GSDMM

Motivation

When I was working on the third homework of data mining course: clustering the short texts, I found this paper in Reference section which turned out be to the one recommended by Mr. Zhang in class. So I tried to implement the GSDMM algorithm proposed myself, of course, with the help of online resources.

NOTICE

This implementation is still on going.

Data Format

• vacabulary.json, with one word and its corresponding id each line.
• train_tokens.json, with one doc-id and its token list each line.
• train_topics.json, using for validation.

Reference

• Paper
  • Yin, J. and Wang, J., 2014, August. A dirichlet multinomial mixture model-based approach for short text clustering. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining (pp. 233-242).ACM.
  • Nguyen, D. Q., Billingsley, R., Du, L., & Johnson, M. (2015). Improving topic models with latent feature word representations. , 3, 299-313.
• Code
  • datquocnguyen/jLDADMM: java version
  • atefm/pDMM: python version