Concordancing Revised or How to Aid the Recognition of New Senses in Very Large Corpora

This paper describes the application of a framework for text analysis to the problem of distinguishing unusual or non-standard usage of words in large corpora. The need to identify such novel uses, and augment machine-readable dictionaries is a constant battle for professional lexicographers that need to update their resources in order to keep up with the development of the dynamic and evolving aspects of human language. Of equal importance is the need to devise automatic means upon which we can evaluate to what extent a (defining) dictionary accounts for what we find in corpus data. A combination of both semi-, and automatic means have been explored, and it seems that Machine Learning might be a plausible solution towards the stated goals.