With the rapid growth of e-commerce and increase in the volume of online user reviews, personalized recommender systems have become essential for helping users efficiently identify items of interest. Although user review sets offer rich information for capturing user preferences and item characteristics, full-length reviews sometimes introduce noise, redundancy, and high computational costs. These issues degrade the recommendation performance. To address these challenges, this study proposes a summarized review-aware recommender system (SRARS), which leverages abstractive summarization to extract essential information from review texts. Specifically, we employed the BART, bidirectional and autoregressive transformer model to generate concise and coherent summaries of user and item reviews, thus improving the representation of preferences and attributes. To further enhance recommendation accuracy, we modeled user–item interactions through an outer product operation that captures complex multidimensional relationships and refined these interactions using an attention mechanism to emphasize informative patterns. Experimental evaluations conducted on real-world review data from Amazon.com demonstrated that SRARS significantly outperformed the baseline models in terms of recommendation accuracy and efficiency. This study contributes to the development of review-aware recommender systems by integrating text summarization and interaction modeling to improve the performance under sparse data conditions.

With the rapid growth of e-commerce, recommender systems have been widely adopted across diverse online services by presenting products aligned with user preferences. Moreover, review-based recommender systems have been studied to alleviate the sparsity of interaction data. However, many studies directly use full review texts, which may contain redundant semantics or noise that is irrelevant to recommendations, thereby degrading data quality and recommendation performance. To address this limitation, this study proposes summarized reviews fusion for adaptive recommendation (SuReFAR), which predicts ratings by summarizing reviews into key information using a multi-summarization strategy. Specifically, SuReFAR utilizes TextRank and bidirectional and auto-regressive transformers (BART) to generate extractive and abstractive summaries of user and item review sets, respectively. Subsequently, we apply an attention mechanism to emphasize salient information within each summary representation and fuse multiple summary representations by adaptively controlling their contributions through a gated multimodal unit (GMU) to predict ratings. We conducted experiments on Amazon and Yelp review datasets, demonstrating that the proposed model consistently outperforms baseline models and captures user preferences more effectively via personalized summary representations.

The rapid expansion of online services has necessitated effective recommender systems that provide information aligned with user preferences and requirements. In e-commerce, user reviews contain rich information and are extensively used as essential data for recommender systems. Many studies have integrated reviews into recommender systems by extracting positive and negative sentiments via sentiment analysis. However, reviews contain diverse user emotions that cannot be sufficiently captured by simple sentiment polarity. Moreover, focusing solely on emotional information may overlook the specific topics and semantic meanings within reviews. To address these limitations, this study proposes a dual-channel emotion- and semantic-aware recommendation (DCESR) model, which simultaneously considers both emotional and semantic information embedded in reviews. The proposed model uses pretrained language models to extract emotional features for users and items and to summarize reviews to capture their semantic features. These two types of information are then effectively integrated using a gate-based fusion mechanism to learn user and item representations for rating prediction. The experimental results obtained from Amazon.com review datasets demonstrate that the DCESR model consistently outperforms baseline models. This finding confirms that an approach that integrates both emotional and semantic information is highly effective in accurately reflecting user preferences.

Personalized recommender systems play an essential role in enhancing user experience by accurately predicting user preferences. Previous approaches mainly focus on modeling long-term preferences or capturing short-term dynamics through sequential patterns, while few achieve an effective balance between the two. This study proposes Rec-SSP, a novel review-aware recommendation model that integrates long-term and short-term preferences through a gated fusion mechanism. Long-term preferences are extracted from aggregated user reviews, whereas short-term preferences are modeled by identifying sequential patterns from recent interactions at both the review and category levels. This multilevel design captures fine-grained opinions across items, ensuring a more accurate understanding of the evolving user intent. This study conducted various experiments on real-world datasets, showing that Rec-SSP outperforms baseline models. These findings demonstrate that balancing long-term and short-term preferences with multilevel sequence modeling can significantly improve recommendation accuracy across diverse domains.