Banapana » search engine

Introducing Powerset

troped — Thu, 15 May 2008 21:49:47 +0000

Despite the fact that “Semantic Web” has been trending down (at least in Google searches), it seems like more web sites devoted to it are popping up these days. I’ve been messing around with Twine for a fews weeks now and it seems pretty useful, although it definitely can’t accomplish what they claim. ¹ A new one that I stumbled into, via KurzweilAI.net is Powerset. My 30 second review is that I tried searches on two pretty obscure subjects that I’ve been reading about lately, the St. Petersburg factor² and kriegsspiel. Basically, I got the same search results at wikipedia, powerset and google—no immediately discernible differences.

A few nice things do stand out about Powerset. The ajaxy integration of content into the search results page is definitely a step up from Google’s fairly wimpy page snippets. This feature alone would make it much easier to peruse results for accuracy and usefulness.³ I’m also impressed with the fact that Powerset appears to be gathering data from Freebase—an important open source database of information that’s a fundamental step towards creating a semantic web.

But if you ask Google and Powerset the question, “What is math?” you get pretty different answers. Google says:

a science (or group of related sciences) dealing with the logic of quantity and shape and arrangement

Powerset says:

MATH is an American Improv/experimental band formed in Woodstock, New York in the mid 90s.

While this may be true, I don’t think it wins the semantic claims any points. True enough, if you ask “What is mathematics?” you get a better answer. But shouldn’t a search engine claiming semantic abilities understand synonyms? Apparently, not yet. But I’ll definitely keep using Powerset and report back as I understand more.

In short, they claim that it can be a replacement for Del.icio.us and your blog and most things social. They don’t apparently understand some important rules about personalization in this regard. ↩
Weirdly, no one, including Google, had any information about this, which seriously makes me question its existence, since I’ve only seen one reference in a book so far! I’ll definitely have to post about it later. ↩
This reminds me, I really wish someone would create a feature that would allow you to eliminate search results from repeated searches. Like, let me save a search and then slowly cull it down to what I think are the relevant results. The data from those individuals who did the culling would surely be useful to the search engines. ↩

Go Trackback Yourself

admin — Tue, 04 Apr 2006 18:35:03 +0000

Most bloggers I read regularly pretty much write their entries and then move on. I myself, like to occasionally cull through what I’ve written about previously and from time to time update my entries. I think this in part stems from the fact that I’m often frustrated when I read a news story and a few months later try to find out how it resolved only to find nothing. It would be nice if news organizations regularly made RSS feeds for individual stories so that you could find the occasional update. Oh well. I suppose as long as thoroughness is dead, the least I can do is try to keep my own stuff up to date. And when I was thinking about how I wanted to go about this I realized that trackbacking isn’t just a nice tool for site-to-site interaction but also intrasite temporal interaction. Now when I update some entry that I wrote a while back, I trackback that entry with the new entry. That way anyone looking at the old entry (having come in from a search engine or what not) will know that new thoughts/facts are available.

Technorati Tags: trackback, blogging, blog

The LUI (Language User Interface)

admin — Tue, 25 Jan 2005 13:41:19 +0000

Yesterday’s post mentioned that I found this article (OpenP2P.com: Is Jabber’s Chatbot the Command Line of the Future?) on O’reilly right after discussing how I thought chatbots could become highly useful interfaces.

Today I just want to expand on that concept a bit. DJ Adams already does a great job of pointing out some of the things that bots do on IRC. Not being much of an IRC user, I had no idea that some of these bots could do the things DJ discusses. Among them he mentions note-taking, moderation and information provision. But what strikes me as one of the most powerful features of bots is the fact that they can get to “know” you. SmarterChild knows my zip code so that when I tell it that I want to see a movie the listings are for my neighborhood. A greater implementation of this might give SmarterChild the ability to discern if I’m at work or at home (although I can always change the zip code or even include it in the request). This knowledge is the very thing that search engine companies are all over right now. As was pointed out in Scientific American this month in “Seeking Better Web Searches” though “most people are unlikely to put up with the bother of entering personal data other than that required for their standard search activities.” Here then is your solution. Take the data in a little at a time with a bot.

Amazon’s recommendations are based on your purchases but a bot wouldn’t have to wait for you to buy anything if it coaxed you into telling it that you liked country music. This could occur through casual conversation but could also occur in other ways. A bot that asked trivia questions might be able to infer quite about you.

It starts to get a little spooky for us privacy advocates at first glance but the bot never had to know your name and it’s easy enough to get a new chat account or change it. Since chat accounts are free there’s also no reason to give them your “real” information (i.e. name, address). So all the bot really knows is a zip code and an account name and some preferences — nothing anyone could trace back to you.

Smart agents and an agent web have been talked about for some time but it looks to me like instant messaging interfaces to bots (and maybe SMS too) might be the interface to finally bring them to life.