Ask Gneeves?

eries writes: "On my way home today, I had a brainstorm that I want to share with the /. community. It's an idea for an open-source Web project similar to the Open Directory Project. The idea would be to provide search functionality similar to that offered by "Ask Jeeves" - users input English-language questions and then get back a list of potential resources that they could use to find the answer. I am currently too busy with other projects (although one of them such as enzyme.sourceforge.net, the engine that backs Catalyst Recruiting has a lot of related code) to really spearhead and maintain this project. I have, however, purchased the domain names that will be necessary to make it work: askgneeves.com (this time the N is silent). Anyone out there interested in working on it? Anyone think it's even a good idea? Is someone already doing this? I think that pretty soon we could have an open-source product that would be useful to 90% of the clueless users out there."

"The problems with Ask Jeeves are two-fold:

1. They only have a few paid editors who try and compile the list of questions and answers
2. They in general only provide a single link to a question that they already know the answer to. Big companies provide big money to make sure that their resource is listed first/exclusively and these big sites are not always the best sources of information. For instance the questions "Where can I buy an Ethernet cable" should take you to a site like Pricewatch and not to CompUSA in order to get the best deal.

The open source community can solve this problem easily. We allow anyone who wants to to submit question-and-answer pairs, and then we allow end-users to vote (or should we call it moderate?) on which answers are best suited to each question. Thus, the system grows as a result of its users activity, and is not really subject to editorial control. Even less editorial control that Slashdot, believe it or not."

Open source this, open source that

[legoboy]

Open source makes about as much sense in this context as it does with that ridiculous "Open Source Iridium" thing.

It sounds like a project which could work, perhaps even wind up useful, but where does "open source" fit in? With a normal search engine, this would be instant death as anyone could falsely inflate their rank, though that's not the case with your moderation idea (which would take a lot of design work as well... being open source also makes it easy for people so inclined to detect vulnerabilities, such as how to make yourself more likely to have 20 different moderating accounts, etc)

I don't know that you'd attract enough volunteers to make it a truly comprehensive thing. I've never been terribly impressed with Ask Jeeves, and I think that most people interested in doing such a thing would be better off helping the Open Directory.

However, if someone thinks they can make it work and manages to do so, more power to them.

------

Re:Open source this, open source that

[dlc]

• Open source makes about as much sense in this context as it does with that ridiculous "Open Source Iridium" thing.

The phrase itself is ridiculous, but the meanging is not. The meaning is, "Let's make this a community project, in which anyone and everyone is allowed to contribute, according to a set of rules patterned after Open Source software."

darren

Cthulhu for President! [cthulhu.org]

A natural language parser and an API

[dlc]

Rather than an attempt at a comprehensive "Ask Jeeves" replacement, perhaps a generic version of the software would be more appropriate. A natural language parser, and natural language query engine, with conversion utilties to go from this format to a variety of backends (RDBMS, maybe LDAP, perhaps even a front-end to commercial/popular search engines).

I think the Open Directory Project [dmoz.org] does a ot of what you have mentioned; what is missing is the ability to feed it natural language queries and fully structured sentences.

If the idea for a new database/repository were to go forward, a very important aspect would be making it very easy to contribute. Perhaps an API that can be easily embedded in software (something like CDDB does), with a dedicated protocol, and maybe even dedicated server software to collect the information.

darren

PS Why is the "n" silent?

Cthulhu for President! [cthulhu.org]

Opensource searching

[Grendel_Prime]

Actually, I have been working on this idea now for several months, and am not technically competent enough to work out the back-end architecture, so I have had to come up with user-driven ways to address the issue of keeping search results fresh and relevant. Your suggestion of opensourcing the search process is in a nutshell what I ultimately came up with, but in order for this to work, a community of search librarians (similar to the /. editors) would have to moderate the channels. I foresee the possibility of creating a shopping channel, several information research channels, and several news channels. The reason that /. works so well in attracting an audience is that it is topic-focused and does a great job of feeding its end-user input back into the channels it is providing to the user base. I would be more than happy to participate in the editorial part of running such a site, but could one of you guys configure this /. site code so that I can understand and use the administrative interface? -Grendel

Already here

[CentrX]

Something similar is already here: Everything 2 [everything2.com]. While it's not perfect, it's certainly a good resource, although not a Web-wide one.

Chris Hagar

MIT AI Infolab/START project

[Fastolfe]

Perhaps this could be a potential use for a beefed-up version of MIT's START system. It essentially does natural language parsing and is hooked into resources feeding it all sorts of geopolitical information, movies/entertainment (via IMDB), dictionary lookups, etc., etc.

http://www.ai.mit.edu/projects/infolab/ [mit.edu]

The system was built in 1993 and seems to be maintained to a minor extent, with features added and updated, but to get it to the scale required for something like this would require a lot more processor power I'd imagine as well as some serious expansion.

clarifications

[eries]

I submitted this story to /. a while ago, and it has been (briefly) discussed on another thread. Here are a few clarifications that came out of that discussion.

1. There is now a limited "beta" version of AskGneeves up at http://www.Gneeves.org [gneeves.org] which is based on some simple code.
2. Ultimately, it makes sense for AskGneeves to basically serve as a front-end to the Open Directory Project, rather than having to re-create all that content. The beta that I have up uses some open-source PHP code to run a simple keyword search against the ODP, but tighter integration would obviously be better
3. Some have commented that this does not, necessarily, have to be an OSS project. However, I think it should be, since that guarantees that the rankings and search results are not being manipulated. Also, it would be the first instance (that I know of) where users could contribute new features to a search engine. Don't like the natural language processing? write your own.
4. There have been many comments about how AskJeeves "sucks" and about how Linux hackers would not have any interest in a project aimed at "lusers" like this one. I think that this is a mistaken view, and I think that it would be cool for us to create a platform-independant app that would be useful to normal end-users, and explain to them just why open-source and open standards are good.

Thanks for everyone who's contributed feedback. Now, all we need is someone to maintain the project...

Want to work at Transmeta? MicronPC? Hedgefund.net? AT&T?

Beware getting sued

[Shook]

This is an interesting idea, but I would suggest changing the name. Companies can (and have) sued people for intentionally creating names that sound like (and can be confused with) their product. I think a non-Jeeves related name would be much more safer, and you wouldn't have to change it mid-stream if AskJeeves threatened you.

Re:Open source this, open source that

[drix]

He got caught up in jargon, but I think I understand what he's trying to say. True; the actual software to perform the searching isn't anything revolutionary, and releasing the source to it wouldn't place it head and shoulders above any other search engine. People tend to use "Open Source" liberally to signify a community based, distributed effort. In this case, that's of questionable worth here as well. Wouldn't it be better just to petition the Open Directory for natural language searches?

The Open Source iriduim thing really was lame, BTW :)

--

Why don't you just ask Google?

[Citrix]

Google [googe.com] ain't open source but it kicks butt. Why don't you save some time and effort and just make something that takes a search in the form of a question and then makes the question more google friendly. heck you could make it "ask anything" and let people select their favorite search engine.

I don't mean to piss anyone off but I think it doesn't have to be open source, just unbiased and free of any corporate agenda. The open should be a method of keeping you honest and a way to enhance your relationship with your visitors.

my 2 cents.
Citrix

Re:Why don't you just ask Google?

[carleton]

This message has no point, except to remove my incorrect moderation.

Try it...

[kcarnold]

Just try giving Google a question. It'll remove really common words, like "a", "for", "which", etc. (not sure exactly), and the others should get put at the bottom in priority, leaving Google with a set of important words from your question that it will search for, which is close to what Ask Jeeves does. A specific product will usually catch a Real Name, and be marked up. So natural langage queries stand a good chance of working already. The only problem might be that question mark at the end. Maybe Google could just have a mode that filtered out more words in a little smarter way, and remove the confusing (to beginners) escape of numbers, etc.