Damien Broderick wrote:
> At 03:24 PM 8/04/99 -0400, Mike wrote:
> >> * Alice Nexis (http://birch.eecs.lehigh.edu/alice): Richard
> >> Wallace's chatterbot has the evasive style of a politician.
> >I just had fun with Alice today. She's a pretty good conversationalis
> I just spoke to Alice. I was disappointed:
> > Does The Matrix kick butt, or what?
> I think it does the matrix kick butt or what
She does have a weakness in that she does not do much or any semantic analysis, and a lot of her responses are triggered by very simple patterns, as follows:
Challenge: Hi, my name is Mike
Alice scans this as fitting in the pattern: Hi, my name is * Alice's database tells her to respond to this pattern with: Hello, * , nice to meet you.
In your example, she saw the pattern: Does * to which her database told her to respond: I think it does * which is obviously an error.
She needs a grammatical lookup table so that the owner can program her to recognize more complex patters of nouns, verbs, adverbs, and adjectives, etc, and to use those words in rearranged ways in her responses.
She does have learning capability, for example, when you say something 'is' somthing else, she will remember that. She knows the sky is blue and my name is Mike (she can even remember WHICH Mike I am, based on the IP address I browsed into her page from). I am dialing up through sover.net, so before I told her my name, she called me Sover-person.