George Bush Conversation Simulator essay
This paper documents the development of a stimulusresponse chat bot, George. It presents information about existing chat bots and natural language generating systems. It goes on to discuss the design and implementation of George and the various methods used to ensure George is capable of performing its function and achieving its goals. The latter parts of this report put forward a procedure for testing and evaluating whether the program has obtained its goal and also documents the results of these tests. The paper ends with a final conclusion on the degree of success the project has, with respect to its aims and purpose of use, achieved.
Natural Language Generation (NLG) and Chatterbots
Natural language generation is described as computer compiled utterances that are created via some manipulation or understanding of an input. A chatterbot is a computer program designed to give the appearance of having an intelligent conversation.
NLG systems attempt to solve the problem of producing discourse computationally. While many chatterbots attempt to pick out keywords in a given input and match them to a database of "knowledge" to compute the most relevant response, there are chatterbots that use natural language processing methods to generate a response.
Existing chatterbots include; ALICE a general chatterbot, whose development began in 1995 and uses a heuristic approach to match patterns in the users input. SmarterChild an IM based chatterbot, launched in 2001 uses a mix of pattern recognition and natural language recognition to take user input and produce database queries. SmarterChild takes advantage of the practical use of it's natural language processing ability and doubles up as an information retrieval system allowing the user to access information about the weather and cinema times etc. Jabberwacky an interesting chatterbot that attempts to learn natural language through interaction with the user. Jabberwacky was launched on the internet in 1997.
The 43rd and current president of the U.S, he was elected as president in 2000. A central issue in his presidential campaign is his "war on terror" and involvement in ordering military action in the east. On a lighter note however, he has become synonymous with making embarrassing and often hilarious grammatical errors when speaking during public appearances. This part of George Bush, the president, is to be "captured" in George the program for the purpose of humor during conversations with the user
Conversation is the verbalisation (or in George's case textualisation) of the ideas and concepts we perceive in our minds with the intention of communicating them to one or more people. Computers do not have minds. They do not have ideas nor do they conceive. However they do have an accurate memory in which information about objects and systems and ideas and concepts can be placed and rules of methods to extract this information can be inserted.
Natural language generating machines, arguably, have two main problems to deal with, deciding what to say, and choosing what words to use. Both of the preceding problems are solvable using grammatical methods. Due to the time constraint on project George an ad hoc method of dealing with "understanding" the users inputted natural language, and choosing the utterances that George will be responding with will be implemented.