A more robust multistep approach

[Deleetdk]

A better idea is to use a multi-step approach:

1. Find all matches.
2. Remove duplicates.
3. Sort by length, shortest first.
4. Try each match against DOI lookup, note status.
5. If only one gets OK status, use that.
6. If multiple get OK status, ask user which to use.

This should cover pretty much everything, but could be slow depending on how long it takes to look up DOIs. Unless a page really has multiple valid DOIs (if it has a reference list!), then this should find all the valid ones fairly quickly.

[Deleetdk]

http://link.springer.com/article/10.1385/MO:23:4:443

This currently fails because there are multiple valid DOIs present and the shortest one is not the right one.

Is there some way other aside from shortness one can grade plausibility by?

[onbjerg]

@Deleetdk Maybe we should check if the DOI is visible. The other DOIs are in the URLs of the cited papers.

[Deleetdk]

Can check if it's in the visible part of the code. Like not in href or

.

On Nov 3, 2016 19:49, "Oliver" [email protected] wrote:

@Deleetdk https://github.com/Deleetdk Maybe we should check if the DOI
is visible. The other DOIs are in the URLs of the cited papers.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#4 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AGcl6fr8F6BHKD5dom223Mk0-P4nLHYNks5q6iykgaJpZM4KoAj-
.

[onbjerg]

Actually, maybe we should check the meta tags as well. The site you linked to has a meta tag for the DOI.

[onbjerg]

Also, the DOI handbook states that one should strive to display it as doi:<doi> and another major registration agency recommends using https://doi.org/<doi> (ref).

As such, we should probably rank them in order of "correct" format:

1. doi:<doi>
2. https://[www.]doi.org/<doi>
3. <doi>