Linguistic Processing of Texts Using Geppetto

We describe the linguistic analyzer of a prototype for Information Extraction from texts. Such analyzer uses information derived from a shallow processor to limit the computational cost of the analysis. At the same time, shallow techniques are used to collapse parse fragments when a complete parse is not possible. The linguistic analyzer has been built using GePpeTto, an environment that allows the development and integration of diierent linguistic resources and processors. GePpeTto includes: graphical tools for editing and debugging linguistic data, a repertoire of parsers. In this paper, we sketch the architecture of the Information Extraction system, then we show how it is possible to build a linguistic analyzer using GePpeTto. Finally, we present the results of some experiments carried on a corpus of Italian economical short news. Summary We describe the linguistic analyzer of a prototype for Information Extraction from texts. Such analyzer uses information derived from a shallow processor to limit the computational cost of the analysis. At the same time, shallow techniques are used to collapse parse fragments when a complete parse is not possible. The linguistic analyzer has been built using GePpeTto, an environment that allows the development and integration of diierent linguistic resources and processors. GePpeTto includes: graphical tools for editing and debugging linguistic data, a repertoire of parsers. In this paper, we sketch the architecture of the Information Extraction system, then we show how it is possible to build a linguistic analyzer using GePpeTto. Finally, we present the results of some experiments carried on a corpus of Italian economical short news.