annotation service

[dagendresen]

Does a discussion on an annotation service belong here, in the resolver repository ... or should this be a separate GitHub repository?

The previous "samlingsportal" (museum collection data portal) included a data annotation function based on and reusing code from the transcription portal ("dugnadsportal"). I added some old slides (from October 2016) with some examples and screenshots from this old service. Maybe useful to develop the specifications for a new annotation service?

https://www.slideshare.net/DagEndresen/20161012-musit-gbif-dataset-portals-243209702 https://drive.google.com/file/d/1TnUYQKMzDbDfGV7C9n-1KowoCxqKvouY/view?usp=sharing

[rukayaj]

That's useful, thanks Dag. We can have the annotation service discussion here for the moment.

Maybe you didn't see this yet (I posted in slack), but there's a discussion on the GBIF discourse about annotations https://discourse.gbif.org/t/annotating-specimens-and-other-data/2419/59. I read the thread through, but it is quite long and I felt rather overwhelmed as you will be able to tell by my comment - there are a lot of possible ways for this to work. I find it interesting that many people are focussed on how to get the suggested annotations back in to the collection management systems.

Anyway, reading between the lines it sounds to me like GBIF are going to seek funding to put together a team to build something fairly large and extensive, and thinking about it this morning, I wonder if they're going to perhaps do it in conjunction with the hosted portals.

I really like Donald Hoburn's vision, We can treat annotations less like emailing data publishers and more like Wikipedia talk pages - every entity can have a page that allows free comments to be added, LOD properties to be connected, structured edits to be proposed, reviewed and handled, etc. but I am surprised he doesn't think it's going to require much development time, I thought will require the investment of a lot of man-hours but perhaps I'm wrong.

Anyway, I think we should keep an eye on what GBIF is doing, but I can of course start with our own annotation system. I am working on the functionality to retrieve the data history snapshots, but there is no immediate need for this and I think the annotation system is more urgent. I will have it separate to the resolver.

From the slides you've shared it looks like the main use case is just to have a single user add structured data on a record level, not to do bulk edits. This functionality should not be a problem to build, but I would ideally ike to have it API powered so it might be possible to integrate it into whatever solution GBIF comes up with.

• So step 1 would be to build an API that allows create + view functionality for annotations. Does there need to be such a thing as user accounts? Do we think users should have to log in to make annotations? This seems to be how dugnadsportal worked?
• If there are users, it would be best if they can log in with their ORCID accounts, and maybe Google? And then I guess users should be able to see a list of their own annotations somewhere, but all annotations should be visible to the public.
• If there aren't users, we have to think about spam prevention
• If there is a concept of 'users', do we need to have different levels of users? E.g. some admin type users who can edit or delete the annotations?
• How should machine annotations (e.g. from wikidata) work? Presumably they should be cached. What happens if the current wikidata entry changes for that record, do we display the old annotation and the new one? Do we just show whatever is current on wikidata? We could also just have a link to the wikidata page, and any human/computer could follow the link to get the information.
• Does there need to be an export function for the annotations? I can see why it would make sense to adapt a transcription system for this, it's basically all almost the same functionality... ... So perhaps if Dugnadsportal was successful as an annotation system there is no reason why we can't just build on that? I have been meaning to look at the code there and fix the new one up, it might be more efficient to do that.

[dagendresen]

• The transcription portal had "users", but one could add transcriptions without logging in.
• Even machines could leave "annotations"/"transcriptions" -- the data quality tests were added as "machine annotations" but these were separate from the "human annotations" and the machine "annotations" were written over each data update cycle (each dataset update from MUSIT to GBIF) -- this way Jenny at UiB (Jenny was an enthusiastic user to fix machine annotation issues for the herbarium in Bergen)
• I recall we aimed at users logging in only with open Authorization from Google, ORCID, etc --- and I think that we could even consider limiting support to ORCID (?) -- to signal the high importance of having an ORCID...! (Does Wikidata provide open user authorization we might include as well...? --- but ORCID is "king").
• Non-authorized anonymous annotations could be there as a large pool of statements treated differently just like the machine annotations, and maybe deleted in roll-over cycles if not picked up by logged-in users...???
• Different levels of users would be VERY USEFUL. Preferably by picking up institutional roles from ORCID!! Maybe spiced by roles in Wikidata. If a user has an official role as curator for a collection then considered admin for that collection.
• Admin users should be able to "delete" (hide not like permanently delete) annotations that are "fixed at source".
• Wikidata "annotations" could be cached for the purpose of data processing for data quality validation. Not just ONLY link to it. There must just be a way for users adding annotations in Wikidata to get these new data into the data annotation processing stream.
• Are we able to trust the Wikidata users with an ORCID to simply add their annotations in Wikidata. That the edits such users make in Wikidata are treated as a valid annotation MADE BY THIS USER. (I did not find out how to, if possible, to add the Wikidata user account to my ORCID profile...)
• Agree that mixing in the annotation system with the transcription portal would cover many use cases --- at least a user interface. But we might need to work on a behind the scene API-based system that is sort of separate??

[dagendresen]

Yes, this Alliance consultation on the Digital Specimen (Extended Specimen) envelope was what trigged my attention to our previous old prototype annotation system based on the transcription portal :-)