Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : September 20, 2022 - September 22, 2022
In this research, methods of extracting product usage and feeling of use from collected user reviews are investigated. Although most conventional data mining methods extract data based on commonality and similarity, we tried two approaches to extract data not only by commonality and similarity but also on differences. In the method using sentiment analysis, it was confirmed that there is a positive correlation between the result obtained from the sum of the polar values and the rating based on the number of stars. In the method using co-occurrence networks, a method of extracting features of two product was proposed. On this method, the results of the difference operation after normalizing the frequencies of words contains each products’ reviews are used to construct a network that integrates the co-occurrence networks of the two products, and to extract features from the obtained network. Finally, as a result of applying the proposed method to a portable fan and a tabletop fan, it was shown that the features can be extracted well.
