Mining Hot Topics from Free-Text Customer Reviews An LDA-Based Approach

Author

Yu, Chuanming ; Zhang, Xiaoqing ; Luo, Huiting

Author_Institution

Sch. of Inf. & Security Eng., Zhongnan Univ. of Econ. & Law, Wuhan, China

fYear

2010

fDate

20-22 Aug. 2010

Firstpage

85

Lastpage

89

Abstract

This study examines how the Latent Dirichlet Allocation (LDA) model combined with natural language processing techniques can be used to identify hot topics from free-text customer reviews. To verify the validity of the proposed approach, 21 580 restaurant reviews are collected. Each review is viewed as a probabilistic mixture of latent topics and each topic is treated as a probability distribution over words in a vocabulary. Parameters are estimated with Gibbs sampling, and the hot topics with top words are acquired. The experiments show that this approach could produce satisfactory results.

Keywords

customer profiles; data mining; natural language processing; probability; vocabulary; Gibbs sampling; LDA model; LDA-based approach; free-text customer reviews; hot topic mining; latent Dirichlet allocation; natural language processing techniques; parameter estimation; probabilistic mixture; probability distribution; restaurant reviews; vocabulary; Dairy products; Feature extraction; Hidden Markov models; Internet; Markov processes; Probability distribution; Vocabulary; Gibbs Sampling; Hot Topic Detection; Latent Dirichlet Allocatio; User Reviews;

fLanguage

English

Publisher

ieee

Conference_Titel

Web Information Systems and Applications Conference (WISA), 2010 7th

Conference_Location

Hohhot

Print_ISBN

978-1-4244-8440-9

Type

conf

DOI

10.1109/WISA.2010.20

Filename

5581396