Akademik Veri Yönetim Sistemi
2024 Innovations in Intelligent Systems and Applications Conference, ASYU 2024, Ankara, Türkiye, 16 - 18 Ekim 2024, (Tam Metin Bildiri)
As the world is rapidly digitalising, there is a significant transformation from traditional shopping to online shopping. Consumer reviews are of high importance for the users when shopping from e-commerce sites. When making a purchase decision, users examine consumer reviews written about the product they are considering to buy. However, since most of the consumer reviews do not contain useful information, users lose time and effort in the process of evaluating the reviews. In this study, a solution to this problem is presented by proposing an intelligent system based on artificial intelligence that can automatically distinguish the useful ones among consumer reviews. Firstly, the authors defined useful reviews for online shopping websites. Accordingly, comments with positive or negative content based on experiences about a feature of a product that may be effective in users' purchasing decisions are considered as "useful reviews". Subsequently, 10, 000 Turkish consumer reviews were collected from a popular e-commerce website in Turkey by web scraping method. These data were labelled by the authors according to the definition of useful reviews. After preprocessing steps were applied on the obtained dataset, BERT models, a transformer-based deep learning architecture, were used in the classification phase. As a result of the experimental studies, it was observed that useful reviews could be distinguished with high performance. In the tests in which 10-fold cross-validation was applied, the best result was achieved in the test using the ConvBERTurk model with an F1 score of 97.04% and an accuracy of 97.02%. It is thought that the proposed intelligent system can be effective in solving the problem and will facilitate users' access to useful comments.