TR-IT-0073 :1994.9.8

Hervé BLANCHON

Pattern-based approach to interactive disambiguation: first definition and experimentation

Abstract:The aim of this document is to have the reader acquire a clear view of the pattern-based interactive disambiguation process developed. The kernel of the disambiguation process has two main entries: the data produced by an analyser and some sets of patterns describing the ambiguities to be recognized and solved. The kernel is thus language independent, which means that disambiguation processes can be developed for other languages using the methodology we used to develop an interactive disambiguation process for French input. The first part of the document gives a brief description of the LIDIA-1.0 mockup to situate the disambiguation process. The second part of the document is an in-depth study of the design of the disambiguation process. This part is an important update and correction of [Blanchon 1994a] which is released in French. This work has been done at ATR. The Appendix A gives a linguistic classification of the ambiguities we have to solve in the field of Natural Language Processing. The Appendix B gives a description of the structure produced by the analyzer used in the framework of the development of the LIDIA-1.0 mockup. The Appendix C gives the list and the meaning of the labels produced by the analyzer which are relevant to the disambiguation process.