Modifiable Semantic Network.

I presume DT has integrated some form of semantic network such as WordNet.  Some questions:

  1. Is it possible for the end-user to modify the semantic network so that categorization becomes ever more ‘intelligent’ for the end-user?

  2. Is it possible to import a semantic network such as UMLS such as done by Polyanalyst (’s website and then importing the derived terminology into DT’s semantic network.

  3. If there is not the above capability, could you consider licensing the dtsearch engine for incorporation of a modifiable WordNet semantic network into DT?  dtsearch ( has a developer’s kit that allows one to use JAVA, C++, etc?

  4. Could you consider integrating Prolog (e.g., Java Prolog or BinProlog from into DT so that a semantic network could be linked both to Agent Technologies (including DA) and to logic programming?  Having a Prolog API in DT would greatly increase its modifiable ‘artificial intelligence’ capabilities.

Just a few suggestions that you can implement within hours, n’est pas / nicht war!!

These suggestions come from a medical educator interested in the application of DT for medical students and doctors taking and applying notes along the continuum of their education and clinical practice.

Charles Beauchamp MD, PhD
Clinical Assistant Professor of Medicine
Duke University School of Medicine ;D ;D

1./2. Actually there’s no semantic network and therefore importing semantic networks or modifying a semantic network is not possible. However, the current classify/see also implementation is really very stupid - this will be enhanced as soon as more CPU power is common (and once there will be some time left to enhance the kernel - we’re almost only application developers at the moment).

  1. Basically we don’t license anything. If more users should request such possibilities, we will add them on our own.

  2. As we don’t want to create bloatware and only add common requests, this is currently very unlikely (I don’t think that there are many users out there doing Prolog)

Thank you for your prompt and frank answers.   Too bad you do not see the inconsistencies in your responses.


Methinks you presume too much. Did you check the prices of the examples you noted, and the level of hardware required to do heavy hitter operations?  The combined price of DEVONthink and DEVONagent is US $60, and they run very quickly on my 3-year-old 500 MHz Mac TiBook. ;D

I’ve been intrigued by neural networks for many years. Having picked up a good many graduate hours in mathematical logic and having served as project director of an information mining and transfer center that used one of the (at the time) fastest mainframes in the US, I’m very optimistic about where computer technology is taking us.

Christian characterized DEVONthink’s classification and “See Also” capabilities as “really very stupid.” I agree, but still find DEVONthink’s contextual recognition capabilities very useful.

In fact, even the best computerized neural network software is “really very stupid” compared to the neural networks in the brain of a human who knows what he’s doing. But computers are useful tools because they can scan enormous amounts of data quickly, and they are interactive with humans – which is to say, if you don’t like the results of a search, modify it and run it again.

DEVONagent has enabled me to run some very sophisticated and useful Web searches. The supplied plugins and crawlers are pretty good, users can devise their own, and the Boolean search query tools are first-rate. Web pages are automatically summarized, further searching and regrouping of ‘hits’ is easy, and any or all hits can be dumped into the DEVONthink database if desired.

Yes, the user does have to be familiar with the technical jargon of a discipline to design searches properly – but if you don’t know the jargon, you can’t understand the literature, anyway.

As for DEVONthink, I continue to be surprised by some of the things it’s “really stupid” capabilities can do for me.

I’ve got thousands of technical references in my DEVONthink database, dealing with topics including environmental engineering, genomics, toxicology and so forth. Individual items range from a paragraph or so in length (such as abstracts) to hundreds of pages.

When I’m doing research on a topic, I can very quickly narrow down the items most likely to be useful. DEVONthink’s Similar Spelling and Similar Context buttons often provide very useful information (even if they’re not very smart!). Likewise, the See Also button often provides insights to relationships that I might not have thought of – as well, granted, as suggestions that I can easily ignore, because they’re dumb.

In sum, if I’m profoundly ignorant about a topic, DEVONthink and DEVONagent won’t enable me to produce solid literature research on the topic. But if I know the topic pretty well, they are really useful tools. I don’t know of any other software, on any platform, at any price, that would give me the document storage and retrieval flexibility of DEVONthink, yet run well on a laptop – except for forthcoming versions of DEVONthink and DEVONagent. (Did I mention the compactness of the database, or the range of file types than can be accomodated?)

Recommendation: try them for a while. I think you’ll be pleasantly surprised.

[Christian, since I’m being so nice, how about adding DEVONagent’s (Near,x) Boolean operator to DEVONthink’s search operators?] ;D

Inconsistencies? I’ve read my posting again and again but can’t find them. Any hints?  ???


thanks for the long feedback - I couldn’t have expressed it better :wink: And ALL operators of DEVONagent will be added to DEVONthink 1.8.x (but probably not the initial 1.8 release which is as usual overdue). But beta 3 includes highlighting of search words in HTML pages and I guess this was another feature you’ve requested ;D

Honey, that’s Official-ese for, Inconsistent with my plans for your life.

Don’t give it another thought. In fact, anybody with a vision ought to stick to it like glue. (Making, of course, just the few small tweaks I need.)


p.s. Anybody old as me might remember David Steinberg on the Smothers Brothers Hour <sigh> authoritatively pointing out the title on his white coat, saying, “See? ‘M.D.’ Me Doctor!”