Code for fast scalable multi-facet intent classification of web search queries using dynamically constructed tree-structured graphical models. This code incorporates (Euro) WordNet, thus exploiting the semantic relationships in the lexicon to improve the classification results.
- WordNet (integrated as MySQL database) http://wordnet.princeton.edu/, http://www.illc.uva.nl/EuroWordNet/
- libdai http://people.kyb.tuebingen.mpg.de/jorism/libDAI/
- matlab (For plotting results. see mcode/ sub-directory)
- MySQL
To compile, goto src/ and run make. The following paths need to be edited for correct compilation:
- WN_ROOT=/usr/local/WordNet-3.0
- DAI_ROOT=${HOME}/Code/libdai
Optional - Only main.cpp uses matlab headers. Present iteration of the code has this commented out.
- MATLAB_ROOT=/chalmers/sw/sup/matlab-2009b
On njlaptop: If the path does not work, add to path the following: /
- See details of configuring for mac on http://developer.apple.com/internet/opensource/osdb.html
- If installing from port, the default location changes to opt/local
- /opt/local/var/run
- /opt/local/share/mysql5