Wenjie Jia

Opinion Analysis of Product Reviews

With more and more reviews on the web, browsing through a mass of the related reviews becomes a heavy work. How to effectively analyzing and organizing these reviews attracts more attention. This paper pursues on the analysis of product reviews. It focuses on the product features that customer commented on and also whether their opinions are positive or negative. Different from the traditional method, we view the product features recognition as an information extraction task. Combined the domain knowledge and lexical information, we adopt the supervised method--Conditional Random Fields to find the opinionated features. For the identification of the opinionated product feature’s orientation, it mainly bases on the domain knowledge, and considers from three levels, including sentence, context and word level. Our experimental results show that the proposed techniques are effective and promising.

Multiple Factors-Based Opinion Retrieval and Coarse-to-Fine Sentiment Classification

Opinion mining is a growing interest task in both research and practical applications. It deals with the computational treatment of opinion, sentiment, and subjectivity in documents. This paper focuses on retrieving the opinion documents and giving their sentiment orientation. Mining and ranking the topic relevant opinion documents are implemented with a sentiment model, combining the existing knowledge and statistic information. Multi-level sentiment analysis approach is proposed to find the topic related sentiment information. Our experimental results on COAE show the effectiveness of the proposed techniques and the feasibility of classifying orientation at varying levels of granularity.