content representation. Finally the result is verbalized using language generation techniques. The chosen approach is limited by the linguistic and conceptual coverage of the domain in hand, but it offers important benefits other methods are missing: current challenges such as multi-lingual and multi-document summaries fall out rather straightforwardly, and the possibility of generating statements about the texts introduces the flexibility that is urgently needed for user adaptive summaries.
- Conceptual text summarisation through language analysis and generation
- Multi-lingual and multi-document summaries
- User-oriented summaries through the generation of meta statements
- Validation within the scope of a medical application
Partners
LexiQuest, France (Technical Co-ordinator)
Istituto di Linguistica Computazionale, Italy
DFKI GmbH, Germany