I was flipping thru a book the other day. An OReilly book
about "Hacking Google". It had direct interface code
examples.
I remembered that one of the problems with old Eliza
code was the huge database required in order to make
jumps in logic (mentioning sports should prompt for
baseball? footbal? cricket?).
I realized that Google is a huge database which can
provide excellent logic jumps. Eliza was able to
recognize key-words (anything not in the normal syntax)
and if it could probe google with those words then the
top-10 responses could be mined for new keywords to
use in a response.
A: "Im trying to find info on Albert Einstein"
E: "Ahhh the theory of relativty, E=mc2"
A: "Well yes but I knew that. Im trying to find out when
he was born."
E: "Albert Einstein. Born: 14 March 1879 in Ulm,
Württemberg, Germany Died: 18 April 1955 in Princeton,
New Jersey, USA."
That wast response was #1 hit on keywords of "albert
einstein" and "born" fed into google.
Anyway, it was just a thought. You seem to have Nicole
parsing human conversation. If she could parse google
responses in the background it might give her a "brain"