I see that you can add ‘keywords’ to a newly scanned file upon import with OCR, but is there anyway to add keywords to a file that already exists in the DT database (or is added ‘comments’ via the Info Panel the only method of applying metadata to an existing document)? Does DT’s AI weight ‘keywords’ and ‘comments’ differently - more heavily one way or another)?
The “keywords” reflect an attribute of a PDF file, they are saved as such. Thus if you want copy the file out of the database or email it to someone, this is preserved. Upon import these are copied to the “Comments” and that is what the AI uses. If you change this it is only saved in the database and not in the original document (whatever the kind is).
If you add keywords or comments to a document’s Comment field, that information is searchable. Tools > Search will let you search only the Comment field, if you wish.
Metadata in the Comment field is not considered by See Also. From my perspective, that’s desirable, as I want the focus to be on the actual usage of text in my document collection of tens of thousands of references.
When I’m scanning documents for OCR I don’t add keywords to the resulting PDFs. I’ve turned off the option to enter Name, Subject and Keywords, so that I can keep feeding copy in without having to pause OCR for each document. After import into the database I’ll often name the scanned documents for my convenience. But if I don’t, that makes no difference to See Also, which looks at the actual documents.
I’ve scanned thousands of pages of paper into my databases. I wouldn’t have gotten as much accomplished if I had taken the time to enter Name, Subject and Keywords for each document.
I’m managing more than 150,000 documents in a number of databases. I use a form of tagging in the sense that I assign material to a topical database, and inside that database to one or more groups. When I’m working on a project I’ll move copies of useful reference materials into that project groups (as duplicates, if I plan to annotate or tag them). I prefer keeping my reference collections untagged, because no keyword scheme could cover all the possible uses of those materials, tagging would take up most of my working hours, and often is more an impediment than a help in looking at material from a fresh perspective.
Having worked with computer information resources for a very long time, my attitude is that tagging is generally a primitive and unsatisfactory approach for most data collections. I restrict textual tagging use to notes about checks in a financial database, or about photos of trips, or to specific collections within a project group, which I will remove from the database after the project has been completed.
The next generation of DEVONthink will make tagging easier. I’ll use continue to restrict my use of tagging. Consider me a curmudgeon; except for limited and simplistic purposes, tagging is too logically limiting to get much respect from me.