Please read http://blog.csdn.net/yangliuy/article/details/8330640, I find it rather interesting.

###First work version run.py method dataset_type data_path model output

• Method: lsa or plsa
• Dataset types: beetle or seb
• Model: 2 3 or 5

###Adout WSD before LSA/pLSA There are 2 options:

• Considering Semantic info for each word: http://svn.ask.it.usyd.edu.au/trac/candc/wiki/boxer
• Replace 同义词、反义词 to a semantic space:http://blog.sina.cn/dpool/blog/s/blog_630aa24f01010xbb.html

###A complete working example

• http://www.codeproject.com/Articles/11835/WordNet-based-semantic-similarity-measurement