bk2vec

Injecting background knowledge into the word vectors

How to train word vectors

• Place your texts under the corpora folder. Script assumes one text per line, splits sentences on dot.
• Launch TextReader.py. It will train word vectors and save them into model.w2v file
• Fill free to implement a proper script that will actually take parameters and won't use the hardcoded ones

How to extract texts from the wikipedia dump

java -cp thewikimachine.jar org.fbk.cit.hlt.thewikimachine.xmldump.WikipediaTextExtractor 
-d <path-to-dump.xml>
-o <path-to-output-directory>
-t <amount of threads>