How is the software project, you ask?

I have more confidence in the machine a I test the various syntax possibilities.  Rules are being generated, not many, three or four.  

I will be am running large blobs of text through the thing in the next few days.  Runs fine using a json attachment along side a plain text attachment.  

Big deal? No, not my puny application, but the concept will spread. The reason is simple, all data used by computers is an N-ary tree, a descend or sets thing. We have a join that only uses the commonality of sets and descents, so any data structure easily adapts.  

Any differences between the two pairs is no problem as long as they agree on what match means.  Matching text and video? Sure, one attachment detects the type of movie action, the other presents a set of movie terms.  Match the terms to the movie, different media but as generic join helps classify them anyway.