Discovering syntactic hierarchies

The acquisition of syntactic rules is predicated upon the successful discovery of syntactic categories (parts of speech). These do not simply constitute a set, but in fact form a nested hierarchy, which allows rules to apply at different levels of generality. Languages provide a variety of cues to syntactic categorization phonological, semantic and distributional. However, the relative reliability of these cues differs from language to language. This paper presents a computational model capable of acquiring the hierarchy syntactic categories from different combinations of cues. Interestingly, the model is domain general and has been successfully applied to non-linguistic discovery of hierarchical structure.