Automatic recognition of complex terms: Problems and the TERMINO solution

While the term-extraction decisions made by terminologists are based primarily on semantic and pragmatic criteria, automated processes have barely started operating at these levels of linguistic analysis. This paper discusses the graphic, lexical, syntactic and semantic difficulties encountered in automated text processing in general and emphasizes in particular certain specific problems that arise in the automatic recognition of complex terms. In order to illustrate the current limitations of existing systems, the article goes on to describe TERMINO, a morphosyntactic text-analysis system developed to help in French-language term extraction. A quantitative and qualitative assessment is made of the system's performance in recognizing complex terms.