Research On Multi-dimensional Sentiment Computing Model Oriented To User Experience Quality

Under the background of experience economy,people's demands for product and service experience are growing,and product and service improvement are focused on improving the quality of user experience.The sentiment feeling in the service process is an intuitive reflection of user service experience.Paying attention to the user sentiment caused by product and service can better improve user experience quality.The UGC is a relatively true and objective description of what users see and feels,and covers the sentiment evaluation information of the product and service attributes from the user's perspective,which can connect product and user.At the same time,there are many product and service attributes,and how to use UGC text to determine the importance of each service attribute is a key issue.The KANO model can classify product and service attributes according to the nonlinear relationship between service attribute performance and user satisfaction,but it has strong subjectivity.Therefore,this article fully considers the service experience sentiment information in the UGC,mines the user's perception and needs of the service attribute,and combines KANO model to determine the importance of service attributes,laying a foundation for improving user experience quality.Firstly,in order to extract the service experience sentiment information contained in UGC text,this paper constructs a text-oriented PAD three-dimensional sentiment computing model,and converts the text into a computable PAD three-dimensional sentiment vector.Considering that prior researches have not fully considered the relevant text features that affect sentiment information,it will cause semantic ambiguity and expression conflict,which will affect the model accuracy.This paper fully considers the influence of different text features on sentiment in different dimensions,constructs a text feature project suitable for PAD continuous dimensional sentiment,and obtains text features that more accurately represent the PAD three-dimensional sentiment information of text.And the obtained comprehensive and non-redundant text features are used as model input to construct a more effective PAD three-dimensional sentiment computing model.Secondly,the PAD sentiment computing model considering multi-text features is solved.The independent variables of text features are extracted from the construction and selection of continuous dimension sentiment text features.Considering multi-channel feature embedding to enrich text representation,the multi-channel CNN-Bi-LSTM neural network model is used to learn the local and global information of different sentiment text features,so as to more accurately extract the independent variables of the PAD sentiment computing model,and solve the PAD sentiment computing model.In addition,considering the lack of PAD sentiment space corpus,the supervised PAD sentiment computing model with multi-channel features cannot be effectively trained.This paper constructs a semi-supervised variational auto-encoder framework,embeds the PAD solution algorithm of CNN-Bi-LSTM fused with multi-channel features,then builds a PAD threedimensional sentiment computing model solution algorithm suitable for small sample problems,which effectively alleviates the small sample problem.Finally,the continuous dimension sentiment KANO attribute classification model is established to measure the importance of each service attribute.The traditional KANO model generally relies on questionnaires to obtain user service attribute requirements,which is highly subjective and cannot accurately reflect users' true feelings.This paper digs out the user sentiment about the service experience from UGC,and combines the UGC data to construct a service attribute system to extract attribute words.Then the PAD three-dimensional sentiment computing model is used to obtain attribute-sentiment pairs,which changes the way of obtaining user service attribute requirements and the form of sentiment expression,then effectively compensates for the subjective influence of traditional KANO model.Finally,combining the attribute-sentiment pairs to compute the sentiment intensity,proportion and importance of service attributes.Based on this,a continuousdimensional sentiment service attribute KANO model is constructed to determine the importance of service attributes and provide a basis for improving user service experience quality.