Most accounts of statistical learning (e.g., Saffran et al., 1996) assume that the learner computes cooccurrence statistics over units, such as syllables, that abstract away from the physical features of the input. This assumption need not hold when the units are underlearned, unfamiliar, or uncategorizable. We tested statistical learning with variable, unfamiliar units and show that if the fea...

How many words are needed to define all the words in a dictionary? Graph-theoretic analysis reveals that about 10% of a dictionary is a unique Kernel of words that define one another and all the rest, but this is not the smallest such subset. The Kernel consists of one huge strongly connected component (SCC), about half its size, the Core, surrounded by many small SCCs, the Satellites. Core wor...

Phonologists and psycholinguists draw a three-way distinction amongst real words, possible words, and impossible words.The distinction between real words and possible words provides the foundation for lexical decision experiments. The distinction between possible words and impossible words reveals implicit cognitive generalizations about words in a language, and thereby contributes to the under...

We propose a proximity probabilistic model (PPM) that advances a bag-of-words probabilistic retrieval model. In our proposed model, a document is transformed to a pseudo document, in which a term count is propagated to other nearby terms. Then we consider three heuristics, i.e., the distance of two query term occurrences, their order, and term weights, and try four kernel functions in measuring...

This paper presents a method for inducing the parts of speech of a language and part-of-speech labels for individual words from a large text corpus. Vector representations for the part-of-speech of a word are formed from entries of its near lexical neighbors. A dimen-sionality reduction creates a space representing the syntactic categories of unambiguous words. A neural net trained on these spa...

This paper devises a novel kernel function for structured natural language data. In the field of Natural Language Processing, feature extraction consists of the following two steps: (1) syntactically and semantically analyzing raw data, i.e., character strings, then representing the results as discrete structures, such as parse trees and dependency graphs with part-of-speech tags; (2) creating ...