I am often saving web pages to DevonThink as PDF files.
It seams that for web pages, the DevonThink hallmark Move To/See Also is not working as well as for ordinary documents. The suggestions are seldom useful. I assume a large part of the reason for this, is that a web page contains so much more information and navigation than just the main text - and a lot of the extra info is not closely related to the main text.
I think the result would be better if a algorithm like the one used by Readability where applied, to find the main text of the page, and index just that. This would give a text similar to the one presented by Safaris Reader function.
Thanks for the suggestion! The only workaround right now is to take a plain or rich note of the main text using services. This reduces also the size of the database and improves the concordance.
I did a quick and dirty script that translated my pdf:s to rtf-documents, using the algorithm mentioned above. The accuracy of the suggestions is vastly improved. Not to the level that I would trust Auto Classify (at least not for my database), but now I get several useful suggestions most of the time.
Extracting the main content of the page, and presenting it in a size and style suitable for reading, also is an improvement.
The major drawback is that the algorithm is not 100% accurate. Some pages I have to convert back to pdf again, to see the relevant content.