| Nowadays,tourism has become the preferred way for people to relax and broaden their horizons.In China's national consumption,the proportion of tourism consumption is increasing year by year.Major tourism websites provide hotel services,ticket prices and tourism strategies for consumers,but also provide opportunities for the majority of Internet users to comment on tourism.These review data include feedback on the services,routes or features of the attractions.These feedback information can often provide decision support for other users when making strategies.But now the network data is huge,browsing data takes a lot of time and energy.Chongqing's special geographical environment has formed a catering and tourism industry chain.Throughout last year's China Business Intelligence Network news,the number of visitors reached a record high of 59,723.71 million in January-December 2018,an increase of 10.13% over the same period last year.At the same time,the increase of tourists has led to the economic growth of Chongqing,with total income of 434.415 billion yuan in 2018,an increase of 31.32% over last year.In this context,taking Chongqing's popular tourist attractions as an example,a tourist attractions evaluation system based on dependency syntax is designed and developed.The main research contributions are as follows.Firstly,based on Selenium crawler technology,a set of schemes for automatically obtaining comments is designed.In view of the limited amount of comment data of single tourism website,using Selenium tool,30000 comment data of Chongqing landmark sites on Ctrip and Honeycomb networks are obtained,and the function of timing update and acquisition is set up,which provides sufficient data support for the system.Secondly,the existing HowNet and NTUSD dictionaries of Taiwan University are combined and screened to obtain the basic emotional dictionary,and the expansion is completed with the help of the dictionary design algorithm of the "Synonym Word Forest Extension Edition" of the Information Retrieval Research Center of Harbin University of Technology.Aiming at the problem of inaccurate emotional classification caused by the small number of existing dictionaries,a basic emotional dictionary and adegree adverb dictionary are constructed with the help of HowNet and NTUSD Dictionary of Taiwan University,and a negative word dictionary is constructed by searching network.Based on Song Jingsheng's theory of comparing Chinese and English subordinate conjunctions,this paper constructs a dictionary of conjunctions,which eventually forms a positive dictionary containing 6440 words,8110 negative emotional words,213 degree adverbs,18 commonly used negative dictionaries and 47 word associative dictionaries.The accuracy of emotional analysis is discussed.Thirdly,based on the dependency syntax and the design of calculation rules,the emotional analysis of paragraph-level tourism reviews is completed,which improves the accuracy of classification in similar algorithms.The system uses StandFord Parse tool of Stanford University to extract the dependency relationship of sentences,and considers the co-occurrence position of negative words and degree adverbs and the influence of conjunctions on the emotional tendency of sentences when designing emotional rules.Through simulation experiments,the accuracy of emotional classification is improved by 4% compared with that of the method without considering these rules in document 12.Finally,Django framework and HTML5 + JavaScript + Python language are used to complete the development of the whole system.In view of the fact that the current tourism websites such as Ctrip and Honeycomb can only see users' specific comments on scenic spots,there is a lack of more interesting topics for users,such as the price,service and transportation of scenic spots,but there is no intuitive display.The system provides a view of specific scenic spots,and can browse users' emotional tendencies on various topics,which provides a platform for the visualization of emotional analysis. |
PDF Full Text Request