Take the case of videos on YouTube. Today’s videos are tagged with only important information that the video up-loader mentions while uploading a video. But this kind of service do NOT need any user to tag information explicitly (well its optional). The service recognizes the audio (speech) pattern and converts it into text and stores the information in a semantic database. It will make the video search even more tangible, effective, and concrete to the user than it is today. The user can search within the video for a particular word or text which was spoken at a certain instance of time. This feature has recently been introduced by Google in the form of Youtube Videos Captioning and the content search is enabled in Advanced Search options. Well, anyway, this can also enable the video streaming servers to give control of streaming to the user as well, hence saving their cost of streaming the whole video and the user can play the video only at specified points in video. It will help solve the problem of efficiently finding the video on the search engine (video host) AND to find the content with-in the video.
Illustration of the concept is shown below, here the keyword “weather” is shown to exist at different seek positions in the video content.
Let’s take a short scenario: The video of some conference is 1:00 hrs long and it discusses the topic of “Marketing” at some instant that people don’t already know that where it is discussed. So, he has to almost see the whole video to find out. Also, there is a problem of extra video streaming that will be done by the video server (which can be saved, if the user is only interested in a particular span of time within a video). Dont forget to read about innovation in online video players: Video Content Like Dislike Jumps