A London based company has released a beta version of its semantic video discovery engine improving video search results.
The search engine, that has recently been patented, breaks down a video into several parts and performs a deep semantic analysis of pictures, audio tracks, and text.
Instead of relying on video-creators to assign text-based tags to it, based on the video content, the new engine extracts such tags automatically and supposedly more accurately. It also generates a Google-compliant XML description of a video to increase search relevance and organic traffic.
The software, developed by Thinkglue - a young start-up company based at the Google Technology Centre in London – can create time closed captions and even flag content inappropriate for children.
The team behind the engine recently received seeding funds and boosts 40+ scientific publications, 3 patents filed, PhDs in Semantic and Media data analysis, and a management team with a past experience in Microsoft.
The beta version of the search engine has been released to a limited number of customers and is believed to help improve monetisation of media assets.