The Research Of Accurate Retrieval And Active Push System For Cross-media Big Data In Tourism Domain

With the rapid development of Web 2.0 social network,massive travel data have been generated on the Internet,resulting in information overload.Users need to make a lot of effort to obtain useful information,and users' demand for the search and push of tourist information has become increasingly higher.Researches on accurate retrieval and active push of cross-media big data in the tourism domain have quite important practical significance.The main work accomplished in this paper is as follows:(1)We use Twitter-LDA algorithm to train attraction-interest-subject model and analysis the topic of attraction comments.Through removing the noise from the data of attraction comments,the results of the word segmentation could express the meaningful words of attraction interest,which implemented the construction of high-quality attraction-interest-subject model.According to the characteristics of the data of the geographical photos,the density clustering was carried out,and a large number of meaningful tourist sites were obtained.According to the time that the tourists take photos,a great amount of users' travel trajectory was obtained.(2)We have proposed an accurate retrieval method based on the attraction-interest-subject model.We used topic query expansion algorithm to realize semantic query,which has improved retrieval accuracy and coverage.On this basis,the final retrieval result was obtained by calculating the semantic similarity of the feature vectors between user query and the subject distribution of attractions,which has realized the goal of analyzing users' query intent on the subject semantic level.Experimental results show that the proposed accurate algorithm gains a promotion of 22%in term of accuracy,compared with query expansion algorithms.After the retrieval result was obtained,the knowledge of attractions was shown from multiple angles to allow users to better understand the attraction information.Image retrieval was realized through cross-media samples such as texts and images to understand the user's retrieval intention.Users' browsing behavior was given different weight coefficients to map their interest subject distribution of attractions and to reflect their current information needs,through which the discovery of users' interest and personalized attractions recommendation were realized.(3)We have proposed a location proactive push method based on users' context information.The probabilistic model of tourist location push was established based on the features of users' context information and the algorithm of context-based tourist location fast push was proposed.The traditional PrefixSpan algorithm was improved to realize effective frequent sequence pattern mining.The tourist location push prefix tree was constructed to realize fast and accurate tourist location push.Experimental results show that the proposed CTLFP algorithm gains a great promotion in term of accuracy,compared with NPP and HTLFP algorithms respectively.(4)We have designed and developed an accurate retrieval and active push system for cross-media big data in the tourist domain.The system was divided into an accurate retrieval module and an active location push module.The accurate retrieval module includes interest topic retrieval,topic query expansion,attractions knowledge abstract,user profile photos,personalized attraction recommendations,etc.The active location push module includes functions such as frequent sequence pattern mining and travel location active push.The system has comprehensive functions and has created a user-friendly interface.