Research On The Fusion Of Sentiment And Opinion For Explainable Recommendation

Explainable recommendation systems not only provide users with recommended results but also explain why they are recommended.Most existing explainable recommendation methods leverage sentiment analysis to help users understand reasons for recommendation results.They either convert particular preferences into sentiment scores or simply introduce the rating as the overall sentiment into the model.However,the simple rating information cannot provide users with more detailed reasons for recommendations in the explanation.To encode more sentiment information,some methods introduce user opinions into the explanations.As the opinion-based explainable recommendation system does not utilize supervision from sentiment,the generated explanations are generally limited to templates.To solve these issues,we propose a model called SAER(Sentiment-opinion Alignment Explainable Recommendation),which combines sentiment and opinion to ensure that the opinion in the explanation is consistent with the user’s sentiment to the product.Firstly,SAER provides informative explanations with diverse opinions for recommended items.Secondly,to enhance the ability to learn semantics and generate explanations that are closer to reviews for the recommendation system,we propose a variant called P2SAER(Introduce Pretrained Language Model to Sentiment-opinion Alignment Explainable Recommendation).P2SAER utilizes pre-training and fine-tuning mode to improve SAER.In the pre-training stage,P2SAER learns semantics from a large corpus.In this way,it reduces resource consumption and improves learning efficiency.And in the fine-tuning stage,P2SAER encodes input representations in different ways to suit the optimization of the downstream task.Experiments on real datasets demonstrate that the proposed SAER model outperforms state-of-the-art explainable recommendation methods,and the introduction of review data information optimizes the efficiency of recommendation.We also analyze the case study,proving the connection between the sentiment and opinion in explanations generated by SAER.Specifically,SAER generates different explanations in different ratings.Moreover,we also compare P2SAER with other methods in the same dataset.It proves that the introduction of pretrained model speeds up the training phase and improves the effect of explainable recommendation.