Base a search on known links?


I’m a bit new to DevonAgent but like the application.

Is there a option to base a search not on terms manually inputted into the fast search field, but give the application a list of known good results.
I would imagine the application could identify the common factors and do an auxiliary search based on these terms.



That’s an interesting idea and similar approaches have been discussed several times in the forum.

Some software attempts to look at the text content of a displayed page and find possibly related items on the Web. The concept is rather similar to the See Also feature in DEVONthink, but so far has been done in a much simpler way. For example, the software might attempt to identify several ‘most important’ terms and base a Web search on those terms. Christian has experimented with several approaches like that and perhaps we’ll see something like this in the DA browser in the future.

To look at the entire set of results of a DA search and ‘distill’ the key contextual items that might support a search for related items on the Web, as you suggested, is a more interesting and more complex idea. In effect, DA would examine the set of results and devise a new search query to expand or limit the search. If the user can ‘tell’ DA which page(s) among the set of search results are the most interesting, such an approach might result in a new search query that would produce richer results for the user, and/or a process similar to the existing “See Also” in DEVONthink.

While DEVONagent searches are supported by a set of very powerful query operators and filters that allow one to precisely ‘define’ the desired pages, DEVONagent already includes artificial intelligence features that look at the text content of the entire set of results and suggest some of the more important ‘topic’ terms. In Digest view of the results, one can click on a suggested term and see a list of summaries of the pages relevant to that ‘topic’ as well as a graphical display of how that term ‘connects’ to other terms in the search results. One can also enter an unlisted term to see if it’s relevant to the results. So far, ‘topic’ terms are single terms (logical analysis and processing/memory demands increase very greatly for multi-term strings).

When i look back ten years at the level of AI assistance and user interactivity available to us back then, it’s clear that we’ve made a lot of progress in the intervening ten years. I would expect even more progress in the next ten years, but perhaps the hardware requirements for memory and processor resources will have to expand well beyond what we can afford at the consumer level today. There’s no reason to think that won’t happen. :slight_smile:

So the short answer is no currently not possible :frowning:
But can you tell me how to add a unlisted term to the results?
This sounds like an interesting start.

If you mean the list of topics in the left column of the Digest view, there’s a field into which you can enter a term, at the bottom.