Research Of Sentiment Orientation Based On E-commerce Online Reviews

Along with the rapid expansion of the Internet,Internet technology is also developing rapidly.While bringing convenience to people's lives,it also changes people's consumption habits.Nowadays,more and more people will shop on e-commerce platforms such as JD and Taobao.They will also share their own experience on them or decide whether to buy or not through the evaluation of others.These review texts all contain the opinions of the consumers.By analyzing consumers' reviews,we can learn about the user's consumption attitude,which aspect of the product is more concerned,etc.Mining the potential information can help companies understand the attitudes and demands of users towards a certain product,which has great commercial value.We mainly studies sentiment analysis of user review texts in the e-commerce environment,taking mobile phone reviews as a case,and using sentiment analysis methods to judge sentiment tendency.The main research contents are summarized as follows:(1)The improvement of the traditional SO-PMI algorithm benchmark word selection method.The SO-PMI algorithm is a traditional method to expand the sentiment dictionary.After analyzing its principle,this article proposes a method for optimizing the selection of benchmark words based on its shortcomings and combining the characteristics and advantages of the TF-IDF algorithm,which initially improves the traditional The effect of the SO-PMI algorithm on the expansion of the emotional dictionary.(2)Based on the improved way of selecting basic words,the FU-PMI algorithm is proposed.The improved reference word selection scheme improves the effect of emotional dictionary expansion to a certain extent,but it has its shortcomings.Aiming at these problems,this paper introduces the community discovery algorithm Fast Unfolding,combined with the characteristics of the SO-PMI algorithm and the Fast Unfolding algorithm,and proposes the FU-PMI algorithm for subsequent emotional dictionary expansion.(3)Comment on sentiment classification experiments.First,we select Jingdong Mall as the data source to obtain the data of the mobile phone sector.After de-duplication and other processing,mark the comment data based on the score.Then classify these data based on sentiment dictionary to get experimental data.The experimental results show that the method proposed in this paper has a better effect than the traditional SO-PMI.Based on the experimental results,the characteristics of user reviews are analyzed,and corresponding reference information is given,which provides users with shopping and merchants to improve products.Certain guiding significance.