search

barbalet / apesdk

The aim of this artificial life simulation is to create a detailed biological environment and a cognitive simulation.
https://www.apesdk.com/
Other
55 stars 1 forks source link

Create standalone language engine #54

Open barbalet opened 8 years ago

barbalet commented 8 years ago

Feature ape names, land names, time of day, weather, verbs and objects.

ghost commented 8 years ago

An idea here is that each individual learns the grammar of its culture with a series of rewrite rules for phrase construction. There are existing C examples of implementing grammars for things like Lisp, but the interesting part would be to add learning and have it work in a similar way to known language learning stages.

There are some questions such as whether the lowest level should be phonemes or whether there should be a pre-defined vocabulary. Probably the phoneme approach would be better because it's more amenable to audio output and also evolution of the language.

aaannndddyyy commented 7 years ago

That would be very interesting if it were not hard-coded but could evolve, as @bashrc suggested. Symbolic, or even better maybe coupled with the braincode somehow. A combination of hebbian learning and reinforcement learning could help them learn languages also from other groups. Or some Bayesian magic that makes them guess form circumstance and importance ( c.f. Grice' implicatures) the meaning of words other utter. This way even random initially meaningless utterances can be adscribed meaning by the listeners, which then coagulates statistically.

ps: From the above, it's pretty obvious that I'd favor phoneme-based. You can have their biology determine the set phonemes they are able to utter. It would be so much more THEIR language if you do not prescribe them the words (form), concepts (ideas) and grammar (structures) they use.

barbalet commented 4 years ago

Part of this is in https://gitlab.com/barbalet/hundred