Our approach
The volume of clips to be reviewed demanded significant personnel time—more than 2,000 person-hours each month—with corresponding costs. Cognizant designed a solution that leverages artificial intelligence to automate the labor-intensive process of duplicate detection and classification. Our AI solution runs on Microsoft Azure in multiple geographies and incorporates a machine-based deep learning model that becomes more effective over time. It easily scales to handle the client’s progressively larger volumes of data.
There are two steps in the operation. First, it extracts low-level data from digital assets, such as video frames, audio and text, using various conversion models. This includes audio-to-text conversion, optical character recognition and image comparison, to process that data. Second, the solution compares and analyzes data from the original "reference" ad-clip with that of other videos suspected of being duplicates.
If it flags a duplicate, the solution provides additional information to help establish if it is an exact copy or an edited version. This information is fed into AI models to provide insights on how these clips are being used and how the media landscape is changing. The intelligence acquired enables advertisers, agencies and media owners to identify, target and reach key consumer audiences.