Algorithm for chatbot
Accept a standard identity matrix expansion of vector representation; the vector of statistical weights of AIML , SRGS, SISR, n-Grams, dictionary, and corpora lookups.
Context should be assigned weights and stored. The statistically most likely candidates are offered for conversation. If a negative response is encountered, such as, "Bad answer, Alice.", the weight of that response should be reduced, and the next most statistically likely answer should be offered.
The stack of recent answers, AIML, should be parsed and assigned weight for current context. Reiteration in the stack should add weight to the topic. The topic should be the most statistically likely semantic interpretation of the symbolic reduction parsed from the stack.
Chatbot should assign weight to identity based on previous contact. She should try to recognise identity eary, and store context and topic weight for the next time the person is encountered. This will help her remember what people like, and make friends.
Context should be assigned weights and stored. The statistically most likely candidates are offered for conversation. If a negative response is encountered, such as, "Bad answer, Alice.", the weight of that response should be reduced, and the next most statistically likely answer should be offered.
The stack of recent answers, AIML
Chatbot should assign weight to identity based on previous contact. She should try to recognise identity eary, and store context and topic weight for the next time the person is encountered. This will help her remember what people like, and make friends.
posted by nobody at 12:12 PM
0 comments
