Searching within Videos

07 Feb 2012

Video has become ubiquitous in today’s environment. With almost 50 hours of videos being uploaded every minute on sites like YouTube, we see videos everywhere (http://www.youtube.com/t/press_statistics). Videos are one of the top consumables on the Internet with social networking sites swarming with short videos. Within the enterprise too videos are making quick inroads. More and more training programs, seminars and talks are being video captured with the intention of being reused by new employees or for the purpose of revision by those who attended.

But there is a difference such training videos are long (usually 1 hour or more) and show only the speaker for a majority of the duration. This brings the problem of searching for relevant content in that long video. Consider a captured training video about writing good official emails. An employee wants to refer to a particular portion of the talk when the speaker illustrated different ways of salutation and recommended ways of beginning official emails. She knows that the speaker talked about it in the program but to find it in the video is a big hassle. She has to seek back and forth “ keep guessing” and try to remember where the speaker talked about the stuff she wants.

kPoint recognizes this problem and overcomes it by indexing as much textual information as possible. One of the key differentiators of kPoint is capturing presentation slides independently of the video. kPoint extracts text from the slides and time-indexes it in the kapsule. kPoint kapsules encourage quick and easy addition of bookmarks to make note of interesting points in the video. Viewers can also ask questions about specific portions in the talk and these are also time-indexed. The result is indeed valuable “no longer is it just an opaque hour long static video” it has been transformed into an interactive, searchable training module. The employee in the above example could search for the keyword salutation and kPoint would instantly take her to the exact point in the kapsule where the speaker talks about it.

In addition, kPoint also allows adding transcript to the kapsule which are seen subtitles overlaid on the kapsule. Needless to say, these are also exploited in search. We are also actively looking at ways to automatically transcribe the speech in a kapsule so that your content becomes more and more accessible to you.

Happy searching!

Enjoy this blog? Please spread the word 🙂