How to implement the soft classifier concept? There are two data structures to edit: the field label and the field values for all the verbs. The field label is easily edited with a dialog box. The problem that bothers me is the breakdown of the labor. First the storybuilder changes the field label, then the storybuilder must set the individual values for all the verbs. What a job!
Even then, there's more work to be done. The Erasmotron should evaluate each field for its information content relative to the other fields. It could then issue an overall assessment of the value of each field. The problem is, once the storybuilder has gone to all the trouble of specifying a field, she doesn't want to throw it away on the chance that she can come up with a better field. Yet, we also don't want to burden the system with a zillion fields. 32 fields is plenty.
I don't know what to do about this.