On the Complexity and Typology of Inflectional Morphological Systems

We lay out a computational model for syntactic category acquisition which combines psychologically plausible concepts from minimally supervised part-of-speech tagging applications with simple distributional statistics. The model assumes a small set of seed words (Haghighi & Klein 2006), an approach with motivation in Pinker (1984)'s semantic bootstrapping hypothesis, and iteratively constructs a hierarchical clustering over a growing lexicon. Clustering is performed on the basis of word-adjacent syntactic frames alone (Mintz 2003) which has been shown to yield qualitatively coherent POS clusters (Reddington 1998), with no reference to morphological features. A prototype-driven labelling process based on tree-distance yields results comparable to computationally complex unsupervised algorithms while maintaining its cognitive underpinnings.