Automated part of speech induction to improve understanding of the parts of speech

While the concept of parts of speech is often treated as being well understood, a review of an assortment of applications shows imperfect consistency and little or no justification for the conventions used. I suggest that automated part of speech induction could be used to improve understanding of the parts of speech. I discuss an established algorithm for part of speech induction and four papers that used that algorithm. I then explain how I modified that algorithm to more clearly reflect the intuitive idea that words that occur in similar contexts should be grouped together. I discuss my implementation of the resulting algorithm and the results of testing it on the Brown Corpus. Finally, I discuss what appear to be the main problems with my algorithm and possible basis for solutions to these problems.