Enhancing the HL-SOT Approach to Sentiment Analysis via a Localized Feature Selection Framework

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 327–335, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Enhancing the HL-SOT Approach to Sentiment Analysis via a Localized Feature Selection Framework Wei Wei Department of Computer and Information Science Norwegian University of Science and Technology wwei@idi.ntnu.no Jon Atle Gulla Department of Computer and Information Science Norwegian University of Science and Technology jag@idi.ntnu.no Abstract In this paper, we propose a Localized Fea- ture Selection (LFS) framework tailored to the HL-SOT approach to sentiment analy- sis. Within the proposed LFS framework, each node classifier of the HL-SOT ap- proach is able to perform classification on target texts in a locally customized index term space. Extensive empirical analysis against a human-labeled data set demon- strates that with the proposed LFS frame- work the classification performance of the HL-SOT approach is enhanced with com- putational efficiency being greatly gained. To find the best feature selection algorith- m that caters to the proposed LFS frame- work, five classic feature selection algo- rithms are comparatively studied, which indicates that the TS, DF, and MI al- gorithms achieve generally better perfor- mances than the CHI and IG algorithms. Among the five studied algorithms, the T- S algorithm is best to be employed by the proposed LFS framework. 1 Introduction With tens and thousands of review texts being gen- erated online, it becomes increasingly challeng- ing for an individual to exhaustively collect and study the online reviews. Therefore, research on automatic sentiment analysis on review texts has emerged as a popular topic at the crossroads of in- formation retrieval and computational linguistics. Sentiment analysis on product reviews aims at extracting sentiment information from texts. It in- cludes two tasks, i.e., labeling a target text1 with 1Each product review to be analyzed is called target text 1) the product’s attributes it mentions (attributes identification task), and 2) the corresponding sen- timents mentioned therein (sentiment annotation task). Recently, Wei and Gulla proposed the HL- SOT approach (Wei and Gulla, 2010), i.e., Hier- archical Learning (HL) with Sentiment Ontology Tree (SOT), that is able to achieve the two tasks in one hierarchical classification process. In the HL-SOT approach, each target text is encoded by a vector in a globally unified d-dimensional index term space and is respectively labeled by different nodes2 of SOT in a hierarchical manner. Although the HL-SOT approach is reported with promising classification performance on tasks of sentimen- t analysis, its computational efficiency, especially as d increases, becomes very low. Furthermore, as d increases it will have more chance to index noisy term into the globally unified index term s- pace so that the classification performance of the HL-SOT approach might be depressed. Hence, we argue that if a locally customized index ter- m space could be constructed for each node re- spectively, both the computational efficiency and the classification performance of the HL-SOT ap- proach would be improved. In this paper, we propose a Localized Feature Selection (LFS) framework tailored to the …

Content truncated to 500 words