PRODUCING HIGH-DIMENSIONAL SEMANTIC SPACES FROM LEXICAL COOCCURRENCE

A procedure that processes a corpus of text and produces numeric vecto rs containing information about its meanings for each word is presente d. This procedure is applied to a large corpus of natural language tex t taken from Usenet, and the resulting vectors are examined to determi ne what information is contained within them. These vectors provide th e coordinates in a high-dimensional space in which word relationships can be analyzed. Analyses of both vector similarity and multidimension al scaling demonstrate that there is significant semantic information carried in the vectors. A comparison of vector similarity with human r eaction times in a single-word priming experiment is presented. These vectors provide the basis for a representational model of semantic mem ory, hyperspace analogue to language (HAL).