Well after reading an interesting article today that Google applied for a patent that will allow them to develop a proprietary system to extract, index, and tag information and text embedded within Video I decided to do a little research myself for my readers…..

Apparently Google applies for patent applications based on employee ideas. But not all patents materialize into products or services.

In this case two computer scientist employees named Luc Vincent and Adrian Ulges are behind a Google backed patent application to make content embedded within images and video content searchable.

Although there is already a site called EveryZing.com, which I mentioned in a previous interview with TalkWithExperts, who is backed by military technology already has a video news portal that allows embedded text within video to be indexed and searchable. Google wants to develop their own video content searchable system by interpreting the content within videos by sophisticated analyzing algorithms.

Their patents seem to largely apply to extracting text embedded within images and making it searchable opening up a plethora of unique ideas including robots meandering up and down store isles using scanning imaging systems and making the content searchable for inventory.

However Google also seems to have its eyes on video and the way video is indexed and searched within portals specifically with YouTube which is most exclusive by users who have to physically tag their videos as well as input descriptions and titles for their videos.

Googles’ idea seems to do away with manual aspects from users who submit their videos and to integrate a sophisticated system that will analyze user uploaded videos and extract data from within the content and using that data to automatically create tags for the videos making it much more accurate.

Google scientist Luc Vincent is a leader of the on-going streetview project and responsible for Google book search as well as head of OCR initiatives. Ulges who is a Google German scientist working on his PHD is involved in helping develop a new video searchable technology system.

“We have developed a system that autonomously learns to tag videos with high-level semantic concepts by watching videos from online portals like http://youtube.com. Our system fuses evidence from several visual features like color, texture, motion, and patches in a robust manner.”

“Commercial systems seem to be neglecting the primary source which is the video content itself”

Most commercial systems like YouTube compile, organize, and make video content searchable Based on traditional user generated Meta data/tags, titles. Indexing methods, which require manual contribution, which is time consuming, incomplete and subjective.

So again it looks like quality content will prevail, and now this is going to not only apply to content on your web pages, articles, and blog posts, but with video as well. I mean can you see the trend here? EveryZing.com is already extracting content and making the text searchable and now Google is ‘kickin it up a notch’ by interpreting the content into searchable tags.

Example: A video on YouTube about a vacation resort with scenes of people swimming in the  beautiful ocean blue water, people laying on the beach getting a tan and the majestic Palm trees blowing in the wind.

No words are spoken only music panning the scene.
Google will analyze and interpret the video and extract its own keywords based on the video content. Google will spit out tags like swimming, beach,etc.

This is going to be interesting. So the times of entering irrelevant keywords titles and tags that you see so much on Youtube to get “seen” will be completely wiped out….

But this also means that your future video content needs to concentrate on specific aspects of a specific subject area and not combine too many different subjects into one video otherwise you could get tags that are not appropriate for your content.

References:

- Here is how Google German Scientist Ulges explains how Data is extracted from Video http://www.iupr.org/~ulges/

- Here is a link to a demonstration of how their YouTube dataset testing system works-
http://demo.iupr.org/videotagging/tagging-description.html

- Download Google Scientist Adrian Ulges .pdf called
Content-Based Video Tagging For Online Video Portals (.pdf)

Frank Bruno
http://www.Custom-Niche-Videos.com
http://www.VideoMarketingTactics.com