Building a Morphosyntactic Lexicon and a Pre-syntactic Processing Chain for Polish

摘要

This paper introduces a new set of tools and resources for Polish which cover all the steps required to transform a raw unrestricted text into a reasonable input for a parser. This includes (1) a large-coverage morphological lexicon, developed thanks to the IPI PAN corpus as well as a lexical acquisition technique, and (2) multiple tools for spelling correction, segmentation, tokenization and named entity recognition. This processing chain is also able to deal with the XCES format both as input and output, hence allowing to improve XCES corpora such as the IPI PAN corpus itself. This allows us to give a brief qualitative evaluation of the lexicon and of the processing chain.