|
The talking program presented here is rather different from a typical ELIZA program - it does not mindlessly repeat your input. Different personality modes were include.
Responses will be given depend on the type of personality you select. In addition, a Diagnostic option is also included so that you can see how the computer analyse your sentence. By checking this option you will realise
the computer becomes less human as you notice certain default responses will be given which appear intelligent but in fact the computer does not know what you are saying!
The core section of this applet is the programming of the parser which determines the computer intelligence. The type of parser we have used is a variation of the so-called noise-disposal parser. It only tracks down a
few keywords that match with the vocabulary database. Unrecognized words are treated as noise and discarded. In this applet responses will be given based on the type of keyword present with almost complete disregard to grammer. For example,
‘I am clever’ and ‘Me is clever’ are treated as same sentences. Obviously, in some sentences the order of keywords appear may also be important and this is also partly being taken care of. For example, ‘I love you’ and ‘You love I (me)’ are
two statements with different meaning.
Different forms of verbs have also been taken into account. For example: sings, singing; and loves, loving are treated as ‘sing’ and ‘love’
respectively. However, adverbs (e.g. cleverly, sadly etc.) are not identified, nor the past and past participate tenses (e.g. sang, sung).
The Talk applet can only recognize around 80 words, including words of similar meaning. You may therefore find the computer to be less intelligent then expected. Obviously, one way
to improve the program is to include more vocabulary in the program code. Note also that the applet is coded to recognized English language only. Appologize to all non-English
speaking folks!
|