Reduction of Noise and Silence in Full Text Retrieval Systems for Legal Texts

Exploitable results

Legal database users using common full text retrieval (FTR) systems have problems retrieving texts as they have to form queries using combinations of terms ignoring the properties of the legal sublanguage resulting in great 'noise' and 'silence'. RENOS addresses a user community composed of legal experts having experience in legal databases. Existing FTR systems use an inverted list with all the terms appearing in the texts. The RENOS system will be integrated in existing FTR systems providing a retrieval method through concepts (groups of terms with their synonyms), organized in a hierarchical network rather than terms, reducing the 'noise' and 'silence'. This is accomplished with the use of a morphological lexicon containing concept groups of single and compound terms and their synonyms, and a conceptual hierarchical network defining a hierarchy of terms that enables access using concepts rather than strings. The system consists of a parser indexing the corpus and a demonstrator performing retrieval. Both the parser and the demonstrator are implemented in a Windows environment using a C++ compiler. An intelligent inverted list will be constructed consisting of a morphological lexicon, a hierarchical network of legal concepts represented by groups containing single and compound terms and a constituent grammar. Pointers from entries of the lexicon to tree nodes are established. Synonymous terms point to the same concept group. Project results include: analysis of legal texts, statistical analysis, morphological lexicon, conceptual hierarchical network, linguistic behaviour, pattern grammar, Software modules. Deliverables produced include: legal domain analysis; morphological lexicon; hierarchical network of legal terms and legal concept groups.

Exploitable results

Diese Seite teilen

Herunterladen