Tonight Ask launched two cool new features, both innovations based on its study of how its customers use the service, according to Jim Lanzone, who runs the company’s search efforts. The first, Zoom, builds on Ask’s original clustering technology but goes several steps further, adding “narrowing” and “expansion” on your search based on Ask’s Teoma technology. I got a preview from Jim, and while I have not really been able to bang on it, it seems quite cool. For example, a search on “the beatles” will offer “Beatles Lyrics” and “Beatles Names” as narrowing results, and “Beatlemania” and “Rolling Stones” as expansion options, among others. Another feature is “related names” which for the Beatles includes Elvis and all the four Beatles. Play with it, it’s pretty neat.
Secondly, Ask is rolling out an expanded answering tool. Now, when you put in a phrase that might be understood as a question (ie “deadliest snake“) Ask will do its best to offer the web’s best answer. Ask will bold and enlargen the words its algorithms have concluded is the best chance to be correct. (It will also offer any number of other possible answers.) But these answers still reflect Ask’s best attempt at discerning truth from what Lanzone calls “the wild west.” It’s a worthy caution, for when you ask “who killed JFK” you will get any number of responses. The first concludes in its bolded text: “The Warren Commission categorically stated that Lee Harvey Oswald was the killer of JFK and that he acted alone.” But if you click through to the actual page from which this text is lifted, you get a conspiracist’s dreamworld (or the truth, depending on your predelictions).
Stepping back from this, Lanzone says that in tests, the new web answer feature increased click throughs on the first result by 200 percent. “Our goal is to decrease the number of people who come to service and can’t find what they want,” he said. He added that it’s Ask’s goal to keep creating new features that, once sampled, will make folks dedicated users of Ask’s searchers. Just in time, for apparently Diller intends to start pushing a lot of new traffic Ask’s way in the coming months. Should be interesting to see how it turns out. My take on this is simple: Ask is resurgent, it’s got a strong service, it keeps innovating, and it’s got IAC behind it now. Don’t expect them to stay 25 points behind Yahoo and Google for long.
PS – If Diller wants to change Ask’s name to Jeeves, he better read Gary – someone else owns the URL.
I must say that though I like the idea of clustering I find the whole structure of the Ask web site rather clumsy. My first impressions are that visitors need to be educated on how to use the site – “your usual web results appear under the fold, sponsored listings below the main sponsor and our clustering technology to the right”. I think a better approach would be to mesh the benefits of both Google, where you have sponsored listings to the right and Clusty (Vivisimo) where clusters appear on the left-hand side.
The answer service is really useful. I particularly like the highlighting of the answer and not just the query terms. Good job, Ask.
If they really live up to the claim to discern truth from the chaos of the Web, then that would be the ultimate benefit. If anybody knows of questions for which false answers are rampant on the Web, please post and see whether Ask dregs up the right or the wrong answer.
Ask still needs to work harder on the prioritization of the main search results listing. Too-often the better sites are down the list and the top site is a lower-quality site that happens to use the keywords prominently.
Thought I’d chime in to shed some light on these comments.
Regarding the layout of the page:
Ask is super-clean and essentially laid out like a standard search site. But we also do a lot of usability studies for how we lay out our features. The main difference with our site vs., say, Google, is the placement of ads in the middle, and Zoom related topics on the right. This is purposeful – we are making that tradeoff consciously. If you see any eye-tracking study (we’ve done some internally and Enquiro has recently done one) of search users, they scan down the left side first to see if there is a relevant result, then immediately scan to the upper right. This is, in some ways, the “bail out” zone, or more constructively, the place they first look to iterate their search. Because less than 1% of users take advantage of advanced search, iteration is a primary need of nearly every searcher. Our Zoom product, which we’ve had for 3 years as “Related Topics”, is the #1 most used feature aside from Web search for precisely this reason. The location is essential to its utility. I do understand that some people prefer sponsored listings out of the way, but hopefully they will only show up if they’re relevant. The left-side, on the other hand, is the wrong location because that’s where people want to see their search results.
Obviously, these are generalizations over an entire user base, and individual preferences may vary.
2. To Jakob’s point about the top result ranking, I have to say I’m not sure what data he is basing it on…and I know Jakob is smarter than to rely on his own qualitiative judgment, or to generalize relevance assumptions based on one query or a handful of biased queries. In addition, our search technology works very differently than he is guessing. The #1 result is the most authoritative site in its “community” based on our local popularity method of ranking. A lot of times that will yield a far different result than Google – sometimes better and sometimes worse, but sometimes just different. This is incredibly valuable, in our view, since it’s a different editorial view of the Web.
Certainly, like other engines, we can still grow the quality of the results over time. The full potential of our technologies has not been realized and we look forward to applying more resources to fulfill that potential in the future.
Jim
It’s great to see you commenting on this post Jim. I think the idea that you view the web at “the wild west” is right on and that’s what we’re trying to solve too at kozoru. Obviously we’re taking a different approach than keywords, but the news that the click through rate went up 200% is particularly heartening.
In any event, best of luck with “Web Answers”. Actually, I wrote about it on our blog too. Check it out if you get a chance.
Best,
-jpg
Where is the be useful?
Wow. Very impressive.
Its a great site. I like this comment.
Good luck
Hi, interesting idea. Your goal to decrease the number
of people who come to service is a good one, I think. Maybe it will take some time to get people to use it