| With the continuous development of service economy,service quality has become the core competitiveness of service industry,and the research on service quality evaluation has attracted more and more attention.Traditional quality of service evaluation methods spends too much time in the process of data acquisition and data processing,with low degree of intellectualization,and the scale and breadth of data cannot be effectively guaranteed.With the continuous development of e-commerce and network economy,online comment data is growing rapidly and the scope of data is expanding.Under the impetus of natural language processing technology and web crawler technology,unstructured online comment data has become a breakthrough to solve the problem of traditional service quality evaluation.Based on the insufficiency of service quality evaluation,this study introduces natural language processing technology based on management methods,breaks the traditional idea of over-reliance on manual evaluation of service quality,and proposes a service quality evaluation model framework which integrates automatic data acquisition,intelligent data processing and real-time data analysis with online comment data as data sources.The model is mainly divided into three modules: text data preparation,fine-grained sentiment analysis and fuzzy cloud evaluation.Text data preparation module is the data support of the model.It is responsible for grabbing comment data from external websites and performing text preprocessing operations such as word segmentation and sentence segmentation to form standard fine-grained emotional analysis input data.Fine-grained sentiment analysis module is the data conversion center of service quality evaluation.It is mainly responsible for two parts.First,it uses annotated data sets to pre-train fine-grained sentiment analysis model based on certain domain,and secondly,it uses pre-trained model to convert unstructured online comment data into structured affective tag data.Fuzzy cloud evaluation module is the core module of service quality evaluation.It adopts the method of fuzzy comprehensive evaluation based on cloud model.It uses cloud model instead of membership function to fuzzily synthesize emotional label data and weight data.The result of service quality evaluation consisting of three digital features of cloud model is obtained.Expectation represents the score of evaluation,entropy represents the stability of customer evaluation,and hyper-entropy represents the randomness of customer evaluation.The core technology module in the research is fine-grained sentiment analysis module.The core of this module is to construct fine-grained sentiment analysis model by using labeled data sets.In the research,fastText model is used as the basic model.Several N-Gram features are mixed and phrase features extracted by dependency parsing(called DP features)are added to improve the classification effect.In order to solve the problem of data imbalance in original datasets,a method of constructing minority classes based on seed sentences(called MCSS algorithm)is proposed in this paper.By increasing the number of minority classes,the MCSS algorithm can balance the data set.The experimental results show that both the MCSS algorithm and DP feature can improve the effect of the fine-grained sentiment analysis model of fastText.Finally,the proposed fuzzy cloud evaluation model of service quality based on DPFastText is applied to the empirical analysis of catering data sets.In the empirical analysis,firstly,the initial weight set is constructed by analytic hierarchy process(AHP).Then,the pre-trained DP-FastText model is used to predict the emotional label of empirical data.Then,the cloud model is used to calculate the weight set and the number of cloud models with single factor evaluation.Finally,the fuzzy synthesis is carried out by combining operator.By comparing and analyzing the results of the empirical model,Meituan and DIAN PING,the feasibility and superiority of the proposed model in the actual service quality evaluation task are proved. |
