Paper: Good Bigrams

ACL ID C96-2100
Title Good Bigrams
Venue International Conference on Computational Linguistics
Session Main Conference
Year 1996
Authors

A desired property of a measure of connective strength in bigrams is that the measure should be insensitive to corpus size. This paper investigates the stability of three different measures over text genres and expansion of the corpus. The measures are (1) the commonly used mutual information, (2) the difference in mutual informa- tion, and (3) raw occurrence. Mutual information is further compared to using knowledge about genres to re- move overlap between genres. This last approach considers the difference between two products of the same process (human text-generation) con- strained by different genres. The can- cellation of overlap seems to provide the most specific word pairs for each genre.