search

mars0i / popco2

Cultural transmission with analogy biases.
GNU General Public License v3.0
3 stars 0 forks source link

Finish model-based bias implementation. #40

Open mars0i opened 9 years ago

mars0i commented 9 years ago

Currently have a new field model-quality in Utterances, but it's not being used in communic/speak or communic/listen yet. Maybe add a field to Person to memoize these. Maybe allow it to be calculated with the new two-argument version of many-times? Or should I get rid of the two-arg version and do something else for model biases?

mars0i commented 9 years ago

Now Person has three new fields in total: quality, quality-fn (which will fill in the value of quality, if the function exists), and bias-fn, which will make use of quality if both exist. Currently this is all implemented, with qualitys being transmitted in utterances, except that bias-fn is not yet used in listen.clj to filter on quality.

mars0i commented 9 years ago

TODO: After getting utterances, sort into collections by speaker-id maybe in the bias-fn (or here?) NOTE: need new example model for this, since I need multiple utterances per speaker, and also should test new utterances, which models so far mostly don't use.