A common architecture to encourage reuse of natural language/text processing tools

Text analysis systems that handle large amounts of free-form narrative text are beginning to use natural language processing algorithms, employing linguistic-based techniques to deduce meaning. Existing natural language/text processing (NL/TP) systems often serve similar purposes and use similar components, but component reuse is rare. We have designed a domain-designed architecture to promote reusability of NL/TP software and data components. It is constructed from an ontology of the NL/TP domain, which was derived from a survey of NL/TP applications and users, and is represented by an object-oriented model. The architecture also includes interoperability standards, interface protocols, and tools for developing reusable components intended to promote reuse of software and data across a wide range of NL/TP applications and technologies, from keyword searching to semantic analysis.<>