Emotion-based Text Analysis Of Online Reviews Of Scenic Spots

With the rapid promotion and development of internet applications,a large number of tourist attractions review information has an important influence on the decisionmaking behavior of tourists.These travel review information appears in various forms of online media,which fully reflects the tourists' different ideas and preferences for attractions or services.Travel reviews have become an increasingly important carrier for sharing travel experience information.However,when tourists write relevant review information,there are strong subjective assumptions and the value of the review content is uneven.A large amount of redundant information makes it difficult for potential tourists to make correct travel decisions in a short period of time,resulting in information overload.How to reduce visitor repetition the time to search and rationally filter information and quickly make correct decisions are particularly important.At the same time,sentiment analysis based on online review data of scenic spots is of great significance for scenic spots to formulate correct solutions to improve scenic spot satisfaction.The main work of this thesis is as follows:(1)Propose a linear fusion algorithm based on SO-PMI(Semantic Orientation Pointwise Mutual Information)and Word2Vec(word to vector)to extend the core chinese sentiment dictionary of HowNet,and add new words in the field to construct tourism domain dictionary.This thesis designs the semantic rules and comprehensive emotional orientation score calculation method that conforms to the expression of chinese text,and uses the domain dictionary and emotional scoring rules to analyze the sentiment orientation of the chinese text.Experimental results show that the model achieves an accuracy of 79.3%.Compared with a single emotional dictionary construction model based on SO-PMI and Word2 Vec algorithm,it not only improves the accuracy of emotion recognition,but also overcomes the objective limitations of a single model.The reliability and effectiveness of the model are verified.(2)Propose a model of scenic spot satisfaction based on multi-angle sentiment analysis.The model first uses keyword extraction technology to obtain the factors affecting the satisfaction of scenic spots(ie scenic image);Then uses the semantic similarity method to associate the scenic spot image with the review data,and uses the sentiment analysis method based on the specific domain dictionary proposed in this thesis to calculate the emotional tendency of the review data,so as to calculates the emotional score of each scenic spot image;finally,a sorting method based on scenic spot satisfaction is designed to provide tourists with a sequence of scenic spots to choose from.The model not only comprehensively analyzes the scores of various factors such as the environment,service,and price of the scenic spot,but also displays the comprehensive score of the scenic spot,provides tourists with reference to the sequence of scenic spots,and provides customers with multi-angle selection guidance.In addition,the results of sentiment analysis also provide good suggestions to merchants in scenic spots to help them adjust their business models and strategies.The experimental results show that the comprehensive scores of scenic spots obtained by this model analysis are roughly consistent with the comprehensive scores of tourists on the ctrip platform.However,compared with the ctrip platform,this model can provide tourists with a more fine-grained emotional analysis of scenic spots,which proves that the model can For tourists' reference.