Are we allowed to guess the DT algorithm?

OK, probably developers will not speak in this thread but may be we users can do some brain-storming on the main classification and suggestion algorithm. I think we can use the software better if we understand its inner workings (sorry I am a physicist, can’t help myself :wink: )

Bruno Haid here says that “As far as we know DevonThink is a hybrid between Vector Spaces and Neural Networks, and definitely doesn’t scale that well…”

Prentiss Riddle guesses here that DT uses “latent semantic indexing

The latter one makes more sense to me. Are there any more guesses or insights?

Even though there is nothing like DT in the personal computers there are many similar corporate solutions. Like this list. Can anybody compare the effectiveness of these to DT?


PS: Why the links are not underlined in the new forum view?

Thanks, this list is very interesting. I’ll bookmark it :wink:

I’m a fan of this subject :wink:. Even started to look around in my university-librarys computer science department.
Ted Goranson made some guesses in this column:
Read about the last time i started this thread:
Perhaps you are also interested in this: The Remembrance Agent, some sort of command-line DEVONthink :wink: (well not really)
I’m very interested to hear more on this subject.

Louise 8)

Hi Louise
Thanks for the links. Have you ever used “The Remembrance Agent”?

It seems that back in the linked thread you asked a very good question I always wondered and never answered:

While creating notes I obsessively copy-paste the title I gave into the document, doubting whether these important keywords in the title will be used by the AI or not. (Because “words” button does not return the title words) How about aliases? Are they indexed in AI or search? Do they have more weight than a word in the text? I think they should be.

We should know at least these simple things.


I know next to nothing about AI, but
in my field (signal processing) we use lots
of statistics, principal component analysis,
belief propagation, etc., all of which I’m
guessing have something in common with
the DA/DTP databasing techniques.