Elizabeth - by Peter Millican

Elizabeth is an automated conversation and natural language processing program designed to provide an enjoyable and easy introduction to natural language processing. However, she has sufficient power to permit the handling of complex grammatical transformations and resolution theorem-proving. Additionally, she has a very wide database of the more straightforward pattern-match/substitution responses.

Elizabeth can be thought of as an adaptation of Joseph Weizenbaum's ELIZA program, in which the various selection, substitution, and phrase storage mechanisms have been enhanced and generalised - increasing both its flexibility and its potential adaptability. The system also incorporates analysis tables to show exactly what processing has taken place, thus providing a learning tool that can give insights into some of the fundamental methods and issues of artificial intelligence.

Elizabeth starts the conversation by randomly choosing one of its predefined 'welcome' messages. From then on, its responses are generated from the user's inputs, by a process that typically involves three steps in the following order:

• Systematic transformation of the input (by means of 'input transformations').
• Identification of 'key words' within the input, and random choice of appropriate messages where this identification succeeds (so-called 'keyword transformations').
• Systematic transformation of the output (by means of 'output transformations' and 'final transformations' – these two types of transformation behave identically unless your script involves Recursion and Text Splitting).

Elizabeth comes with her own resident Script Editor and is backed by comprehensive help pages to help you customize the chatterbot easily.

Elizabeth can be downloaded at http://www.etext.leeds.ac.uk/elizabeth/