This week, blinkx announced that it had packaged its video search and metadata extraction technology, making it available to publishers and content owners as licensed software or on an SaaS basis. The blinkx Advanced Media Platform spiders video content, processes it, and makes it searchable. It also optimizes the video content by adding metadata so that Web search engines that rely on text to search for videos will have more information about each video. Although video downloads are increasingly popular, accurate video search is hard to come by. That's because Web search engines have no way to delve into the actual video in order to find out what it's about.
blinkx has made a name for itself in Web video search because it can analyze video content. When it
processes a video, the sound track is converted into text and any embedded text passages are passed
through an optical character recognition (OCR) engine. It categorizes the content and analyzes the video
for scene changes. Based on this analysis, it creates a "thumbnail" description of the video that is
searchable by their Conceptual Recognition Engine (CoRE). The result of this processing is threefold.
First, because of the thumbnail description, the video itself can be matched more accurately against a
query, because that description is richer than the typical one or two words of metadata that most videos
come with. Second, the metadata that are added to the file make it possible for text-based Web search
engines to find it more accurately, driving traffic to a content provider's site. Third, because the description is more accurate, it's possible to target any kind of ad more precisely. This increases user satisfaction, but also the flow of ad revenue to a Web site.
blinkx has aimed its video processing modules at the consumer market. This typically contains vast
numbers of videos that are poorly tagged. Matching those videos on its blinkx.com site, as well as for a
variety of well-known Web search sites, brings them 4 million to 5 million searches each day. Their
business exploits a sweet spot. Better targeting of ads brings them higher CPM rates and funds their
business nicely. They match not only the familiar text ads but also banner, pre-roll, or overlays. With the
Advanced Media Platform, they are enabling customers to monetize video content just as text has been
monetized. One way of looking at this new product is that it provides search engine optimization for video,
just as many services do the same for text content, or provide it for video content, but manually. blinkx is
not manual, though. Its machine learning approach uses training data to train the system to categorize
and tag automatically. It can therefore process large volumes of video content once the system has been
trained. Because it uses a variety of clues — visual, text, and sound, its accuracy can approach 90%. In
the messy, voluminous world of consumer content, that is quite a leg up for content providers. As traditional offline publishers move online, they are eager to attract the online audience and to monetize their most attractive content.
One would naturally assume that some of the media companies would be the first to move to this platform. What is particularly intriguing, though, is that blinkx' first software customers are from the financial services arena: Kiplinger, WallSt.net, and Portfolio.com. Next to follow will probably be large consumer video producers — anyone with a high volume of video content that needs to be organized, processed, sorted, and monetized. blinkx can both serve and match ads, or provide highly targeted ad matching for ad networks. The system is compatible with third-party ad networks like Yahoo!, Google, or Tribal Fusion.