Quad-tuple PLSA: incorporating entity and its rating in aspect identification

  • Authors:
  • Wenjuan Luo;Fuzhen Zhuang;Qing He;Zhongzhi Shi

  • Affiliations:
  • The Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China,Graduate University of Chinese Academy of Sciences, Beijing ...;The Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China;The Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China;The Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China

  • Venue:
  • PAKDD'12 Proceedings of the 16th Pacific-Asia conference on Advances in Knowledge Discovery and Data Mining - Volume Part I
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

With the opinion explosion on Web, there are growing research interests in opinion mining. In this study we focus on an important problem in opinion mining -- Aspect Identification (AI), which aims to extract aspect terms in entity reviews. Previous PLSA based AI methods exploit the 2-tuples (e.g. the co-occurrence of head and modifier), where each latent topic corresponds to an aspect. Here, we notice that each review is also accompanied by an entity and its overall rating, resulting in quad-tuples joined with the previously mentioned 2-tuples. Believing that the quad-tuples contain more co-occurrence information and thus provide more ability in differentiating topics, we propose a model of Quad-tuple PLSA, which incorporates two more items -- entity and its rating, into topic modeling for more accurate aspect identification. The experiments on different numbers of hotel and restaurant reviews show the consistent and significant improvements of the proposed model compared to the 2-tuple PLSA based methods.