Method for Product Selection Support Based on Reliability of Features in Reviews
Siqian Yu, Taketoshi Ushiama · 2021
When users selecting products on e-commerce sites, they always use reviews to guide their purchasing decisions. However, if several reviews are posted below one product, it might become difficult for users to select products efficiently. In this study, we propose two methods that use natural language processing technology, such as Word2vec and TF-IDF, to help users judge the reliability of features described in reviews effectively and efficiently. One method uses word embeddings and another method uses sentence embeddings. Finally, we evaluate the results of the two methods and find that the sentence embeddings method seams have a better performance.