AIML Bots

Local ANDYBOT

WARNING: ANDYBOT server will be undergoing personality audit and modification. Sorry for the inconvience but ANDYBOT is undergoing a period of expansion. This is to help make her more 'ANDY' like. The connections bewteen categories are not fully intergrated and so conversation may not be scattered at times. With the additional information she has also taken to asking questions about connections and relations between things.

Access ANDYBOT here. If your system supports javascript then ANDYBOT should be in a new window that just popped up. If the remote server is down, just close the popup. The Internet connection should be available full time. If you want to talk more than once then please register a username and password so she can remember you later.

This is an experimental AIML bot being tested for providing a prototyping shell for ANDY and to function as a site guide on android technology. Remember ANDYBOT is just a prototype used to test AIML development.

AIML and Bots

A.L.I.C.E won the 2000 and 2001 Loebner Prize for being the most lifelike machines for those years. The prize uses the Turing Test, named after British mathematician Alan Turing, to determine if responses from a computer can convince a human into thinking it is a real person. Richard Wallace created A.L.I.C.E using a library of over 30,000 stimulus-response pairs written in Artificial Intelligence Markup Language (AIML). AIML is a XML type language originally designed to have a HTML like syntax. This allows AIML to be created rapidly using existing editing tools, and to be quickly learned by those familiar with HTML.

The development of ALICE is based on the fact that the distribution of sentences in a conversations tend to follow Zipf's Law. While the possibilities of what can be said is infinite, what is actually said in conversation is surprisingly small. In fact, 6000 patterns cover 95% of inputs. AIML is designed to easily match and specify responses to these core patterns.

AIML is of interest in several ways:

  • Personality Prototyping: it is very easy to specify the responses, in order to prototype new personalities of androids and bots.
  • Information extraction: AIML provides an easy way to match patterns and extract the relevant information. This information can be used as a query to the web or databases, or can be stored as data in a database.
  • Natural Language learning: One area of AIML research is allowing AIML to create more AIML as the bot interacts with the environment, thus writing and extending itself. While proven to work, it assumes the users are trustworthy, an assumption that is violated on the open Internet.
  • Body Browsers: AIML does not specify the type of client used. So one can develop an interface to an android, and connect it to an AIML server. Doing so would allow existing AIML tools to be used on real robots. (For an example search for "BrainBo") We are researching this possibility.
  • Wearable Assistants: AIML can make both a good simulation and a working front end for audio only wearable computers. Using voice recognition and synthesis software, and a wireless headset, we have tested an audio wearable interface using AIML.

 

Alicebot.Org The main site for AIML based bots. Contains free implementations for various bot servers. Also includes access to mailing lists.

Alicebot.Net A major developer of AIML systems is Jon Baer. ANDYBOT uses one of Jon's earlier servers.

Alicebot and AIML Architecture Committee. Kino is on the committee. The committee defines the features and syntax of the AIML language used by all AIML-based Bots.

Project CyN

 Project CyN is Daxtron lab version of the AIML interpreter Program N. CyN combines AIML scripting with access to the OpenCyc inference engine. Allows access to the OpenCyc commonsense inference engine inside AIML chat bot code.