We are pleased to announce the release of WordNet-SenseRelate v0.02. This
is an "all words" word sense disambiguation program that relies on
information from WordNet-Similarity. It assigns a word the sense which is
most related to the senses of its neighbors.
This is distributed via SourceForge and CPAN, and you can find links to
all of those at:
A number of new features are included in v0.02, perhaps the most
significant of which is the idea of "part of speech coercion". Most of
the measures in WordNet-Similarity (except for lesk, vector, and hso) are
limited to performing noun to noun or verb to verb measurements. However,
in carrying out word sense disambiguation with one of these measures, how
do you find out how related a noun is to a nearby verb? Well, you can
either use a measure that supports such comparisons (lesk, hso, or vector)
or you can "coerce" the part of speech of the verb to a related noun.
This is a very experimental feature, and while we think it's interesting
we aren't quite sure yet how well it works. But it's fun. :)
In addition, the README has been significantly expanded to explain the
algorithm and the support provided by the package. For example, there is
considerable support provided for tracing the algorithm, which can be very
interesting as you move through a large amount of text.
Please give this new version a try, and let us know if you have any
questions about the algorithm or the package itself!
Ted and Jason