When Google spent $1.65 billion on YouTube, it did so to get out in front of an exploding Internet trend — on-demand video.
YouTube grew its monthly unique audience 559 percent from January to
September this year. The site last peaked in August with over 34 million unique
visits, according to Nielsen//NetRatings.
And YouTube isn’t the only on-demand property with huge numbers. In
September, MSN Video had a unique audience of 12.6 million, MySpace
saw 10 million and Yahoo Video got 5 million visits.
That’s a lot of video brought to you by savvy search engines and
social networks.
So why is video search so bad?
Compare Google’s news search to its video search, for example.
Searching for “Mets” in Google News will get you all the latest
content from the many external publications Google crawls that
contain the word.
But if you conduct a search in Google Video, only Google Video
content will appear.
And don’t hope to find a video if the word “Mets” isn’t in its meta-data.
Google doesn’t “crawl” through the videos themselves, but instead
catalogs the various meta-data collected around it, such as the
video’s title, comments, labels, descriptions and closed captioning.
Other video search engines have the same issues. They do not work
well because they do not crawl through each video’s content the way
Google’s Web search, for example, crawls through the content of each
page it indexes.
Good luck finding a specific video on YouTube by any means other than
searching for its title or one of the few keywords haphazardly tagged
to it.
For example, you’re not likely to find every video with a Ferrari in
it on YouTube by searching for “Ferrari.” Not even if somebody says
the word “Ferrari” in the video.
Videos have become popular with broadband adoption, and search
marketing has proven to be a lucrative business model. So why don’t
video search engines crawl the actual content of videos?
The answer is that people are working on it. It’s just that maybe they’re having a hard time, is all.
It’s simply very hard to write an algorithm that recognizes any kind
of content in a video, Alex Laats, CEO of video-search company PodZinger, told internetnews.com. That’s because it’s all speech and moving images and no text.
PodZinger is one of many trying to solve the problem. The company wants to use
speech-recognition technology developed by its parent company, BBN.
The technology works by turning each spoken word into a
keyword. Then it gives each keyword a confidence score and time stamp
and then tags it all to the video or audio file.
It’s essentially a much more comprehensive set of meta-data for users
to search. It’s like if everyone who ever uploaded a video onto
YouTube bothered to transcribe their videos into keywords.
More and more comprehensive meta-data might be the best that video search
engines can do, Gary Price, Ask.com’s librarian and director of online
information resources, told internetnews.com.
Price is just not sure if word-by-word speech recognition technology
is the best way to collect it.
He favors the speech recognition technology from Nexidia.
Price said what makes Nexidia’s algorithm different is its processing
speed and that it breaks speech down into 42 phonetic sounds.
It’s easier for an algorithm to tell one sound from among 42 than one
word from among the millions in the English language.
Given Price’s engagement with
the video search problem, you might wonder if his company plans to
make a grab for that brass ring anytime soon.
Ask.com likes to brag about market-share growth in its television
commercials, and there’s plenty growth opportunity in the video search
vertical.
PodZinger’s Laats puts the potential revenues of a successful video
search engine in the tens of billions of dollars.
“We’re very interested in any possible searchable information,” Ask’s
Senior User Experience Analyst Michael Ferguson told
internetnews.com.
Other companies have been more open about their video search
efforts.
You can start with the big three search engines.
Google tried to go it alone with Google Video but then relented and
bought YouTube last week. On both properties, however, you can only
search for videos uploaded onto that particular platform.
As for Yahoo, it re-launched its own video platform in June.
The site’s approach is similar to social search and tagging sites del.icio.us and Flickr, which happen to be Yahoo Video’s corporate siblings.
Microsoft launched a YouTube clone earlier this fall called Soapbox. Users can browse categories, find related
videos, subscribe to RSS feeds, and share their favorites. Soapbox
search is still as bad as the rest of them.
In the second tier outside the big three, AOL has made the biggest push. Their video search actually attempts to search the entire Internet.
Laats’ PodZinger is of course going for it and so are other smaller players, such as the aforementioned Nexidia, an oft-lauded company called Pixesy and another called Blinkx, which is already licensing technology to Microsoft and Lycos.
Don’t count the smaller companies out, because they get it, too.
“If you have search it creates a way to structure advertising,” Laats
said.
Effective video search engines give people what they are
looking for. It shows them what they want to see.
You can imagine how advertisers would love to pay to have their
products included on that list.
But again, that’s only effective video search engines. And so far,
there aren’t any Google-like successes in the mix.
Not even from Google.
