OHRank： An Algorithm Integrating Mentality and Influence of Opinion Holder for Opinion Mining

One of the most important tasks in opinion mining is opinion element extraction such as opinion expression, opinion holder and opinion target from a review text. Majority existing methods for opinion mining only concentrate on the review text itself. The mentality and influence of opinion holders is usually overlooked in opinion mining. The isolation of between opinion element extraction methods and mentality and influence of the opinion holder will make it very difficult for users to decide which mined opinions are worth being utilized. This paper introduces a novel concept of opinion network and proposes a PageRank-like algorithm called OHRank. It is first attempt to analyze the valuable opinion from the constructed opinion network by integrating mentality and influence of opinion holder into opinion mining. This proposed approach has been applied to real-world datasets and initial experiments indicate that the mentality and influence of opinion holder and his/her extra activity are helpful for finding valuable opinion and that the OHRank method outperforms benchmark methods that ignore those information.