whuir / stm

The implementation of 'Effective Document Labeling with Very Few Seed Words: A Topic Modeling Approach', Chenliang Li, Jian Xing, Aixin Sun, Zongyang Ma, https://doi.org/10.1145/2983323.2983721

STM

The implementation of the Seed-guided Topic Model (STM), as described in 2016 CIKM paper:

Effective Document Labeling with Very Few Seed Words: A Topic Modeling Approach. Chenliang Li, Jian Xing, Aixin Sun, Zongyang Ma.

The STM is a topic model for the dataless text classification task.

Description

1. catalog file

Catalog file is uesd to describe category information. For example, the catalog file of data set '20 news groups' sholud be wrote in this form :

cate alt.atheism cate comp.graphics cate comp.windows.x .....

If we want to combine several categories into one , take 'comp' and 'sci' as examples , then the file should be wrote like this :

cate comp.graphics comp.os.ms-windows.misc comp.sys.ibm.pc.hardware comp.windows.x comp.sys.mac.hardware cate sci.crypt sci.med sci.space sci.electronics .....

remember that data file path must be same as category name. "20_news_group/comp.graphics/..." : a path like this is valid "20_news_group/othernames/..." : and this one may cause run-time error

2. seed word file

Just like catalog file , seed word file also use 'cate' as sign of the begin of a category :

cate windows dos microsoft ms cate graphics image gif animation tiff cate gay homosexual sexual ....

Make sure that the category order in catalog file is same as it in seed word file, like this:

cate cate comp.graphics graphics image gif animation tiff cate cate sci.space space orbit moon earth sky solar ..... .....

3. launch the program

The main Java entry is in class 'stm'. To launch the program there are several parameters must be specified , such like "model.testSetPath" which means the path of test path , and 'model.catalogPath' which means the path of catalog file.