In as much detail as possible, please describe what you would like to see.
Adding to the TODO list feature request suggested in Issue 859, it would be
nice to also suggest improvements to the rules. In particular, if a user has a
state with a NumericInput interaction and has a single answer group with
several rules:
answer equals 4
answer equals 5
answer equals 6
answer equals 7
There can be three logical conclusions made about the numeric domain, based on
the inferred intention of the user:
answer is greater than or equal to 4
answer is less than or equal to 7
answer is between 4 and 7, inclusively
All three of these possible suggestions simplify the above 4 rules down to 1,
which might be what the user originally wants.
Now, this particularly is an issue when training a fuzzy rule. Many examples
may be provided in the process of training the rule. This suggestion system
would make an effort to try and find a single explicit rule to add such that:
1. A maximal amount of training data may be removed from the fuzzy rule for
that group
2. All of the pieces of training data being removed will still be properly
classified with the new explicit rule
This potentially weakens the general classification, but it strengthens these
core cases. This suggestions system may also lead to the eventual manual labor
needed to add answer groups and just infer all of the answers a user wants
through the process of teaching Oppia possible responses.
Original issue reported on code.google.com by bhenn...@google.com on 15 Jul 2015 at 1:19
Original issue reported on code.google.com by
bhenn...@google.comon 15 Jul 2015 at 1:19