Video Classification Based On Text Data Mining

In the era of technology continuous developing, multimedia data on the Internetsuch as text, sound, images and other various forms are filled with human audiovisual,combined with real-time broadcasts and a variety of electronic devices enabled userseasily play and store large amounts of multimedia content. However, with the explosivegrowth of these data, the share of the storage space is growing, which the video data isparticularly the most obvious. People want to manual processing, and analysis of thesemultimedia data becomes unavailable actual. Therefore, efficiently retrieval andnavigate vast amounts of video data and classified them, not only enhance userexperience, but also identify potential commercial value.The study found that the common video classification method based on visualcharacteristics, not only time-consuming, inefficient and costly. On the other hand, thetext messages are often found in the description of the video information, reviews, aswell as to provide a personalized tag data has become a widely used carrier. Textprocessing technology is simple, efficient. Thus, the video text information becomes themost direct, feasible and effective feature to video classification.In this paper, the design and realization of video classification system based onvideo text description and popular tags, the main research content is:1. To dig out the hidden video content information as the goal, combined with thevideo text description, text classification technology is applied to video classification.Design and implementation of video classification system, it could assigned new videoautomatically to its predefined categories according to the video description text.2. For feature selection, feature weight calculated to be the research focus ofVideo classification based on text data mining, proposed chi-square entropy featureselection method, suitable for video text categorization.3. Though video classification method based on the video description text featureis good performance, high efficiency, the semantic gap between low-level textdescription to the video content ca not solved. Because of folksonomy tags from socialnetwork usually carried with video content and category Information, combined with these tags and text mining, this paper further classify the current share videos online.4. Video text descriptions are not the opposite of video tag data, they arecomplementary and indispensable to each other. This paper presents a probabilisticmodel to compromise them, and experiments show that video classificationperformances to achieve best while the probability value is0.5.