5/12/2006

Professor Marguerite Murphy, Professor Dragutin Petkovic

This project report describes the development of a C++/XML framework for evaluating and reporting equivalences among multi-word idiomatic expressions of different natural anguages. The project is split into two parts: the Lexical Framework and the Knowledge Discovery Framework. The Lexical Framework imports and stores information about individual words, multi-word (non-idiomatic) expressions and conceptual relations among already-stored items. The Knowledge Discovery Framework employs the Lexical Framework, compares known idiomatic expression pairs and reports the degree of relationship of the pair. While the results suggest that it is computationally possible to determine the equivalences of multi-word idiomatic expressions, these results have not been shown to be statistically significant in this project. Some suggestions for making them so are offered.

Computational linguistics, machine translation, WordNet, multi-word expressions.