I imaging the following logical components of the application:

• Retrieving the Sefaria text and splitting it into lines and (where applicable) sentences
• Setting an "anchor" that highlights prosepective matches (e.g. a gematria number that may be a Seif or Siman)
• The ability to select and tokenize text that contains either:
  • The book name
  • A specific level of the hierarchy
  • An exclusion (e.g. the string preceding the anchor shows that the word "siman" is not a book in this context
• Running all past patterns on each of the following string with an anchor in order to automatically classify them (compile tries with xregexp ?)
• Submitting batches of tokenized results to Sefaria (user would have to supply own API key).

I have not given much though tot the UI yet; frankly, I wouldn't mind if the first pass was a console application (though the tokenizing would be hard to implement in that version.)
Thoughts?