We are happy to announce the release of WordNet-SenseRelate-AllWords
version 0.05. This version features a number of significant changes,
perhaps the most important of which is a new definition of windowing that
is more flexible than that used in version 0.04 and earlier.
You can download from:
In previous versions the option --window N meant "N WordNet words to the
left and N words to the right." This means that you would always try to
have an equal number of words on either side of the word being
The new --window option is more flexible in that it specifies the total
size of the window (including the word being disambiguated). So, you can
specify --window 2, which now means 1 word to the right and the word being
disambiguated. --window 3 means 1 word to the left, 1 word to the right,
and the word to be disambiguated.
So, when the value of --window is odd, you have a balanced number of
words on either side (as we did in previous versions). What is new now is
that if you have an even number of words, you have 1 more word to the left
of the word to be disambiguated than you do to the right.
In our experience, --window 2 (new formulation) often works quite well
(this means assigning a sense to a word by considering its relatedness to
one word to the left, that is prior to it). So that might be a good
starting point for experiments.
There are other changes as well, including some renaming and improved
documentation for our utilities that convert SemCor-formattted data into
a form that our program wsd.pl can process (semcor-reformat.pl), and that
produce output files suitable for scoring by the Senseval scoring program
Please let us know if you have any questions!
Ted and Jason
Get latest updates about Open Source Projects, Conferences and News.