Wednesday, May 23, 2007

Parents, teachers, and kids needed to beta test our search engine for kids

A recent comment on this blog prompted me to advertise that we need beta testers for the Piffany search engine. Though we intend to have an offering for tweens and teens in the near future, and believe it or not, we have something in mind for pre-schoolers too, for now our search engine will be tailored for kids ages 5-12. Unlike other search engines, we are just for kids and we are a real search engine. In fact, some of the technology under the hood exceeds that found in adult search engines. If you follow this blog, then you will learn more about that, but for now I would like to call on anyone interested in beta testing our product. If you're interested, let me know by clicking on one of the following:
We should be online for testing around June 1, 2007. In the meantime feel free to snoop around the website or watch our demonstration video (very amateurish, I admit, but it gets the message across).

Tuesday, May 22, 2007

You do plan on having a search engine on your search engine, right?

It’s approximately two weeks from our official launch date, but we were supposed to have everything online and ready to go on May 15, 2007 so that the press and investors could have a look at it. It’s a week after the deadline and still no search engine online. The reasons, of course, are the usual suspects: technical issues and uncertainty. It makes me wonder how anything ever gets done anywhere.

I understand that there are a few technical issues regarding the installation of the search engine front end—that portion of the search engine that users interact with. Only Ilan, who is in charge of this component, can fully appreciate those issues, but I have issues of my own on the back end—that portion of the search engine where search results are generated from inputs taken from the front end. I still haven’t put it in the code to allow for multi-token search, or half a dozen other things that will ultimately be needed for the official launch. I gave up long ago on seeking perfection in any single component of the project before moving on to the next component. Instead, I have adopted a strategy that is similar to learning. First cover a topic broadly with little depth on any single sub-topic (breadth first), and then revisit the same material, several times if necessary, each time going deeper (depth first). This works fine for the most part. Where it fails is when new knowledge gives rise to doubt about previous assumptions. A recent example of this process and its shortcomings are taken from my work over the past few days.

Yesterday, I completed the first objective evaluation of Piffany. The analysis compared the top search result for about two hundred different keywords—some general and some specific—on Google, Google Kids Directory, and Piffany. Scores for kids, teens, and adults were assigned based on the relevancy to topic and age. We totally dominated the kids and teens searches for keywords that were more school/homework related. However, we did not fare as well with keywords that were derived from pop-culture. A 16 year old searching for Aguilera resulted in The Christina Connection, an unofficial website about Christina Aguilera, rather than the official Christina Aguilera website, which was number one on both Google and Google Kids Directory search but only number three on Piffany. The official Christina website is entirely Flash based (animation based WebPages that work more like an interactive video than a conventional webpage). We haven’t put in a parser for Flash sites yet. Flash sites tend to favor the kind of sites where presentation is more important than content, and they also tend to get heavy use by commercial sites so it has been low on our list of priorities. But should it be? Many of the top results on Google for pop culture are done in Flash. I can’t really say whether the lack of Flash parsing is the reason Piffany scored lower on pop culture without actually putting in a parser for Flash and trying again.

Keeping with the conventional wisdom of not changing horses amid stream, I will add the Flash parser on the next iteration of version upgrades to the search back end. I think today I will put it the code to allow for multi-token searches. And before you know it, our search engine will have a search engine as part of its offerings.