The implementation of “Fast, Flexible Models for Discovering Topic Correlation across Weakly Related Collections” (EMNLP 2015).

Disclaimer: this research code is nasty, lack of proper design or comments. Use at your own risk :-)

Prerequisites

Cython, CythonGSL

Usage

Please refer to the README in each model folder for usage.

Data format

In a corpus file, each line represents a document, which has words follow the collection_id, all separated by space.

<collection_id> <word_id> <word_id> <word_id>...

All ids begin with 0.

Vocabulary file

The vocabulary file is used by read.py, each line is a word:

<word0>
<word1>
...

So <word0> has id = 0.

---

If you use C-LDA/C-HDP for research purpose, please use the following citation:

@InProceedings{zhang-EtAl:2015:EMNLP2,
  author    = {Zhang, Jingwei  and  Gerow, Aaron  and  Altosaar, Jaan  and  Evans, James  and  Jean So, Richard},
  title     = {Fast, Flexible Models for Discovering Topic Correlation across Weakly-Related Collections},
  booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
  month     = {September},
  year      = {2015},
  address   = {Lisbon, Portugal},
  publisher = {Association for Computational Linguistics},
  pages     = {1554--1564},
  url       = {http://aclweb.org/anthology/D15-1179}
}