Distributional typicality: A new approach to estimating noun and verb usage from large scale text corpora

This paper reports a new approach to estimating the extent to which words h ave predominant noun and verb usages which do not require human judgments a bout parts of speech. The Hyperspace Analog to Language model (HAL, Lund Br Burgess, 1996) was used to computationally estimate noun vs verb usage bas ed on the statistical regularities present in a large-scale electronic text corpus. This measure can be used to estimate the extent to which a given w ord occurs in typical noun or verb sentence contexts (i.e., its distributio nal typicality) in informal contemporary discourse, (C) 2000 Academic Press .