General Overview of Program Functions

The program �Arthur� that we have created in C++ is a trial program for a fully functional �Chatterbot�. The core of the program has been successfully modelled and is fully extensible.

The program has four distinct components: - The Header file, the Database, the Template file and the program itself. The program is also divided into six functions: - main(), parse(), response(), search_database(), display() and templ_database(). An overview of each of these functions and components will be given with individually.

The input from the keyboard is read into an array using the empty spaces between the strings as separators. The sentence is assembled from this array of strings. At this point, the sentence is sent to a function in the header file (See header file definition) where it is cleaned of punctuation and converted into lower case. Each string is now a word.

Before any parsing is done to the sentence it is now possible to look through the database to see if there is a match between any topics, in the form of key words or phrases and the input sentence (See Database definition). If a match is found, one of a number of possible replies can be generated. If no match is found in the database, the sentence is passed to the parse function where it is analysed in order to create a response.

In the Parse function (See Parse function definition) the input sentence can now be read as a bit stream pattern. The header file contains a lexicon of known words including, verbs, nouns, adjectives and interrogatives etc. The pattern uses integer values to identify the input words as belonging to one of the different word groups. It does this by comparing each input word with the lexicon in the header file.

The patterns are then used to see if a sentence is acceptable to the rules of the program. In our program, we have elected to define two patterns as being acceptable in order to demonstrate the �Chatterbot�. In future versions, any number of bit stream patterns could be used. Each pattern, defined as being grammatically correct, by the program, requires additional code to deal with the structure of the response. The response is generated in the Response function (See Response Function Definition).

The response to the input pattern is generated by the rewrite rules within the program. The rules attempt to deal with all of the verb tenses and the grammar needed to respond to the input sentence. The response generated is then passed to the template file. The template file consists of a number of possible phrases within which to place the response (See Template definition). A template is selected at random to house the response. There is no reason to limit the number of choices possible for a match to a particular response type.