شماره ركورد كنفرانس :
3926
عنوان مقاله :
A Two level Probabilistic Topic Model
پديدآورندگان :
Rahimi Marziea marziea.rahimi@shahroodut.ac.ir School of Computer and IT Engineering Shahrood University of Technology Shahrood, Iran , Zahedi Morteza zahedi@shahroodut.ac.ir School of Computer and IT Engineering Shahrood University of Technology Shahrood, Iran, , Mashayekhi Hoda hmashayekhi@shahroodut.ac.ir School of Computer and IT Engineering Shahrood University of Technology Shahrood, Iran,
كليدواژه :
probabilistic topic modeling , LDA , Gibbs sampling , over , generalization
عنوان كنفرانس :
بيست و چهارمين كنفرانس مهندسي برق ايران
چكيده فارسي :
In this paper a new probabilistic topic model is introduced which can provides two levels of topics called specific and general topics. LDA model as a basic probabilistic topic model suffers from over-generalization and to the best of our knowledge this problem is unsolved. By adding another level of topics we tried to overcome this problem. Th e proposed model is applied to a corpus of 250 documents and 6143 unique words. Th e results show that the proposed model can produce more specific topics and also can produce clusters that are more similar to human-assigned categories.
