Greetings all, We are very pleased to announce the release of version 2.05 of WordNet-Similarity. This includes some significant improvements to the /utils/*Freq.pl programs that are used for computing information content from different corpora, as well as some improvements to our methods for identifying WordNet compounds in corpora and in WordNet definitions.
You can find links to the CPAN and sourceforge distribution sites here, along with a link to the Change Log that will give more details on this release: http://www.d.umn.edu/~tpederse/similarity.html If you use rawtextFreq.pl to compute information content you should certainly upgrade, as we uncovered several bugs in that program during the development of 2.05. In addition we have modularized the code that is used for counting concepts and propagating those through WordNet, so that now all the *Freq.pl programs use the common module WordNet::Similarity::FrequencyCounter, rather than having their own individual copies of that code. This will make it easier for us to maintain this code, and it will also be easier for users to write their own programs that need to propagate counts through WordNet. We have also modularlized the code that identifies compounds and have placed that in WordNet::Tools. In the process we have uncovered a few small bugs in how we have been identifying WordNet compounds when computing information content and also when processing definitions for the lesk, vector, and vector_pairs measures. The errors in identifying compounds affected a small minority of WordNet compounds, including those that were beyond 8 words long, and those that include punctuation other than _. If you have not upgraded for a few releases, or if you use rawtextFreq.pl, lesk, vector, or vector_pairs, an upgrade to 2.05 is strongly encouraged. Please do let us know if you have any questions or concerns. Enjoy, Ted and Sid -- Ted Pedersen http://www.d.umn.edu/~tpederse
