|
|
|
|
|
|
by vunderba
268 days ago
|
|
|
I remember reading the source of the original MegaHAL program when I was younger - one of the tricks that made it stand out (particularly in the Loebner competitions [1]) was that it used both a backwards and forwards Markov chain to generate responses. [1] https://en.wikipedia.org/wiki/Loebner_Prize |
|
|