A Maximum Entropy Model of Phonotactics and Phonotactic Learning

The authors present a model of the phonotactic and orthographic constraints of Australian and Standard Southern British English monosyllables. This model is used as the basis for a web-based psycholinguistic resource, the ARC Nonword Database, which contains 358,534 monosyllabic nonwords--48,534 pseudohomophones and 310,000 non-pseudohomophonic nonwords. Items can be selected from the ARC Nonword Database on the basis of a wide variety of properties known or suspected to be of theoretical importance for the investigation of reading.

One ubiquitous problem in language processing involves the assignment of words to the correct grammatical category, such as noun or verb. In general, semantic and syntactic cues have been cited as the principal information for grammatical category assignment, to the neglect of possible phonological cues. This neglect is unwarranted, and the following claims are made: (a) Numerous correlations between phonology and grammatical class exist, (b) some of these correlations are large and can pervade the entire lexicon of a language and hence can involve thousands of words, (c) experiments have repeatedly found that adults and children have learned these correlations, and (d) explanations for how these correlations arose can be proposed and evaluated. Implications of these phenomena for language representation and processing are discussed.