Earlier this week, I had the pleasure to speak with the CEO of Asterpix, Nat Kausik. Tomorrow, Asterpix will announce the launch of the web's first robot to crawl the web, discover and then automatically tag objects within the videos themselves. Nat was kind enough to speak with me and provide a comprehensive overview of this new technology, called, Asterbot.
Aterpix originally launched in December of last year as an excellent, free web tool for creating hyperlinks within videos, dubbed, "hypervideo”. Analogous to hypertext, hypervideo provides an interactive experience by allowing viewers to select objects of interest to get more information or navigate the video.
Nat provided me with a comprehensive overview of Asterbot, the web's first robot to crawl the web, discover and then automatically tag objects within the videos themselves.
Basically, Asterbot scans each video, discovered from crawling the web, in order to find the most relevant contents and mark them up with interactive and track-able "hotspots" which can have related ads, text, links, photos, keywords, etc…
More specifically, the technology scans the entire video as well as the page to which the video resides and analyzes the content in order to identify objects and rank them in order of relevance and significance in terms of the level of attention given by the camera.
The algorithm is designed to follow the assumption that the most relevant subjects within a video tend to be not only the main foal point for the camera, but also tend to be elements that reside within the the description and text on the page for which the video is embedded. (I mention that this is an assumption in that clearly not all content producers do the best job at describing their videos or including relevant text within the copy of the video page. However, I think that for the most part, this is a fairly safe assumption for some video content out there.)
Using this assumption, Asterbot breaks up the textual elements around the video content into clustered topics which it then ranks in order of importance. Asterbot then uses proprietary technology to identify objects within the video itself and matches these with the clustered topics from the on-page text. Using these learnings, objects within the video that are discovered and ranked, are then marked up with hotspots.
Here is an example of how this works:
The Asterbot will automatically recognize the killer wave and track its movement in the video. What's more, it will actually add more information about the wave—its height, its location, and even the link to its Wikipedia page. Take a look about 20 seconds into the video below and hover your mouse over the dotted rectangle hotspot.
If you want to see any of the videos that have been tagged this way, you can go to their site, and search for videos by Asterbot88. At the time that I spoke to Nat last week, they were crawling and tagging a few thousand videos per day. Asterbot88 has almost 23,000 videos since I last checked.
Nat mentioned that they may eventually package this capability into a technology component that can then be licensed.
Stay tuned for their announcement tomorrow!