KB-LDA

This is my implementation of the work [1]. For any questions regarding the work, please contact the original authors of the paper.

Setup:

To get the project's source code, clone the github repository:

$ git clone https://github.com/madhavcsa/KBLDA.git

Dependencies:

Please install the following dependencies for KBLDA to run: numpy, scipy, scikit-learn, joblib.

Usage:

To use kblda, you can start with these following samples commands:

 $ python KBLDA.py <TriplesFile> <CIFile> <NPFile> <VPFile>

Positional Arguments:

• TriplesFile Enter the path to Triples File
• CIFile Enter the path to Hearst Patterns File
• NPFile Enter the path to file corresponding to Noun Phrases of Documents
• VPFile Enter the path to file corresponding to Verb Phrases of Documents

Optional Arguments:

• -h, --help show this help message and exit
• --alpha_R ALPHA_R Enter the alpha for SVO
• --alpha_O ALPHA_O Enter the alpha for Ontology
• --alpha_D ALPHA_D Enter the alpha for Documents
• --gamma_I GAMMA_I Enter the gamma for Noun phrases
• --gamma_R GAMMA_R Enter the gamma for Verb Phrases
• --K K Enter the number of Topics, default 100
• --iters ITERS Enter the maximum number of iterations, default 2000
• --Odir ODIR Enter the output directory to which the results will be saved
• --sampling SAMPLING Enter seq for sequential and parallel for distributed; parallel is faster and is default choice
• --Threads THREADS Enter number of threads

Refernce

[1] Dana Moshkovitz-Attias and William W. Cohen. Kb-lda: Jointly learning a knowledge base of hierarchy, relations, and facts. Proceedings of ACL, 2015.