Dictionary validation through a clustering technique

Barriere and Popowich (1996) have developed a technique, implemented by Barriere (1997), that takes as a "trigger" a word sense defined in a dictionary and finds its concept cluster a group of word senses related to the trigger that shows its larger semantic context in the dictionary. The technique can be extended to take a concept cluster as a "trigger" and find an extended concept cluster by combining the clusters found for the word senses in the trigger cluster. In this paper, we present two applications of the concept clustering technique to validating dictionary definitions. The first application is a qualitative tool for seeing connections and noting inconsistencies among dictionary definitions; the second application is two quantitative measures of the coherence of dictionary definitions and interdefinitions. The first measure can help in writing tight-knit interdefinitions of related word senses; the second can help ensure that individual word sense definitions are of high quality.