The detailed instructions will be updated soon.
The sparse constrained LDA code is designed and built upon python package lda: https://pypi.python.org/pypi/lda, as it utilizes Cython for fast arithmetic operations.
It currently supports four types of constraints:
- Document Seed
- Document Pairwise
- Word Seed
- Word Pairwise
Reference:
-
Limin Yao, David Mimno, and Andrew McCallum. Efficient methods for topic model inference on streaming document collections. KDD, 2009
-
Yi Yang, Doug Downey, and Jordan Boyd-Graber. Efficient Methods for Incorporating Knowledge into Topic Models. EMNLP, 2015.