Cross-media Education Big Data Intelligent Search Based On Deep Semantic Learning

In recent years,the rapid development of Internet and big data technology has penetrated into all aspects of social life,and has an important impact on social life.In the field of education,the development of education big data has become an important strategic choice to promote China’s digital campus construction and education reform,and education big data has become a new driving force that cannot be ignored in the field of education.The rapid development of Internet plus society has made Internet plus products continuously emerging.People are not only limited to obtaining information from the Internet,but also sharing information with the outside world through the Internet.On this basis,network users not only put forward higher requirements for real-time information resources,but also require more personalized and accurate search results.Therefore,for the emergence of a large number of texts,images and other cross media information in the network,we need to deeply mine the semantics contained in the information,and improve the quality of search and the accuracy of the results from the real needs of users,so as to realize the intelligent search of cross media deep semantic understanding.This thesis mainly includes the following four aspects:(1)In terms of the acquisition and feature learning of big data in cross-media education,a big data acquisition model for cross-media education is proposed to solve the problems of data redundancy and noise in the existing network platform in the field of education.Aiming at the shortcoming of the existing cross-media feature learning methods,which can only obtain the global semantic information and ignore the fine-grained semantic information,a cross-media feature learning method based on dual-pathway attention was proposed.For the acquisition of cross media education big data,we first build a keyword thesaurus related to education resources,then crawl and clean the relevant resources,and further refine the content.Finally,we get the data related to the field of education,including more than 27000 pairs of image text pairs.Aiming at the feature learning of cross media education big data,this thesis proposes a cross media feature learning model based on dual-pathway attention by integrating deep convolution neural network,attention mechanism and recurrent neural network,so as to fully learn the fine-grained features and context features of cross media education big data.Experimental results show that the map performance of this model is significantly improved compared with other feature learning methods in cross media search task.(2)Regarding the deep semantic relevance learning of cross-media education big data,in view of the key problem of the semantic gap that is difficult to bridge in cross-media semantic learning,a supervised adversarial hashing model based on dual-pathway attention features is proposed.Based on the existing dual-pathway attention feature learning model,combined with existing technologies such as generative adversarial networks and semantic hashing,it deeply explores the semantic associations between different media data,and combines feature learning with adversarial learning,hashing learning,and so on to build a unified semantic space of different media data.The experimental results show that the proposed method not only has a significant performance improvement on public datasets;but also has a significant performance improvement on the dataset constructed by itself,that is,the MAP on the NUS-WIDE,MirFlickr25K,and the educational dataset.(3)In terms of cross-media education big data intelligent search based on deep semantic learning,in view of the problem that user queries in existing search engines only start from the perspective of keywords and do not fully understand the user’s search intent,a query semantic expansion method based on embedded topic model is proposed.The embedded topic model is integrated with the query semantic expansion method to expand the user’s query and fully understand the user’s search intent.In addition,combining the semantic extension method with deep semantic learning,a cross-media education big data intelligent search algorithm based on semantic extension and deep semantic learning is proposed,and experimental verification is carried out on the public dataset and the constructed educational dataset.Experimental results show that the proposed algorithm has a significant improvement in MAP compared with the comparison algorithm.(4)Combining the three parts of research content:acquisition and feature learning of cross-media education big data,deep semantic relevance learning of cross-media education big data,and cross-media education big data intelligent search based on deep semantic learning,a cross-media education big data intelligent search system based on deep semantic learning is designed and implemented.Three functional modules in the cross-media education big data intelligent search system based on deep semantic learning are realized:the feature learning module of cross-media education big data,the deep semantic relevance learning module of cross-media education big data,and the deep semantic learning-based cross-media education big data intelligent search module.The function of extracting fine-grained features and contextual features of different modal data in the feature learning module of cross-media education big data is realized.The function of learning deep semantic correlation between different modal data in the deep semantic correlation learning module of cross-media education big data is realized.Realize the function of semantic expansion and search for different modal data in the cross-media education big data intelligent search module based on deep semantic learning.This thesis implements the acquisition and feature learning of cross-media education big data,the deep semantic relevance learning of cross-media education big data,the intelligent search of cross-media education big data based on deep semantic learning,and a cross-media education big data intelligent search system based on deep semantic learning is designed and developed.Both the experiment and test results show that the system realizes deep semantic learning and intelligent and precise search of educational big data,designing an easy-to-expandable algorithm interfaceand user-friendly interactive interface,and fully displaying the operating results of each module function.