I would like the option to specifically exclude certain groups and documents from DEVONthink’s Classification algorithms.
Justification:
[1] There are instances when I wish to segregate material within the database from my general reference collections of documents. For example, when I’m writing a draft about a subject that is covered in the reference collection, I don’t want my draft notes to be considered by the Auto Classify algorithm when adding new material to the database. Likewise, when I’m keeping daily journal or project journal notes within DT, I don’t want Auto Classify to consider the group within which these notes are placed when it looks at groups for placement of new reference materials. (That’s why I often use NoteTaker to contain such notes and drafts external to my DEVONthink database. But I would really like to be able to write this content within DT if I could do so without confusing DT’s classification routines.)
[2] DEVONthink frequently makes ‘mistakes’ when recommending classification of new content. Based on observation, I could drastically improve the utility of auto classification if I could ‘tell’ DT not to consider certain groups or documents for classification purposes. Here’s a good example. My database contains some groups with personal information, including one about the purchase of a Bosch dishwasher. The group about the Bosch dishwasher contains a short note including purchase date and serial number, and a scan of the warranty information. DT has recommended that I place more than a hundred new content items within that group! Why? As nearly as I can figure out, DT does so because the address for Bosch is in the same city as a couple of research institutions that are cited in many articles I’ve downloaded. Almost always, DT’s second choice for placement of the new item is a good one – so we could dramatically improve the auto classification routine if we could exclude a single short document’s content when DT is considering classification of new material!
Another example: many HTML pages downloaded into DT contain my name as a subscriber to a journal site. When DT sees this, it recommends classification into a group that contains a collection of papers I’ve written. While this is somewhat flattering, it’s a bad classification decision. That problem would go away if I could exclude my “papers” group from DT’s classification decision.
Christian, I would think it would be much less of a programming chore to give the user a bit more control of classification in this way, by comparison to a major upgrade of the classification routines. 
However, I would like "See Also" to continue to pick up "excluded" groups and documents.
)