Allow users to choose which documents contribute to the automatic matching model

[jgillula]

I have a lot of documents dating back over two decades, and the format of the document type has changed a lot in that time. In other words: my Bank of America statement from 20 years ago has very different text in it than my Bank of America statement for the last couple years. However, I want all of those documents to be the same document type.

This presents a problem for the machine learning that does the automatic document-type and tag matching, though, since the 20-year old bank statements and the recent ones have very different text.

To solve this, I propose allowing the user to set a per-document flag that indicates whether or not the specific document should be used in the model (with the flag set to on when a document is first consumed). That way, when the format of a bank statement changes I can bulk unset that flag for all the old statements, and only the new statements will influence the model.

WDYT?