NATURAL LANGUAGE ACCESS TO FREE-TEXT DATABASES

摘要

Problems of indexing and searching free-text databases are discussed. As they are of general interest for on-line users, they are presented in detail. Especially the possibilities and limitations of Boolean search are shown. An experimental system COPSY (context operator syntax) that was built in order to avoid common errors connected with Boolean search is outlined. This system permits, as input, any natural language search question formulation and yields, as output, documents ranked on the basis of an automatically calculated correspondence between natural language search questions and a content-based analysis of documents.rnCOPSY is part of a text processing project at Siemens, called TINA (Text-Inhalts-Analyse: text content analysis). Software from TINA is actually being applied and evaluated by the US Department of Commerce for patent search and indexing.