Extracting product features from online reviews based on two-level HHMM

Author

Xiaoli Wang ; Zhang Lu

Author_Institution

Sch. of Econ. & Manage., Tongji Univ., Shanghai, China

fYear

2014

fDate

14-16 June 2014

Firstpage

1

Lastpage

4

Abstract

With rapid development of E-commerce, obtaining product features from online reviews effectively is both important consumers and product manufacturers. In this paper, we proposed a two-level Hierarchical Hidden Markov Model (HHMM) to extract product features. In HHMM-1, we use segment tags to divide comment text into Feature-Contained Segment and Non-Feature-Contained Segment. Then the product feature in Non-Feature-Contained Segment is further marked and extracted in HHMM-2. The experimental results of online reviews from Amazon show the HHMM method is very effective in product feature extraction.

Keywords

electronic commerce; feature extraction; hidden Markov models; Amazon; HHMM-1; HHMM-2; e-commerce; nonfeature-contained segment; online reviews; product feature extraction; segment tags; two-level HHMM; two-level hierarchical hidden Markov model; Data mining; Educational institutions; Feature extraction; Hidden Markov models; Information retrieval; Maximum likelihood estimation; Training; HHMM; data mining; product feature extraction;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer & Information Technology (GSCIT), 2014 Global Summit on

Conference_Location

Sousse

Print_ISBN

978-1-4799-5626-5

Type

conf

DOI

10.1109/GSCIT.2014.6970125

Filename

6970125