Font Size: a A A

Research On Fine-grained Video Tag Mechanism And Application

Posted on:2013-11-19Degree:MasterType:Thesis
Country:ChinaCandidate:X QuFull Text:PDF
GTID:2248330392457836Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Video tags refer to the phrases describing the characteristics of the video, which enableusers to search and recommend the target video. There are some problems with the existingresearch on video tags: Firstly, existing video tags describe general characteristics for thewhole video but not detailed video clips, causing that the description for the video is not soaccurate. Secondly, the video hotspots are added manually with lower efficiency, whichdoesn’t apply to the case with massive video hotspots to be generated. Lastly, there are alsosome defects for video advertising. The content of ad is unrelated with the video and theplaying time relatively fixed, not receiving expected advertising effects.To solve above mentioned problems, we propose a fine-grained tags system. The sys-tem proposes a new method to generate fine-grained video tags and a solution of generatingvideo hotspots automatically with the generated fine-grained video tags. In the system, us-ers could mark the wonderful video clips to generate fine-grained video tags. The systempartitions the tags and extracts the keywords of tags with text weighting algorithm. Thenthe system will weigh the keywords with negative exponent model and further weight thevideo timer shaft based on the weighted keywords. According to the weighted video timershaft, hot time points are chosen as the video hotspots. And every video hotspot is recordedas the time point and description words. The ads could be delivered on the related videohotspots with target to solve the problem that the ad subject and the video content are unre-lated.The experiments show that, the average time of generating hotspots is3.033s. The ac-curacy of extracted keywords for video hotspots which are generated with the improvedtext weighting algorithm is24.7%higher than traditional text weighting algorithm. Theproposed fine-grained video tags system could more accurately describe video clips fea-tures and realize the automatic generation of video hotspots, which improves the users’video sharing experience and provides data supports for directional advertising on videoclips.
Keywords/Search Tags:online video, video tags, fine-grained video tags, video hotspots, target ad-vertising
PDF Full Text Request
Related items