The program timbl_wsd implements a WSD system based in a machine learning based engine called TIMBL. TIMBL immplements a k-nearest neighbor machine learning approach. TIMBL is highly configurable in terms of features used,similarity metrics and parameters for them.
The timbl_wsd.py can be used for training a WSD model, to annotate a text with senses, or to evaluate a WSD model. The software is written in python, and to install it there is no any special requirement apart from having the python interpreter installed. The external dependencies of the system are:
1) SONAR corpus (http://lands.let.ru.nl/projects/SoNaR/) 2) TIMBL package (http://ilk.uvt.nl/timbl/) 3) Pynlpl NLP library (http://github.com/proycon/pynlpl)
Once the packages 2 and 3 are installed and the SONAR corpus has been downloaded, we can run the system just by typing: timbl_wsd.py.
We can specify all the parameters of the system by means of arguments to the python program. A short description for all the parameters can be found in the package.