• by Steve Smith , Staff Writer @popeyesm, September 11, 2007

Video search is not an easy trick. I have covered this field for several years now, watching companies like Blinkx and Truveo evolve algorithms for indexing visual material and presenting it in usable ways. Every time I dip into the topic the quality of results I get from the major search engines remains marginal at best.

If I come to the PC in the morning and type in “Letterman,” the odds are pretty good I am looking for last night’s monologue or Top Ten List. Pop that term into Yahoo and I get a hit for a documentary called “Electric Jihad” and some link to a standup routine from 2004. Run the same search on Google and I do get more clips from Letterman. but most go back several years. Oddly enough, if I type just the first few letters of the query into Veveo’s new vTap mobile video search engine, I get a folder labeled David Letterman and some of the most recent Top Tens, featuring Hillary Clinton and Bob Barker. I don’t get last night’s list, but I get closer to my goal on mobile than I get online. If I were Google and Yahoo, I would be wondering how “those little bastards” did that.

I don’t know what spidering techniques are going on in the background of vTap.com, which launched Monday for the iPhone and Windows Mobile platforms (Java to follow), but it does an impressive job of delivering video clips that are closer to my desires than the big two Web engines. If I tap in “PAV” for instance, it uses incremental searching to narrow down the possibilities in real time, so a folder of Luciano Pavarotti surfaces by the time I type the third letter. Entering that folder offers up clips from the recent funeral in Modena. Likewise, a search of “LENO” or even “JAY” brings up a Jay Leno folder with clips of Fred Thompson’s candidacy announcement. “BRITNEY” gets me Sunday night’s MTV Video Music Awards fiasco. Even the embedded YouTube client on the iPhone didn’t deliver results so close to my intent.

I don’t bother mentioning vTap to fluff up the company. In many ways vTap looks and feels more like a very good tech demo in search of a buyer. There is no current ad model, although the company tells me it will monetize search eventually with contextual ads in the results pages themselves. There is a sharing mechanism for pushing the video to a friend, but the engine is weak on media browsing. VTap doesn’t let you browse laterally very easily into other clips from the same media source, for instance. Nor can I subscribe to shows in the iTunes fashion, or the way I can in some of the recent mobile video portals.

Video search experts tell me this platform is as much about browsing as actual searching. People search for a rough target and then need tools for exploring. In video search we are looking to be entertained, not searching for a specific answer. VTap tends to offer up a roster of the same hit from multiple sources. Again, it feels like a very good but incomplete tech demo. In fact, Veveo Vice President of Marketing and Business Development Guru Pai tells me the company is talking to carriers about becoming both a direct-to-consumer product as well as a white label solution on the deck.

What vTap does demonstrate is the way the disciplines required by a mobile application may help Web designers solve problems that continue to bedevil all digital interactive platforms. Incremental search routines are nothing new, of course. Even on mobile, I recall Verizon’s VCast music using a real-time iterative search to make its massive over-air catalog more manageable, and it worked well when I first tested it.

The input constraints on the new crop of media devices may force all digital design to find more efficient paths. Verizon’s FIOS division for IPTV, for instance, developed a new set-top-box video search tool that had to tackle problems similar to mobile search. The FIOS team not only uses iterative search to solve the discovery problem, but also folds into the search algorithm information about recent hot news and entertainment topics. The engine becomes smart enough to recognize the likely intent of a query at a given point in time. An engine should be smart enough to know that if someone types in “Britney” this morning, they probably want to see that cringe-inducing clip.  

We are so accustomed to thinking of mobile as the bastard midget child of the Web, it may escape our notice that the Web could use some of the discipline and restraint mobile invites. For instance, a content provider would have to be crazy to slap a 30-second TV spot onto a two-minute mobile video clip. The video podcast pre-rolls and mobile TV spots I have seen have been blessedly brief so far. And yet for some strange reason Web publishers seem to have no problem letting advertisers cram their video players with atrocious ad-to-content ratios. I was browsing news clips at a major TV brand online the other day and saw the same 30-second HP spot in front of every short story. I have to wonder if mobile video will come up with more effective pre/mid/post-roll models faster even than Web video… because we have to.

Just as the Web helped teach TV about brevity and interactivity, the narrow mobile pipe might force the bloated Web to think harder about filtering, personalization and easier interfaces. At some point, we may even start defaulting to mobile for their superior interactive experiences. Keep your eye on that midget bastard; it’s crafty.