Design And Implementation Of Video Search And Recommendation System

With the continuous development of the Internet and the accumulation of video data,people have higher requirements for the effective acquisition of video.At present,general retrieval systems do not assess the quality of resources,nor do targeted searches for users,nor do they consider the intentions of current users when personalizing recommendations.In order to enable users to obtain videos with relatively high quality and greater relevance to users when retrieving video resources,and to enable users to obtain personalized recommended videos,speed up the efficiency of users in obtaining suitable resources and ease the user's retrieval effect The experience is not good,this project calculated and implemented a video website retrieval and recommendation system.extracts the user's historical retrieval text and the relevant text of each video resource,and uses named entity recognition technology based on information entropy and mutual information to obtain named entities in the video field as an important basis for subsequent word segmentation.By designing the inverted index table,the matching score of the retrieved text and the video resource is calculated.Calculate the quality score of the video resources by designing calculation rules,and use the word frequency-inverse word frequency(TF-IDF)technology to calculate the relevance score of the video and the user,and finally reorder the retrieved candidate videos according to the above three scores,according to Display search results in scoring order.This topic uses Text-Convolutional Neural Network(Text-Convolutional Neural Network)to classify the retrieved text,identify the user's retrieval intention,and obtain the popular videos under the intention classification as the video to be recommended.According to the user's historical behavior records,the user-based collaborative filtering algorithm and the video resource-based collaborative filtering algorithm are used to obtain two kinds of videos to be recommended.Finally,the quality scores of the three parts of the video to be recommended are comprehensively considered and sorted to personalize the recommended video for the user.The principle and implementation process of retrieval and recommendation has been introduced in detail.After completing each module,the functional test and performance test are performed on each module.The test shows that each module can work normally.The retrieved resources are consistent with the retrieval text.Video is reasonable.