Research On Key Technologies Of Semantic Retrieval Based On Multimodal Data

With the development of science and technology,the Internet has penetrated into every aspect of life,and it has become a part of people's daily life to find the required information from massive network resources.Information data is exploding,multimedia data has developed from a single text in the past to many different media forms such as images,videos and texts,and network information is multi-modal.Because machines can't understand the semantics of information,traditional keyword-based information retrieval methods can't accurately provide information that meets users' real needs.Semantic retrieval has become an urgent need.Because it is difficult for computers to understand the complex semantic information expressed by multi-modal data,this paper uses OWL-S service description framework to label multi-modal data,describes text data semantically according to ontology concept,and uses ontology description model to establish the relationship between image and video metadata and high-level semantics.The annotation of video is based on the annotation of key frame images.Different from images,the object semantics of video is the moving object in video.Therefore,Gaussian mixture model is used to separate the foreground and background of frame sequence images to obtain the moving object,and then the video annotation documents are represented by the set of annotated key frame documents.In order to realize semantic retrieval,this paper proposes a semantic Web service matching algorithm based on improved Naive Bayes classification.In view of the fact that the traditional Naive Bayes is realized based on feature independence,this paper uses text clustering algorithm to obtain different categories of service description texts,and combines the feature weights of keywords in different categories with Naive Bayes algorithm,which can effectively reduce the influence of feature independence on Naive Bayes classification algorithm.At the same time,the ontology concept similarity algorithm based on distance is improved,which not only improves the timeliness of Web service matching retrieval,but also improves the accuracy of service matching.The recall rate and precision rate are used as evaluation indexes in the experiment.Through a large number of experiments,the ontology-based multi-modal data annotation method proposed in this paper effectively supports semantic retrieval,and the semantic Web service matching algorithm proposed in this paper improves the recall rate by 12%,the precision rate by 33% and the timeliness by 26%,which can better meet the needs of users in semantic retrieval.