Emotional Analysis Of OTA Website Review Texts

With the development of economy,people's demand for tourism is increasing,making the online travel market is explosive growth.Therefore,Ctrip and Qunar as the representative of the tourism OTA website has accumulated a huge amount of user comments text data.How to get useful information from a large number of text data to improve the user experience has become an urgent problem to be solved.In this paper,the author analyzes the sentiment of OTA web site review text data.Specific work as follows:First of all,this article through the web crawling technology has obtained the comment text data of tourism website OTA as the research object and constructs the corresponding classification thesaurus and emotional lexicon.Because of the particularity of OTA tourism website comment text,the comment text data,some of the current open source emotional lexicon can't be very effective,but the mainstream emotional lexicon mostly two emotion classification,can't reflect the emotional tendency of user specific;nor will the user's emotional tendency according to the evaluation elements or by individual segments user preference for emotional tendency.So,this paper constructs a proprietary thesaurus classification according to the evaluation factors and emotional lexicon subdivision.Thus,it is more effective to obtain the user's emotional preference value.At the same time,it is also an indispensable part of this paper.Secondly,this paper proposes a depth learning model based on LSA(latent semantic analysis)and DBN(deep belief network).Because the traditional text feature matrix of text vector space based on the construction of the only reflects the frequency of text information in the information,semantic information and did not contain words in between the words hidden(such as: polysemy or synonymy etc.)so in the model fitting process the effect is often lacking.Therefore,the LSA method is used to decompose the original text feature matrix by SVD,and then the text feature matrix is reconstructed by selecting the number of singular values.Finally,the depth learning model of DBN is constructed based on the reconstructed text feature matrix,in order to effectively obtain the sentiment tendency of the text through the training of the text data.Finally,the six groups of data and models are designed to verify the validity of the model.Based on the results of the cross validation,the depth learning model based on LSA(latent semantic analysis)and DBN(deep belief network)has better performance.