Semantic Processing: Statistical Approaches

The development of methods for representing meaning is a critical aspect of cognitive modeling and applications that must extract meaning from text input. The ability to derive meaning is the key to any approach that needs to use or evaluate knowledge. With the advent of more powerful computing and the availability of on-line texts and machine-readable dictionaries, novel techniques have been developed that can automatically derive semantic representations. These techniques capture effects of regularities inherent in language to learn about semantic relationships among words. The techniques operate on large corpora, permitting automatic development of lexicons on large samples of language. The techniques can be incorporated into methods for cognitive modeling on a wide range of psychological phenomena such as language acquisition, discourse processing, categorization, and memory. In addition, the techniques can be used in applied settings, in which a computer can derive semantic knowledge representations from text. These settings include information retrieval, natural language processing, and discourse analysis.