Interpreting the digest view


Could someone offer me a somewhat technical (or at least “specific”) explanation of what the grey lines represent in the keyword diagram of the digest view? I understand that they represent a relatively “close relation” between connected keywords, but what does that mean?

What does it mean, say, in such queries as:

  1. keyword1 AND keyword2
  2. keywords1 OR keyword2
  3. keyword1
  4. keyword1 NEAR keyword2

Also: What is the difference in connection strength between two keywords that are indirectly connected and three nodes apart, for example, and two nodes also indirectly connected yet four nodes apart? (Beyond the tautological answer that in the former case the connection is ‘stronger’ or ‘closer’.)

And what if, to give an extreme example, the two top-ranking keywords in terms of occurrence frequency are not directly connected in the graph? Or not connected at all??

I hope this is not a proprietary knowledge (trade secret) because DA is software for knowledge workers, therefore people who need to know what they are talking about (and what’s more, people who cannot suppress that urge to know).

The manual is unfortunately not clear on this matter, and therefore, much as the diagrams look compelling, they do not mean much (at least not to me).

Thank you very much!

UPDATE: I discovered this thread in the forum but Bill DeVille’s answer is not quite specific: [url]Graphical Window Digest]

Basically just that related words are used in similar contexts/documents.

The digest is actually independent from the query and only based on the accepted results.

Thanks for this (perhaps a little too elliptical) answer! :slight_smile:

So I understand that for every pair of nodes there is a certain “threshold” below which they are not directly connected?

That’s right. We could even add a preference/slider to adjust this but I’m not sure if there’s demand for this.

A slider to adjust the distance between thresholds (therefore the sensitivity of the graph) sounds like a wonderful idea! Sometimes the system produces graphs that seem to me too “flat” (too many equal connections), so I guess that making the thresholds scale more fine-grained would provide more information.