Cluster word vectors with cosine

I have vector sets, of the format: wordA|wordB|wordC; 0.8|0.5|0.3 wordA|wordC|wordD; 0.7|0.3|0.1 wordA|wordE|wordF; 0.9|0.2|0.2 wordC|wordE|wordF; 0.3|0.3|0.1 (and 8 more) The numbers are relevance scores for each word. The goal is to reduce/cluster this into fewer sets. e.g., to 2 sets. There are existing methods for clustering, e.g., cosine similarity or [login to view URL] Your input is a textfile similar to the above (my file format is slightly more complex and larger). The output should include a measure of similarity. i.e., a way to know when to stop clustering. In some cases, we will reduce to 8 clusters, in others we cluster down to 1-2 clusters. If you are familiar with nltk or similar, this should be a simple task. Thanks.