Use of Content Tags in Managing Advertisements for Online Videos

We consider the problem of managing advertisements (or ads for short) for ad-hoc online videos, such as those generated by end users. We propose a novel ad selection approach based on video tags. We first compute intrinsic values of individual video tags to each ad, called ad-specific weights of tags. The computed weight of a tag indicates the relevance of the tag to a given ad. Then based on the tag weights, we select the ad which is the most relevant to the current video which a user is viewing.Using lab experiments, we show that our tag-based approach is able to select relevant ads. For example, experiments show that 58% of the selected ads are among the top 5 most relevant ads in a pool of 64 candidate ads.