Kozoru, a new question answering engine, is talking more openly about its approach to search on its home page. From their first blog entry:
We believe that kozoru is a necessary first step towards the next stage of search technology. Instead of entering keywords into the search bar, you’ll be able to ask a question and get an answer. In fact, kozoru doesn’t understand keywords. We know that might be a little hard to believe, but you can really only ask kozoru questions.
There are some big claims on the site:
After all, how can we do this in less than 9 months when others are still trying to solve this problem after 20 or more years? However, that’s probably the easiest question to answer.
Basically, we’re free to do whatever we want. It may sound odd, but the secret sauce at kozoru is that we aren’t constrained by keywords or neural nets or the traditional academic notions of what A.I. should do.
And I really feel that’s how it is with any disruptive technology. A few key people come together and say, “Why has everybody been doing it this way?” It was that way with the Manhattan Project. It was that way with James Watson. It was that way at PARC.
We’re betting it’s that way with kozoru.
Well, we’re waiting to see….
Barry and the folks at Rusty Brick have built a search engine relevancy test – essentially its a “blind” engine which gives you results randomly from either Yahoo, Google, Ask, or MSN, then asks you to rate the results (more on the details here). They call it “RustySearch” and Barry just pinged me asking me to get the word out – he needs a larger sampling of searches before he can draw any further conclusions. (His initial results are here.)
Pamela Parker interviews Dick Costolo of Feedburner, who we’ve worked with at BB and who I talk to frequently for Searchblog. Some interesting tidbits in here.
A happy surprise this week: my book is available for sale up on Amazon, and has been for some time. Its sales rank – at about 409,000, clearly indicates that it’s not exactly tearing up the charts, but given that it’s not even printed yet and I just found out about this, I suppose that’s to be expected. But it’s just so….real…to see it there on Amazon, with the cover and everything. Today the final, final manuscript came and I’m spending the next week poring over it, hoping to make it as good as it can possibly be. If you’d like to pre order the book now, why I’d be honored. Here’s a link to the order page. Just know that the information about the book is not yet accurate – it will be longer than 288 pages (don’t know where they got that, it will be more like 350), for example.
I’m terrified. But thrilled.
UPDATE: Holy crap! You guys pushed it up to like #5400 on Amazon! I guess that’s to be expected – start from a small base and work up, but wow. Thank you.
A new vertical search engine plans to launch this Monday, with a twist. Koders.com has the goal of helping programmers find open source code they need to do their job. From the site’s “about” page:
A significant portion of application development involves a process of find, copy, paste, and integrate. This process can be greatly accelerated when you can find existing source code that provides a solution to the task at hand.
Koders makes it easy for software developers to find existing source code that solves many common development problems with our vast index of working source code from a variety of open source projects. In many cases you may find code that solves the exact problem you are working on, and in other cases, you can find an 80% solution – where existing code can be suited to your needs with minor modifications.
This is a cool idea, though I have to admit I have no idea if it is an unsolved problem. Plenty of programmers/coders read this site, so let me know if this is the kind of service you might need or want. I do know that as someone who is about to embark on building an open source-based platform, this could certainly prove a valuable resource.
More details from a release I was just sent:
The company currently indexes over 190 million lines of code from more than 28 thousand projects in hundreds of open source repositories that users can search by keyword, language, and license or perform advanced queries using enhanced syntax. Search results are then displayed in a developer-friendly format that makes it easy to understand the code in its original working context.
Koders.com also features a unique Project Cost Calculator that presents a side-by-side analysis of leveraging existing code versus developing it from scratch. Not only does this enable easy assessment of build-versus-buy alternatives, it also provides developers with a quantifiable perspective of their contributions to the open source community.
After a successful campaign for Sun on Slashdot, Feedster is taking the wraps off an RSS ad network. For now, you can sign up, and that’s about it. More to come soon, I am sure….