ECTOR is a multilingual chatterbot that is able to learn from users' entries.
ECTOR is an acronym standing for the French Entité Cybernétique Totalement Obligée de Réfléchir (Cybernetic entity totally compelled to think). It is also a name similar to Hector, the Troyan hero, Achille's enemy. Actually, Achille killed Hector, but I hope here, they will make peace (Achille is a kind of ECTOR's ancestor, see this little story)
I want ECTOR to be an intelligent chatterbot (more intelligent than Achille, but it is likely that it will not be as impressive as Achille --in the beginning--, it has to learn many things before).
ECTOR is based on a artificial intelligence architecture, I called BAsCET (it is shown, in French, in my PhD Thesis, or on Everything2, in English), that is inspired from Copycat, an AI system from Mitchell and Hofstadter.
It is a mix between neural and semantic networks, uses agents and a blackboard. Its Concept Network uses co-occurences to compute influences of one semantic node on another.
What's New in This Release:
· Modified: nodes with a null activation are deleted from the ConceptNetworkState, but only if their age is greater than 50! That keeps the size of the ConceptNetworkState reasonable.
· Modified: generation procedures (forwards and backwards) quicker. One loop erased.
· Added: break recursion after 100 calls in sentence generation (forward and backward).
· Modified: repetitions in the same generated sentence are limited by the length of the token to be repeated.
· Added: URL treatments, one can now place an URL within a sentence, without it being considered as several small sentences.
Product's homepage
Find us on Google+
