Description of How AI Works in DEVONthink Pro 2

I have used the earlier version of DEVONthink Pro but probably didn’t take full advantage of it. I just upgraded to DEVONthink Pro 2 (not the office version), and I would love some examples about how the AI features work in real-world situations.

Also if you import PDFs into DEVONthink Pro 2, can the application use the AI function if the OCR functionality isn’t there? I am kind of confused by all this.

Thanks in advance!

If you can select or search text in a PDF, it does’t need OCR. But if the PDF is merely a picture of text, OCR is required to make the PDF searchable.

The AI features such as Classify and See Also analyze the contextual relationships of the words in a document and compare them to the contextual relationships of the other documents in a database.

These features become more and more useful as you populate a database.

Let’s talk about ‘Classify’ first. You have to initially create groups that contain related documents. As you create and populate groups, the database analyzes these relationships of documents within groups, in terms of their text content.

Now, when you add a new document, you can click on the ‘magic hat’ symbol in the navigation bar above the document pane and a sidebar will display a list of suggested groups that may be appropriate for that document. (Classify) If you agree with one of the selections, click on it to move the document into that group. As your database grows, and if you are reasonably consistent in the way you organize material, you will find that Classify gets better over time. It’s looking at patterns of word usage in that new document and comparing it to patterns of word usage among documents in the various groups in your database.

Also in the sidebar you will see a list of documents that are possibly similar in contextual patterns of their words. (See Also) Whether you are a scientist or a historian, remember that DEVONthink doesn’t hold a degree in your specialty and, in fact, knows nothing about science or history or any other discipline. But DEVONthink is very good at looking at frequencies and patterns of word usage, and if it detects similarities of these kinds among documents it will suggest them.

It’s up to the human user to decide if the suggestions made by See Also are useful to you. Some of them may seem useless. But I often find new relationships between ideas that I hadn’t thought of.

An example that I talk about is a suggestion made by See Also when I was reading a paper about ecological impacts of invasive species on native populations. One of the suggestions was another paper in my database about factors that affect the equilibria of chemical reactions. That was a very fruitful suggestion, as similar quantitative principles are in play and it got me thinking about how populations of organisms are dynamically related to factors such as food and habitat resources, and likewise to competition with other organisms for those resources. .

Bill’s note is so good I clipped it into my DTPO daybook database. :smiley:

Thanks, Bill. Your response is exactly what I was hoping for! :smiley:

I just upgraded to DTPO, and is there any way to get its OCR capabilities to decipher PDFs that were inputted beforehand, when I just had DESKTOP?

I have several typed letters of my grandfather’s that I would like to make searchable, but I don’t see where I can access OCR capabilities unless I’m scanning in a new doc.

Thanks in advance!

Go to Data/Convert/To Searchable PDF. That will convert your pdfs to pdf and text.


This last response reminds me that Bill de Ville has been intermittently driving me wild with fury since, oh, 2003. He regularly pulls stunts; as if you rang Information and said “Can you give me the number of a dry-cleaner near my home?” and they said “Well, no, not yet, though it’s in the pipeline. Meanwhile, what I do is fly to Nice where you can get the best bouillabaisse in France at a place called Tetou. Or failing that, another workaround is to replace the RSJ in your living-room extension with two pairs of Oliver Peoples sunglasses and a female iguana. It works for me.”

On the other hand sometimes, as here, people occasionally raise issues of terrible complexity at which point Bill dashes off a carefully-worded reply of such limpid lucidity that you want to make it compulsory reading for everyone whether or not they use DT.

I think the truth is Bill is an old-style Greek god and probably lives on Olympia. Exactly that combination of capriciousness, omniscience, self-certainty, the desire to help, the opposing desire to impose his will on people who have done him NO HARM AT ALL, deliberate obscurantism and remarkably incisive clarity of expression. What I want – what we all want – is, extended from just DT/DA to cover every aspect of our lives. This cannot be beyond the capabilites of the developers, surely?

It would be d@mn useful. You’d start to welcome intransigent problems in life simply to see what the hell actually said about them. The answer would be somewhere along a continuum from barking to revelatory, but you wouldn’t know where until you asked the question.

Bill: long may you reign!

“mbywater”: you will be hearing from my attorney quite soon to extract a reasonable sum to pay for the repairs to my 23" Cinema display due to my having splattered a mouthful of espresso on it at the end of your first paragraph. Oh, and the cost of that fine cup of espresso as well. :smiling_imp:

I can only assume that “RSJ” refers to the recto-sigmoid junction in one’s lower gastro-intestinal tract (do your own lookup - there are some very distoibing images on pages I wanted to embed here.).

Bill definitely does not live in Olympia, WA, or even Mt. Olympus in Greece. He has a very nice palace in Brown County, IN, with the requisite seraglio suited to his qualities, which you so accurately enumerated. :smiling_imp:

@twicks: RSJ = “rolled steel joist” in this case. As for the other RSJ, I gave up being a medic so I didn’t have to look at or hear about that sort of TOILET TALK any more. :smiley:


Damn! You blew my cover.

As you were entering that post I was sitting in a flying cattle car, returning from a DEVONtechnologies meeting in Bavaria. I was ruminating (pun intended) on the life, death and influence of Mad King Ludwig.

We visited Ludwig’s Neuschwanstein Castle. He had been profligate in spending in support of the arts and in building his castles. This engendered a growing amount of resentment among his ministers, who arranged to have him declared insane. There’s a lingering suspicion that his death (supposedly by drowning) may not have been accidental.

At any rate, the enterprising Bavarians later turned Ludwig’s castles into a major tourism asset. The wealth he had poured into them was repaid many times over by tourist dollars pouring into the region. I doubt, however, that was the result contemplated by Ludwig himself. In software jargon, that would be an undocumented feature. :slight_smile:

Christian, Annard and Eric have been busy working in the code mines; more neat features coming along.

Just the sort of term that springs to mind and is part of everyday coffee shop talk! :laughing:

Actually the first thing that came to my mind, as an old telephone man, was RJ-45 but that clearly wasn’t the case. The other term was the top definition in the first 10 google hits I looked at so forgive my venturing a guess. :smiley: