How to use the AI features to file documents into databases

I file my documents per hand into my databases, but I was thinking maybe your AI can help me. :slight_smile:

Here my workflow:

  • All my DT documents arrive in the Global Inbox. Then I file them in my databases.

  • I have five or six databases, but here I talk might just talk about two: a database Computer and a database General (both have already many years of documents).

  • I file the documents into one of my databases and then I add additional tags, for example the tag game for a game info.

So let’s assume I start a filing session. I see in my Global Inbox the webarchive of Truberbrooks. I know I give it the game tag, but I don’t remember in which database do i file games? In General or in Computer?

Not a problem to find out with a search in a 2nd window for the game tag. I see straight away: “Aha, I always file games in General.” - No problem, but a little bit tedious when a couple of weeks later I file another game and I have to do the same procedure. (Because by then I have forgotten where I file games…)

Of course I can put everything in one database, but let’s assume I would like some separation of different stuff.

Is there any way that your AI could help me? Can the AI, when I tag something with game, automatically suggest to move it in the General database, because all other documents with the game tag are in that database?

Just a thought…

Take a look at the DEVONthink manual about “Auto Classify” and then after that look in the manual about “Rules”. Also get and read the free copy of “Take Control of DEVONthink Pro”.

PDF’s above at https://www.devontechnologies.com/support/download/extras

1 Like

DEVONthink has a powerful See Also & Classify feature and you can open it in the inspection pane by pressing Ctrl S, I use this feature extensively when filing documents from the global inbox into specific databases and groups and the results are stunningly accurate.

I’m not sure what affects the accuracy of this feature though. Obviously the document content is taken into consideration, but I would like to know if other factors (tags etc.) play a part too.

2 Likes

Tags are not considered for a classification (or See Also). Content and location are used for classification.

2 Likes

See Also & Classify is a GREAT tool! When you’ve selected a document in the Inbox, the SE&C pane lists some suggested folders in the top half of the pane and similar documents in the lower part. The correct folder is usually the top one listed, but not always. Just click on the correct folder, and then click on the “Move To” button at the top, or just type Ctrl-C (^C). If the proper folder is the top one, you don’t have to click it first, just type Ctrl-C to move your document into it.

3 Likes

You can also double-click a result to move the file or even drag and drop to a suggested location.

And if the suggested group is in the same database, hold the Option key to replicate the file to the suggested group.
:slight_smile:

and

and more precise

:slight_smile:

2 Likes

Talking about replication, I should have mentioned that you can highlight multiple groups in the same database and then click Move To or type Ctrl-C, and the document will be replicated in the multiple groups. I use that feature as well.

Indeed, I was a bit loose in my response :slight_smile:
Thanks for the assist.

1 Like

Thanks heaps, @xurc! This kind of tool is what I was after. :slight_smile:

It might not be perfect, but so far I have tried it it works pretty well. Thanks again for your clear suggestion!

Great info. Thanks!

I use classify all the time. What I find is key to auto classify is whether there is a discernible pattern pattern. I am an attorney and the vast majority of my documents come to me consistently have the same information listed on the first page (eg, case number, party names, etc). This makes devonthink’s AI work extremely well. But sometimes, I am pleasantly surprised to find that the AI has zeroed in on something I didn’t notice even where the normal identifier information is not in the document. I find the AI works best also if you are consistent in which group you file those documents. So ultimately where you file things isn’t so important. The main thing is that wherever you file it, you know you that location is a place where you would file documents similar to that document. After you file 3-4 documents according to the pattern you have decided on the AI will “pick up the ball” from there. I don’t use auto-classify because I still find that the AI picks the wrong group, probably 10-15% of the time. But most of the time the first group the AI selected is the right one, so all I have to do is hit the keyboard shortcut “ctrl-c” in rapid fire fashion. When the AI is wrong, usually the second option the AI suggests is the right one. I can process hundreds of documents in a couple hours using this approach.

2 Likes

Do you mean to select multiple destinations or multiple files to move?

The “See Also and Classify” tab can with one doc selected go to multiple destinations (but only in same database), but will not move multiple-doc’s selected into any destinations. For the latter, use the “Move To” command (one of the actions under the “Actions” button) to move the files (plural or singular) to one destination. At least this how I use/see it. Try for yourself and see what happens.

As rmschne said, moving a document into multiple folders, but since you’re using replication, there’s still only one document, it just shows up in multiple folders. Also as rmschne said, this only works for multiple folders in the same database.

Yes, when there are multiple destinations involved, one is the original document and the rest are replicants. I forgot to say that.

I just started using the “See Also & Classify” inspector and I agree it works rather well for locations/groups.

As mentioend by the DEVONthink team it does not for tags. I have sorted my documents by tags more than groups, so for me it is overall not that much of a help…

However it does sort out the problem I began this forum thread with, so I’m “happy enough”. :wink: