pandora charms mbt sandals shop monclers jackets buy runescape money tiffany co jewellery montblanc pens online ghd uk links of london jewellery online movie baratos mbts mbts shoe mbts calzature
EllaZ Logo

Chat Technology at EllaZ Systems

[ Privacy, Policies, and Disclaimers ]             

Dr. Alan Turing first proposed an imitation game as a test for Artificial Intelligence (AI) in the 1950's which has become known as the "Turing Test."  He believed that if a computer can convincingly imitate a human, then it can be declared to be intelligent.  Dr. Hugh Loebner established the annual international Loebner Prize Contest to determine the “most human computer.”  It was first conducted in 1991 and has been an annual occurrence ever since.  Members of the EllaZ Systems team have won this award in 1998, 1999, and 2002, giving them practical experience in the performance of conversation systems.

Perhaps the first “chatterbot” program was Eliza, which was created by Dr. Joseph Weizenbaum in 1966.  Eliza could respond with text to keyboard input, carrying on a simulated conversation.  Keywords found in user input were used to select prewritten responses.  She often re-used user input in her replies, creating the impression that she understood more than she actually did.  Some early users were actually fooled into believing that Eliza was an intelligent and understanding conversationalist.

Stanley Kubrik produced the landmark motion picture 2001: A Space Odyssey in 1968.  An intelligent computer, the HAL 9000, was a central character.  Astronauts talked to HAL as they would a human, and HAL responded with a human voice.  In addition to controlling spaceship systems, HAL could entertain by playing chess and conducting friendly conversation.  Since then, people have been waiting for the development of an intelligent computer they can talk to.

AI has not developed as quickly as early researchers envisioned.  Many successful special purpose examples of AI have been built, such as chess programs that compete at the highest level of play. However, we are still at least a few years away from a computer system with human level general intelligence.

A wide variety of individual, corporate, and collective efforts have produced natural language systems since Eliza was born.  Perhaps the best known collective effort is the GPL licensed ALICE chatbot.  This effort uses over 40,000 stimulus-response sets to take this technique to a high level. Kiwilogic is a commercial operation that provides natural language powered service and information systems.  Another commercial effort is Artificial Life, Inc. of Hong Kong.  A few automotive activities are using voice enabled systems for access to mobile phone and navigation functions.  Of course, there is also EllaZ Systems, a cooperative international effort headquartered in Tianjin, China.

The chatterbot algorithms developed over the years provide a starting point for natural language interaction between Users and AI's.  Basic chatterbot algorithms include simple pattern matching, phrase normalization, abbreviation expansion, typo correction, reuse of User input with tense and pronoun shifts, and so on.  Other basic strategies include the use of monologues, giving priority to function type responses, and limiting repetition.  These all help Ella produce  natural responses in ordinary English.

The imprecise natural language of humans provides a major challenge with figures of speech, synonyms, and implied phrases.  A combination of techniques can be used to address these ambiguities, but often imprecise and tangential responses are deemed satisfactory to users. 

Our approach is to produce an AI that relies on accomplishing a lot with the detail work of piecing together available algorithms and techniques.  Increasingly sophisticated analytical and responsive behaviors can be layered as we have time and ability to implement.  Ella will be a blend of creative content and analytical programming. As various choices are made regarding functions, content, and priorities, computer personalities are emerging as a new type of art form.

Despite the fact that human–level AI has not yet been achieved, we have the ability to provide many types of conversational interaction with computers. Progress in speech recognition, voice synthesis, and natural language interfaces makes a surprising range of conversational actions by a computer possible. Further, continued improvements in the cost, performance, and size of computer hardware components make such systems practical today for a wide range of applications.

   Copyright © 2007 EllaZ Systems All Rights Reserved