Research On Tourist Sentiment Classification Mining Technology Based On Multiple Methods

With the rapid development of tourism and the Internet,"Internet + Tourism"brings convenience and benefits to the people,at the same time,it generates tourism big data at an alarming rate.Among them,the text of tourism big data provides convenience for tourists to express their feelings and exchange information,with its convenience,simplicity,rapidity and low threshold,it has become an increasingly important place in tourism big data and is expected to become the main source of future tourism big data.These textual data contain rich emotional information,mining these information can help to perceive the emotional state of tourists,obtain tourists'views and attitudes towards tourism products,and have very important commercial value and social significance for tourism planning,tourism product development and marketing and other tourism activities.The study of tourist emotion in the context of tourism big data has become a hot spot in tourism research,the emotional classification technology is the basis for constructing the tourist emotion evaluation model,the advanced technology or emotion classification method directly affects the effect of the tourist emotion evaluation model.Under the guidance of big data theory and emotional theory,this paper takes the text of tourism big data and tourist emotion as the research object,and combs the related research results of tourist emotion research,emotional analysis and emotional classification technology at home and abroad,through artificial intelligence logic/algorithm programming method,machine learning method and deep learning method have carried out multi-method analysis and exploration on the tourist sentiment classification based on travel text big data.The main research contents and conclusions are as follows:(1)This paper summarizes and analyzes the use of emotional dictionary methods in the study of existing,it's concluded that the technical paradigm of this method is generally:constructing emotional dictionary,word segmentation,designing emotional calculation rules,and calculating emotional values.The core of this method is to build an emotional dictionary and design emotional calculation rules.This method is simple,easy to implement,and the wide range of applicable corpus is its advantages.The shortcomings are:dictionary matching is difficult to improve its accuracy due to the richness of natural language semantic expression;the dependence on dictionaries makes it expensive to construct a high-quality emotional dictionary manually;and the design of emotional calculation rules depends on manual experience and poor generalization ability.(2)This paper summarizes and analyzes the use of machine learning methods in the study of existing,it's concludes that the technical paradigm of this method is generally:word segmentation,text representation,feature extraction and selection-selection of machine learning classification model and training model,and calculation of emotion values through model.This method uses statistical methods to extract feature items in the text,and its nonlinear features greatly improve reliability with respect to the linear characteristics of the dictionary.The emotional classification process based on machine learning is easier to implement and the amount of calculation is small,but it is easily affected by text feature selection and training data scale and professionalism,which will limit its generalization ability on complex problems to a certain extent.(3)This paper explores the application of deep learning technology in the study of tourist emotional classification,realizes the high-dimensional vector representation of words through Word2vec,and constructs the RNN deep learning emotional classification model based on Word2vec.The experimental comparison shows that the classification of tourist emotion based on deep learning technology can also achieve good results,and the accuracy rate reaches more than 85%.This method uses word2vec to automatically discover features when generating word vectors,and at the same time reduce feature dimensions,which solves the problem of word order information and semantic features.Text corpus of multi-domain training is easy to transplant to multi-topic use,practical and generalization ability is better,and more suitable for the study of tourist emotional classification in the era of big data.However,this method needs more data,lack of large-scale training data sets has become the bottleneck of using in-depth learning to carry out the research of tourist emotional classification.(4)In the practical application of tourist emotional classification,the use of technical means should not only consider its accuracy,error,advanced level,etc.,but also should consider the speed of its operation(time complexity),the degree of resource consumption(space complexity),whether the implementation and stability are acceptable.There are two innovations in this paper:one is to summarize and analyze the existing methods of using emotional dictionary and machine learning in tourist emotional research,and summarize the general technical paradigms of these two methods,which will provide help for the follow-up related research and enterprise application;The second is to explanatorily apply deep learning technology to the tourist emotional classification,and has achieved good results.It provides a new way of thinking for the study of tourist sentiment classification under the background of big data era,which has the significance of technical guidance.