Opinion Strength Identification in Customer Review Summarizing System Using Association Rule Technique

The purpose of this research is to study opinion strength identification in customer review summarizing system in Thai language. We expect that opinion strength identification will help refining results from customer review summarizing system and improving quality of the results. In this paper, we present results from our preliminary study that was conducted with an aim at creating rules for identifying opinion strength by using an association rule technique. We used a paper survey to collect human opinion towards words used in 180 reviews of three groups of cosmetic products: blush on, facial cleanser, and foundation (60 reviews per group). The survey was done with 225 samples and resulted in 11,686 phases that had effects on samples' opinion towards the products. Association rule technique was applied to words in those phases to generate opinion strength identification rules. In our further study, those rules will be used to extend a capability of a customer review summarizing system. We plan to compare an effeciency of the system which is extended with the opinion strength identification rules to the one that is not, and also to a summary result obtained from human.