This is the first version which found a difference between interleaved and blocked learning. In the current implementation curriculum is controlled by the number of stories within a block. In the next implementation, curriculum will be controlled by the probability of staying versus shifting.

cswEngine

CURRENT IMPLEMENTATION

Objects: Exp, Node, Questions, RFC
RFC is a dict, RFs are dicts
Exp has .get_path() which returns a sequence of nodes and an RFC

NEXT STEP

NOTES

current RFC mechanism

  schema_role_fillers_D, {roles:[fillers] } and 
  filler_properties_D, {filler:{property:value}} 
the get_filler_properties method takes an RFC_str = {"role":"filler"}
  and returns and RFC, which is {"role":filler} = {"role":{"property":"value"}}

make_RFC_bag_full does most of the heavy lifting
it first makes a list of tuples of "role-filler" 
then uses .split('-') method to make a list of RFC_str
then pass these RFC_str to get_filler_properties to make RFC

richly filled

more roles and fillers roles that alter transition probabilities are consequential, roles that don't are innocuous.

RFC's will continue to only have two aggressors and one victim
now also includes: setting.name, victim.emotion, victim.yell, victim.type, drink1.name, drink2.name, drink3.name, dessert.name

further ideas

want .get_info() returns str with info about the entities in RFC .query(condition) returns true or false. currently encoding as a string with the info of role.property.bool

node.pr is {subj.viol.true:{nodename:pr,}, subj.viol.false:{nodename:pr,}} should pr and RFC interact better for querying?