Improving the performance of E-Commerce Recommender System Using BART-Based Text Summarization

The rapid growth of e-commerce has generated massive amounts of data, increasing the need for personalized recommendation systems. Previous studies have used recommendation systems based on full review texts, but they suffer from noise due to unnecessary details and redundant information, as well as high computational resources. To overcome these limitations, this study proposes a model that extracts key information from review texts using the BART and predicts user preferences based on it. Text summarization can effectively extract important information from reviews to reduce noise, improve computational efficiency, and more accurately reflect user preferences. Experimental results using data from Amazon Video Games category show that the proposed model outperforms various baseline models in terms of MAE and RMSE. This study demonstrates that text summarization can improve the performance of recommender systems by accurately reflecting user preferences based on important information in reviews. In the future, we plan to evaluate the generalization performance of the proposed model using different datasets and state-of-the-art models. This is expected to provide more sophisticated and personalized recommendation services in e-commerce platforms.