ECTOR

There is a C version in development.

These two versions are under GNU General Public License

ECTOR uses occurrences and sequences of terms to modelize the language used by the utterer.

Terms just used by the utterer are "activated", and this activation is propagated through the model, to activate the nearest terms, or concepts". A sentence is built from the more activated terms, taking into account the most likely sequences.

Expressions are detected, and built, making the built sentences more realistic.

However, the sentences resulting are not that nice. Sending back already read sentences (as Jabberwacky does) should be more human-like, but less creative.

Also, the activities in the model stay a while, along some sentences, constituting a kind of mean-term memory.

External links