Open AccessConstruction of Dirichlet Mixture Allocation total probability model based on multiple class text analysis
Author(s) -
Xiaoming Liu,
Feng Zhang
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2031/1/012055
Subject(s) - latent dirichlet allocation , topic model , dirichlet distribution , inference , computer science , probabilistic logic , class (philosophy) , hierarchical dirichlet process , mixture model , statistical model , artificial intelligence , mathematics , data mining , algorithm , mathematical analysis , boundary value problem
The LDA model is a total probability generation model for analyzing a large number of documents. It extends PLSA, another text analysis model. In this model, each document is treated as a topic hybrid model, and the topic’s proportional prior distribution is a Dirichlet distribution. The LDA model does not reflect complex dependencies between underlying topics. Based on LDA, this paper introduces a new topic generation model, DMA (Dirichlet Mixture Allocation), which models document collections more accurately than LDA when documents are obtained in multiple classes. In this paper, we build a probabilistic topic model of DMA, use the method of variational inference to approximate each parameter of the model, study the model, and finally solve the estimation of each parameter.