I believe audio/video search is only useful in recruiting and sourcing if it goes beyond metatags to full text conversion and searchability. On that score, there are already big distinctions between the major players. Two main players with millions of hours of video, PodZinger (parent company BBN is a general Internet technology pioneer) and Blinkx, compete (and sometimes cooperate) with the video search component of the major search engines.

Most players offer RSS feeds so you can track new matching results as they are added to, or spidered by, the site. However, in the results list, Podzinger uniquely previews the text around your desired keywords and lets you jump right to those parts of the video when those words are said. Likewise, Blinkx automatically parses all text within videos and makes it searchable, too. (I ran tests myself to verify that the search terms can occur deep into the video and not just in some introductory metatag.)

Unfortunately, unlike Podzinger, Blinkx does not index audio files currently, nor does it show where your search keywords appear in the video, nor can you jump to those points. But Blinkx's special features are a preview mode that shows the first few seconds of each video and 2 lines of text captioning, which is nice for visually-oriented folks. Possibly more helpful from a searcher's perspective is the ability to create a video wall -- a series of thumbnail images tied to your favorite search terms (thanks for the tip from my colleague, Jim). Jim's post linked to a story indicating Blinkx powers the video search on major sites including AOL, ITN, Lycos and Times Online, and recently added parts of Microsoft's MSN and Live sites. It also indexes video for many major news services. That's helpful in the momentum department.

Nexidia uses unique speech recognition technology that processes audio and video faster, and breaks speech down into 42 phonetic sounds. According to search expert Gary Price in a recent article, it's "easier for an algorithm to tell one sound from among 42 than one word from among the millions in the English language." However, I have not found any public search engine/portal that uses their technology. To date, they seem to be selling to enterprises; I suspect this won't last long.

The rest are way behind in that they only search the metadata (typically, title and introductory caption), but they have a few redeeming features that the two leading players might want to integrate:

• Pixsy, thanks to a large number of major content partners, is the video and photo search technology behind various sites, most notably PureVideo. The company claims its technology searches for both photos and videos by “relevance”, “category”, “provider”, or “freshness,” though the public implementations at their powered partner sites only let you search by keywords (relevance) and category (genre).
• CastTV has a unique approach to metadata, grabbing more context than the others below do, and in unique ways, which is promising but can't be tested properly until it launches publicly with critical mass of content.
• Google lets you search for videos that can be purchased online by price range, video length and/or content genres (currently available in the US only). Unfortunately, relatively few videos are being genre-coded, so don't depend on the usefulness of that search criterion. What's better is Google video search lets you use some of its special commands (e.g., title: to search only within the video's title), which goes beyond the AND, OR, NOT, exact phrase search capability common to all the video search engines (Podzinger's interface doesn't reveal it also supports Booleans, but it does.)
• This is also true for Yahoo's Advanced Video Search, which also lets you limit results by video file format (multiple select checkboxes). Ditto for Microsoft, whose video search is in beta as is its Soapbox initiative to solicit voluntarily-contributed video. Google is ahead on the latter, but jumpstarted its repository with its recent high-profile acquisition of YouTube, the web's largest dedicated amalgamation of contributed video, with a high percentage of amateur videos.

The impact on recruiting search of these grass-roots solicitation efforts are unclear, but I would still bet that true Internet-wide crawling for video would be better search results-wise. None of the major search engines seem to be spidering for video or audio as well as they do for text-based pages. AOL claims to be, but the results thus far are anemic.

BUT HERE'S WHAT'S MOST IMPORTANT, AND MISSING

More important for sourcers and other professional researchers is that you need critical mass of business-focused multimedia content. Not just major media sources like MSNBC, CNN, etc., but also analyst conference calls about corporations, company product demos, industry conference sessions that have been recorded, etc. I realize much of this is proprietary content and unlikely to appear on any site (other than the host organization's), and certainly not on a free search engine until some major licensing deals come to pass. Unfortunately, all the action to date in that arena seems to be in the entertainment space (no surprise, since most of the aggregated video content of "quality" is from the cable and broadcast TV networks, and audio is from the record labels).

However, with some major players opening up their APIs to developers and the history of the major search engines being able to make deals to search special content (in this case, it will probably be some kind of pay-per-video -- see Google's foray four paragraphs above -- or packaged subscription and online advertising hybrid), I expect new innovations to speed up.

AOL's offering (thanks in part to buying Singingfish and Truveo for their technologies) which claims to have "millions" of videos is clearly only searching basic data: my search for 'earnings report' turned up only 25 results, and you know more than 25 videos on company earnings exist! It does not appear the full power of AOL's affiliations (also including Blinkx) have been implemented here.

Blinkx is much better, finding 18 video results for "earnings report" as an exact phrase (use quotation marks), or about 1,000 results for just having both words (no quotes). Podzinger only finds two video results for the phrase, but 198 if you add audio search (106 videos for just having both words, and over 5,000 results if you add audio).

However, Podzinger apparently uses some algorithmic estimate: if you search for software, it says "Results 1-10 of about 1205 for software", but as you go deeper into the results pages, you see that about number dropping to 925, and finally when you get to page 80 (i.e., "Results 791-800 of about 925"), you can't see any more results: the word Next at the bottom page number results links trail is no longer a link! This is disturbingly similar to Google's dirty little secret of estimating results (often showing a number in the millions) on its regular web search: if you try to go past result #1,000, there aren't any displayed! Google never goes past 1,000, probably a protective measure to boost its server performance. (Score a point for search engines like Microsoft's Live, formerly MSN Search, which keep on going and going...)

BEST PRACTICES AND TOOL RATINGS

As of now, the key seems to be a search string combining unique technology terms with company names on the search engines that parse full-text speech. Any video (or audio) file that contains all these terms is likely going to be a presentation by someone knowledgeable in the field. Some will be the individual contributors (e.g., the engineers during some tech podcast) and others will be the public-facing people (e.g., executives, product or marketing managers appearing in a news story).

The benefits are you get the names of various qualified prospects (if it's a group presentation), a lot of context around the people and where the technology and market are going, and the sources of the most relevant results may give you new channels to focus on, both for recruiting purposes and as leads for your sales/marketing people on where to market yourselves!

So I have to give the #1 slot to Podzinger for the targeted searchability and overall volume of results (thanks to audio), Blinkx slides in at #2, all the rest provided far fewer relevant results, though I do like what I'm hearing about Nexidia, so if they can make a public search deal or some reasonably-priced subscription model with major content suppliers, it might earn at least the #3 slot. Team that up with a major search engine and you definitely have a winner. Stay tuned!

So which of these folks do you think is doing it best, from the recruiting and sourcing perspective? Any other players of note? Please share your successes and frustrations with audio and video search.

Read Glenn's other posts on RecruitingBloggers.com.

This past week marked the launch of a couple of promising additions to groupings in the recruiting blogosphere.  RecruitingBloggers and TheDayInRecruiting.  These join other useful amalgamations that have achieved critical mass that I profiled in a previous blog post.

 

As long as the low barrier to entry remains, the recruiting blog count will rise.  While Shally may have a point about the signal-to-noise ratio, it is inevitable that a diversity in voices temporarily results in a cacophony.  Then some of these build enough steam (viral marketing buzz and/or a flurry of promotional spending and PR) to hit our radar.  Not all deserve our attention.  It is up to each of us to sift what's important, and realize that the "biggies" aren't always the best. 

 

To help in this effort, my Microsoft colleague Jim Stroud has made a few significant improvements on a concept I implemented (I won't claim "pioneered", but if anyone was doing this even earlier, I'd be curious) about 18 months ago, back when RSS readers were far less ubiquitous:  Making feeds of various selected recruiting industry-related blogs visible on a web page, so you didn't need a news reader.

 

In my case (at www.recruiting-online.com/newsfeeds), I just grouped them in a couple of categories with pulldown menus so you could select the feed you wanted to view the latest posts from.  What Jim has done at TheDayInRecruiting is:

 

The other new amalgamation is RecruitingBloggers, started by Michael Kelemen, better known to many as the Canadian Headhunter who helped get Recruiting.com rolling and left to start RecruitingAnimal earlier this year.  He has invited a number of recruiting bloggers he likes (the site is subtitled "The Shamans of Search") to contribute posts on an ongoing basis, which also may be crossposted on one's own blog.  I guess this falls under the category of group blog, which is becoming increasingly popular in order to build and sustain critical mass to help keep that buzz going.  Again, I appreciate being invited to contribute (my first post there appears today).

 

The cacophony is getting too loud, and we all need help.  I look forward to seeing more as this evolves.  I suspect the blogosphere will end up following the same rules as every other communications category in order to bust through media clutter, but I'm hopeful that Web 2.0 as well as non-technologic creativity will somehow facilitate visibility for useful yet underfinanced voices. 

 

Technorati Profile, Recruiting Blogs, Internet Recruiting, Recruiting Blogosphere

 

A: This is a superb example illustrating the benefit of Google's numrange search command, which is unfortunately missing on other major search engines. This command lets you find web results where numbers within a stated range appear on the page. The format is low-value, then two dots, then high-value, with no spaces. In this case, since you are apparently interested in four-digit values, you would use 1000..9999 (though you could go higher, of course).

 

Since you wanted people with this certificate number on their resume, we would add that to the standard Google resume template string as follows (for a deeper explanation of the keywords/commands surrounding ccie 1000..9999, see Google's help page, or for more specific recruiting-related detail, Shally Steckerl's recently-updated Google cheatsheet):

 

~cv ccie 1000..9999 -benefits -candidate -careers -eeo -eoe -example -job -jobs -opening -post -preferred -reply -sample -send -submit -template -your

 

Not bad, but the results aren't ideal because these resume pages also contain mentions of "Windows NT/2000", year numbers on their resume, etc. So you could use a more limited range to knock out 2000 and the likely year values to appear on resumes of current workers (past employment through "degree expected" dates), such as:

 

~cv ccie 2010..9999 -benefits -candidate -careers -eeo -eoe -example -job -jobs -opening -post -preferred -reply -sample -send -submit -template -your

 

That's a little better, but again, equipment model numbers and other noise -- particularly the final 4-digit strings within phone numbers (almost ubiquitous on resumes!) -- isn't yielding ideal results. Adding "cisco certified" to the above string only helps a little, and doesn't address the main problem.

 

If you stayed with me this far, you're in for the big payoff. For very pure (but fewer results), try this, which forces the number to be near the CCIE (regardless of whether it has a space, a #, or another word in between) but is otherwise forgiving with however the candidate may have formatted it on the resume:

 

 

But if you want the mother lode, take out the ~cv -- it's still going to find you web pages mentioning people with those CCIE certificate #'s, but those mentions are not necessarily on resumes! You won't have the full resume, but you'll usually have enough info to contact them:

 

 

P.S.  To submit your sourcing question for possible answer here (personally-identifying details removed), post a comment or email it to blogquestion [at] recruiting-online {dot} com.  To learn more about this and many other aspects of Internet recruiting and sourcing, check out the free samples from the Recruiting-Online.com course.

CollegeRecruiter.com offers the latest on internships and entry level jobs.

cybersleuthing, Internet Recruiting, Sourcing