Inxight proposes to develop an extensible linguistic approach for resolution of Named Entity (NE) Coreference. We employ a pipeline approach - where text is first fed through a set of linguistic preprocessing services, then through a series of finite-state grammars to extract candidate entities, and finally through the entity co-referencing process. The new work to be performed in coreference resolution includes new knowledge sources, corpora, parsers, and a complete Coreference Resolution Module (CORM), implemented as a rule-based linguistic system. Syntactic, semantic, and discourse capabilities are provided. Our approach includes innovations in the application of Optimality Theory for constraints reasoning, and the development of discourse features in a hybrid approach that employs finite state transducers and grammar formalisms such as Lexical Functional Grammar (LFG). The proven implementation of finite-state grammars and techniques in the base platform, combined with new coreference research adapting a lexicalist approach such as LFG, provide a performance optimized approach necessary to perform the intermediate-level parsing required to meet the computational demands of coreference chaining. The technology platform also supports the language and domain coverage, and services oriented component architect as may be required by future operational applications
