| Summary: | 碩士 === 國立政治大學 === 企業管理研究所 === 101 === A growing number of consumers have written product reviews to share their own experience on the Internet. The development decreases information asymmetry between consumers and manufactures and causes e-word of mouth effect that firms could not ignore. According to the survey of more than 2000 adults in the U.S., 81% of Internet users had searched for product information they planned to buy at least one time. Between 73% and 87% Internet users said the product reviews influenced their purchase intention, especially in high involvement products.
Consequently, it is essential for manufactures to have the ability to summarize thousands of consumer product reviews into useful information in a short time. Thus, review opinion mining becomes an important issue in the recent years. In the field of review opinion mining, it is critical for manufactures to differentiate product features in terms of their importance. According to the data from aspect opinion mining, manufactures can determine which product feature significantly influences sales volume and customer satisfaction. Therefore, our research focused on identifying “critical product features.”
We found existing studies did not address the time-effects on product reviews. That is, consumer review might be influenced by the foregoing reviews. The time-effects will cause inconsistent between the overall score and the feature score while the data based on the traditional aspect opinion mining method. Our research took the inconsistent situation into consideration, and developed an imputation method for features missing in reviews. In addition, we analyzed the sentiment polarity of Canon digital camera (SX210, SX230, SX260 generations) on Amazon with GPA, MPA Matrix. The results clearly identify the positive and negative features in different product generations. Using the automatic sentiment analysis framework we propose, manufactures could find the critical features that receive very favorable responses from consumers or need improvement in an efficient and more accurate way.
|
|---|
