Feature request: Better knowledge discovery by finding online papers with higher relevance

[raindropsfromsky]

Qiqqa starts its value-addition after the user has searched the internet and collected the documents in his library.

In comparison, some reference managers go an extra mile and help the user identify good papers by searching the online catalogs for the desired key phrases.

Mendeley Suggest is a good example. But its document list is curated manually.

Docear had a recommender system (the product is discontinued).

But Qiqqa does not need such manual intervention (or even collaborative effort), because it has three built-in features:

1. Its Expedition feature finds common themes automatically.
2. The Autotag cab be used as search words. In addition, tags that are added manually also serve as searchwords. .
3. Qiqqa can use the nodes of the Brainstorm diagram to search for relevant papers.

At present, Qiqqa does not have a facility to log in multiple online catalogs and search them for suitable papers.

But with the features outlined above, Qiqqa can automatically search all catalogs and then rank the found documents based on their scores, just like how it scores the documents in the user's library. This would save a lot of time for the user.

Also consider that this search is actually a "think ahead" feature: While the user is busy analyzing the document that he has got in his library, Qiqqa already looks ahead and finds the matching documents from the internet and keeps it ready for his perusal. This engine can be set to work only when the PC is idle, so that only idle resources are used to the maximum, and the PC does not slow down because of the extra load.

[GerHobbelt]

Brain is shutting down tonight, sorry. Got to reconsider this later.

Things that pop up immediately though (brain-💩)

• Qiqqa has "something" (requires user action as in click) in the sidebar of the PDF View pane: this search-in-advance was done via Google Scholar and that got kicked out in latest v82 versions (issue #188) as that approach was increasing the risk of the Sniffer / RECAPTCHA / lockout situation, which is a grave showstopper. (#241 + #155 + #113 + ... -- 😨 not a good sign sign that I now know these issue numbers off the top of my head 😓 )

• search ahead (or shall we call it 'prefetch' or 'presearch' perhaps) is a great idea, but I'm not riding on top of paid search services, so that makes this a dangerous proposal when you're not funded and thus able to buy into the APIs/whatnot of the search engines you're going to look through: while prefetch enhances the search, it also implies more search queries and thus "hammering". See #241 for a nice wake-up call about that one: google isn't the only pal with that attitude out there!

So I wonder how this might be done without getting kicked in thy bottom for bad boy behaviour 😉 when you dial this feature on and used it for a while.

[raindropsfromsky]

Ok if the "automatic" and "search ahead" are both objectionable, then let us skirt both behaviors.

Let Qiqqa create a list of related topics (ontology) of the subject, using the auto-tag and theme-detection features. Allow the user to select any of those terms (or even manually add new terms) and then launch a search in the online catalogs.

This is not much different than a completely manual search in Google Scholar, except that Qiqqa will provide a ready list of search terms/phrases.

BTW both auto-tag and theme-detection features are broken at the moment, and you mentioned six months ago that it was due to some debugging effort that had gone wrong. So it should be easy to reverse those experimental changes and make them work again...