Contextual Distinctiveness: a new lexical property computed from large corpora

Scott McDonald, Richard Shillcock

Research output: Working paper

Abstract / Description of output

We describe the computational formulation of Contextual Distinctiveness (CD), a new lexical property derived from the distributional information present in natural language corpora. CD measures the quantity of information a word conveys about its contexts of use, which we demonstrate to be an interesting and objective indicator of the distributional differences between words. CD is computed from co-occurrence vector representations created using similar methodology to that of Lund and Burgess (1996) and Landauer and Dumais (1997), but provides a means to quantify between-word differences in contextual behavior. We establish the psychological relevance of CD to lexical processing behavior by showing that CD values are significantly correlated with published lexical decision and naming latencies obtained in an isolated word recognition task.
Original languageEnglish
PublisherSchool of Informatics
Publication statusPublished - Jul 2001

Publication series

NameInformatics Research Report
PublisherSchool of Informatics


Dive into the research topics of 'Contextual Distinctiveness: a new lexical property computed from large corpora'. Together they form a unique fingerprint.

Cite this