Personalized Sentiment Analysis Of User Generated Content Based On Sentiment Dictionary

With the change of people's living habits and the emergence of 5G and other advanced technologies,online operations continue to occupy people's daily life.Due to the great change of shopping methods,online shopping has become the main way for consumers to purchase.There are many researches on e-commerce platform.First of all,most of the existing sentiment analysis methods of user generated content based on sentiment dictionaries focus on sentiment words,ignoring the influence of the text context,especially the modification of sentiment words on the sentiment color of the text;secondly,users usually make consumption decisions based on product descriptions,product ratings,online reviews and other user generated content,but due to the asymmetry of information,users When receiving these information,it is impossible to judge the authenticity of the information,so it is particularly necessary to identify the subject of user generated content.Finally,in the current research of sentiment analysis of commodity reviews,we usually only focus on the text itself,but ignore the personalized characteristics of the subject.Based on the above problems,this paper proposes the following solutions:First of all,for the single text-based user generated content,an sentiment extreme value calculation method based on sentiment dictionary is proposed.Compared with the traditional coarse-grained sentiment analysis,the Chinese dependency grammar analysis is added,and the existing dictionaries are expanded to make the Chinese text sentiment analysis more fine-grained,which lays the foundation for the next main body's sentiment expression intensity calculation and commodity comment's sentiment preference calculation.Secondly,from the perspective of users,the historical comments of the target users are taken as the research object.Based on the four features,the comment text representation vector is constructed,and a user generated content subject recognition model based on clustering is proposed,which realizes the distinction between real users and false users.In order to reduce the low value density of user generated content caused by false users' frequent publishing of false information,the identification model of user generated content subject is used to eliminate false users and ensure the authenticity of data.Finally,from the perspective of commodity evaluation,we take all the reviews of the target commodities as the research object.Considering the different sentiment expression habits of the comment subject,a personalized sentiment analysis model of commodity comment is constructed to eliminate the ambiguity caused by the different sentiment expression habits of the user and get the sentiment preference of the user for the target commodity.Sorting out sentiment preference scores in different time periods,drawing dynamic sentiment curve of the target goods,making the results ofcommodity sentiment analysis more intuitive,which is conducive to providing reference for other potential users.The innovation of this paper includes:(1)On the basis of HowNet dictionary,Chinese emotional words are added.From the direction and degree of emotional tendency deviation(QCNSP),which can accurately quantify the emotion contained in UGC.(2)In view of the historical comments of the target users,the evaluation time,evaluation length,emotional expression intensity and language habit factors contained in the user comments are extracted,the UGC feature vector with the dimension of 20 is constructed,and the user generated content subject identification model based on clustering is proposed to realize the identification of the true and false subjects by the judgment standard of the distance between the clusters.(3)Considering that different users have different sentiment expressions in the online environment,the same UGC contains different sentiment colors for users who do not use it.Therefore,taking the evaluation of the target goods as the research goal,the sentiment expression habits of the comment subject should be taken into consideration when conducting sentiment analysis of the comment on the goods.By combining the influence of user's sentiment mean on the sentiment analysis of comment text,the value of user's sentiment preference for goods is obtained,which makes the result of sentiment analysis more objective and real.