Research On Exploiting The Evolution Of Fine-grained User Opinions In Product Reviews

With the rapid development of intelligent devices and social media,e-commerce platform has become an indispensable part of people's daily life.With the increasing transaction data,recommendation system can help users to filter products more quickly.In the current e-commerce field,online reviews and ratings information play an important role in shaping the purchase decisions of customers,so how to make better use of these information has become a key challenge.In recent years,many researches have been done to make proper recommendations for users,by exploiting reviews,ratings,user profiles,or behaviors.They dig deep into the user's preferences information,so as to make more personalized recommendations for users.However,the dynamic evolution of user preferences and item properties haven't been fully exploited.Moreover,it lacks fine-grained studies at the aspect level.It mainly includes the following three challenges.Firstly,existing works lack a deep understanding of the evolution(dynamic feature)of users and items.Second,it lacks comprehensive time dynamic analysis for users and items at the aspect level.Then,there is a pressing need to integrate aspect dynamic features of users and items into sentiment prediction.To address the above issues,we define two concepts of user maturity and item popularity,to better explore the dynamic changes for users and items.We strive to exploit fine-grained information at the aspect level and the evolution of users and items,for dynamic sentiment prediction.The main research contents of this article are as follows:1)We propose a novel problem of Aspect-based Sentiment Dynamic Prediction(ASDP).To address the problem,we define the concepts of user maturity and item popularity to capture the dynamic features of users and items,and exploit them for better sentiment prediction.2)We deeply study the dynamic changes(i.e.,gradual changes and sudden changes)in user aspect preferences and item aspect attributes,with three real datasets from both the overall-level and the aspect-level.3)We propose a new model named ASDP to dynamically capture the change patterns of user aspect preferences and item aspect properties with uniform time intervals.We also propose an improved ASDP+model with a bin segment algorithm,which sets time intervals non-uniformly based on the sudden changes.4)We evaluate the proposed models on three real-world datasets.Compared with the baselines,the proposed model has a significant performance improvement.For example,the improvements on the Beeradvocate dataset are about 2%-20%on the F1 score.