Just like collaborative effort has made Wikipedia one of the best encyclopedias in the world, so can collaborative effort help us create the best bots ever seen.

To this end, I propose to open a Mediawiki-based site for sharing bot code.

The site will support several common languages, such as AIML, ChatScript and RiveScript.

Developers will be able to create a bot using a single page, for example, “AIMLBot:Eliza” can be the name of a page that contains AIML implementation of Eliza.

Additionally, to facilitate code reuse, developers will be able to use MediaWiki’s “include” features to include other pages and combine them to a single bot. For example, there may be pages named “ChatScriptTopic:Childhood”, “ChatScriptTopic:Movies” etc., and then a page called “ChatScriptBot:Morti” which will include all these topics.

The code will be available by a CC license (like Wikipedia), so that anyone can copy and paste it in their own bots.

Later, we can create sites that will compile and host these bots. So, for example, one can tell the site: “I want to chat with ChatScriptBot:Morti”, and the site will compile the bot (if it’s not uptodate in the cache) and start a conversation.

What do you say?

Well, it’s certainly worth considering, as far as I’m concerned. I’m not all that sure about “compiling” a chatbot in the manner you’ve described, but I suppose it’s possible.

Compiling is for a more advanced stage. For a start, we can just have a wiki site where people can work on bot code together.

Do you think it is possible to open such a site in chatbots.org?

That would be up to Erwin and the rest of the team. He should be online at some point soon, seeing as it’s just past noon where he is.

This was already tried with AIML a few years back but unfortunately died a death.
http://alicebot.wikidot.com

Yes, sadly it has been taken by a horde of spammers…

This doesn’t mean that we shouldn’t try again. Maybe we will have more luck with a MediaWiki system, which is currently more well-known.

Most common languages, such as AIML, ChatScript and RiveScript already have terrific support sites.  What about the little guy’s chatbot, newer and less well known than the contest level class of chatbots?

@Erel: great idea Erel! Thanks for bringing this to the table.

We’ve planned a wiki already ( http://www.chatbots.org/admin/viewthread/382/) but this is a great addition!

We also like to include transscripts, as that’s demanded by the community as well (http://www.chatbots.org/admin/viewthread/230/)

As Steve mentioned, there are quite a few initiatives out there. The difficulty with a good wiki:
1. it needs focus. Have new pages every week and content updated regulary should be the most important thing there is for the website owner. We’ll have to work towards version 3.0 of the website, introduce our new home page, and generate sufficient revenues, and then we can go on.
2. it needs good moderators. Like Dave here on the forum here, we will also need good moderators on the Wiki. People who are able to create a positive spiral in the community and love to learn as well, people who are passionate to the topic.
3. it need people who feel responsible for pages. Who seriously go in discussion with others when they don’t agree on the content of a page. That really makes the difference.

From a technical perspective, having a wiki is not really difficult (although I’d like to connect it to the Chatbots.org chatbot, which is complication factor of 1000% :-s ). To make it successfull, the organizational perspective is vital.

Would consider to have a role in the technical and organization development of the Wiki and later on in the content production?

I think this is a great idea, but I think you should really differentiate between code-sharing and a potential “crowdsourced” bot.  I’m really interested in the second of these because I think it has great potential if it can draw in significant numbers of contributors.  To achieve this, I think you need to choose an accessible language - I’m not qualified to judge, but maybe Chatscript or AIML have achieved this - and provide instant-ish results.  If you think about Wikipedia, a key ingredient of its success is that if you see something wrong or incomplete, it’s easy to correct and you can instantly see the fruits of your work.  With a bot, in addition to the ability to edit the scripts, this would also mean pointing people to the right place, i.e. which topics or rules led to the exchange they just had.  I think the ability to come in and chat with the bot(s) to try it out would be a real draw.

For people who want to help but don’t want to write script, it’s useful to have a “report bug” function which just logs the current status / exchange plus a comment about why the reply seems wrong / unhelpful.

Then, as pointed out already, you need moderators and enough people who care, to improve the quality in the face of vandalism or just misguided edits. 

We may be close to the point where the tools are available to crowdsource a really powerful bot, given enough “brute force” input.  Rather than spread the effort, I think a single new bot should be the focus.

Finally, you need to think carefully about the licence.  There are lots of variations of CC and the key question is whether you put the content under a “sharealike” restriction.  If you don’t, people can come in and lift sections to use in their own bots.  That’s great for a code-sharing site, but less good if you’re trying to motivate lots of people to come in and co-create a bot that can win one of the competitions.

@Erwin: It’s good to know that it’s on the plan

I will be happy to help in development.

Most of my wiki experience is in MediaWiki (as a bureaucrat in one of Wikimedia’s sites). I already have many templates, Perl and PHP functions for automatically handling MediaWiki sites.

@Oliver: I agree that it would be best to be able to instantly try the code. I am working on something like this now.

I think that a single bot is not enough because there are many different kinds of bots, for different purposes. But we can define several purposes and work on a single bot for each one.

@Erel: thanks for your offering! I’ll contact you when the time is ripe (can you say that in English???)

Erwin Van Lun - Jun 13, 2011:

@Erel: thanks for your offering! I’ll contact you when the time is ripe (can you say that in English???)

The phrase has slipped out of common usage, but it’s still considered a valid aphorism.

@Erwin, sometimes the English version just doesn’t sound right, right. I was just thinking this morning: I will pull my plan.

At last, I have a proof-of-concept - a site where you can easily edit chatbot code just like a Wikipedia article, and immediately chat with it:

http://chatbot.wikia.com/wiki/Chatbot_Wiki

It’s currently hosted on a free site, so there are a lot of ads. You can remove some of them by registering.

The main page has a guided tour, which will show you how to edit and chat with bots.

Currently the chatbots in that site are written in Javascript, as this is the only language to which I have a Java interface. I would like to add other languages such as AIML and ChatScript, but to do this, I will need Java interface to them.

Your feedback is welcome!

Erel, I’ve found you on LinkedIn; but, I can’t find you or http://trosinteractive.com on Twitter, which seems a bit odd since you’re in the “social knowledge” and “social analytics” business.  Anyway, you can find me on Twitter at http://twitter.com/mendicot .

Dave Morton - Jun 13, 2011:

Erwin Van Lun - Jun 13, 2011:

@Erel: thanks for your offering! I’ll contact you when the time is ripe (can you say that in English???)

The phrase has slipped out of common usage, but it’s still considered a valid aphorism. :)

I think the word you’re looking for is “idiom” :) Wikipedia says:

An aphorism (literally “distinction” or “definition”, from the Greek: ἀφορισμός, aphorismós, from ἀπό + ὁρίζειν, apo + horizein, “from/to bound”) is an original thought, spoken or written in a laconic and memorable form.’