ABSTRACT
This chapter focuses on the natural language processing (NLP) system developed for the Athena Language Learning Project (ALLP) and in particular, the computational details of the NLP and conclusions reached in the course of its development. The NLP system developed in the ALLP supports Spanish, French, German, English, and Russian. The system addresses both recognition and generation of language. The recognition module accepts typed sentential input from students and analyzes its morphological, syntactic, semantic, temporal, and, to some degree, pragmatic structure. The NLP system in the ALLP uses language-independent processors to control both recognition and generation of input and output at every level of analysis. In the ALLP NLP system, recognition and generation use separate algorithms and grammars. This has the disadvantage of requiring more software, but found it to be worth to maintain separate systems because of the huge effect of error detection on recognition grammars.
