Tourism Cross-media Big Data Semantic Learning And Modeling

As a kind of development of Internet application,smart tourism greatly enriches the tourism information and the related fields.In this thesis,we aim at the cross-media characteristics of tourism big data,and based on the text depth representation model and convolutional neural network,we make a unified modeling and expression of the text and image information of tourism attractions,and propose the semantic learning and analysis of tourism cross-media data method to correlate travel texts and images based on deep features and topic semantics,and learn semantic information between tourism users and tourism texts based on deep neural networks.Experiments show that the proposed method can get better results for semantic analysis and cross-media search of tourism big data.The main work of this thesis is as follows:(1)Achieve cross-media travel big data access to information.For websites related to tourism information in the Internet,the tourism data are obtained by combining the relevance of tourism topic,the characteristics of time and space and the structure of web page,and the obtained travel texts and travel images are displayed.The accuracy of the number of travel webpages obtained based on TD-IA based on the topic,spatio-temporal characteristics and webpage structure is as high as 95%.(2)Realization of semantic representation of big data in cross-media travel.For the travel texts and travel images acquired from the Internet,the semantic features of tourism texts and images are extracted based on the travel corpus data using the text deep representation model and the convolutional neural network,and the semantic representation of the tourism travel images and images corresponding to the deep feature vectors is obtained.The accuracy of tourism image search method based on the deep features and tourism topic semantic classification(DF-TC)is improved.(3)Realization of semantic learning and unified modeling of cross-media travel big data.For the semantic depth features of tourism texts and travel images,semantic learning is conducted based on the potential semantic topics of tourism,and the training set of cross-media data is modeled using semantic analysis based on the topic of tourism data integration to achieve the unified expression of cross-media tourism big data.This thesis proposes a semantic learning algorithm based on deep neural network to construct semantic features of tourism users and semantic features of tourism attractions.The accuracy of cross-media search algorithm(CMS-WCTS)is improved.The DNN-TUSL algorithm based on deep neural network has an accurate rate of more than 73%.(4)Achieve cross-media tourism big data search.The test set of tourism cross-media data is preprocessed,and based on the semantic learning of tourism latent topics,a cross-media search of tourism big data test set is implemented,so that the research methods of this subject can better represent tourism texts and images and obtain better cross-media search results.Based on the above four research contents,the semantic learning and modeling based on the potential topic of tourism and deep neural network are realized for the Internet cross-media tourism big data,and the cross-media search of tourism big data is realized.