资 源 简 介
The latest code has been moved into Gibhub, and this site will no longer be updated.
BTM: https://github.com/xiaohuiyan/BTM
online BTM: https://github.com/xiaohuiyan/OnlineBTM
bursty BTM: https://github.com/xiaohuiyan/BurstyBTM
The papers can be found on my homepage: http://www.shortext.org
Biterm Topic Model (BTM) is a topic model developed for short text
(need to set a window length when generating biterms for nomral text),
like microblogs and webpage titles. It learns topics by modeling the
generation process of word co-occurrences (referred as biterms), rather
than word-document co-occurrences.
### 1. Model Description ###
In BTM, the distribution of a biterm b=(w1,w2) is
> P(w1,w2) = sum\_k{P(w1|
