Allow moderators or administrators to curate tags

[conferences-gesticulees]

• PeerTube version or commit: 1.0.0-alpha.8

• What is the expected behaviour?

Tags will only be useful if videos about the same topic all use the same tag. It would be good to be able to list all tags and to be able to delete or “rename” a tag.

• What do you see instead?

There is currently no way to browse tags. Changing a tag in multiple videos requires to visit the metadata page for each video, which can be pretty tedious.

[maxlath]

Following up on the conversation with @Chocobozzz at the JDLL and on mastodon, here is how I think the task of deduplicating tags could be eased by the use of Wikidata items:

Each Wikidata item represents a concept: as a first approximation, 1 Wikipedia article => 1 Wikidata item.
Example:

• the city of London, a.k.a. Londres in French or Londra in Italian, is uniquely identified in Wikidata as Q84
• exoplanets, a.k.a. extrasolar planets, Экзапланета in Belarusian, or كوكب خارج المجموعة الشمسية in Arab have the unique id Q44559

This means that you can search for any of those terms in any of those languages and be gently redirected to the uniquely identified concept:

• search London in English
• search Londres in French
• search Londra in Italian
• search exoplanet in English
• search extrasolar planet in English
• search Экзапланета in Belarusian
• search كوكب خارج المجموعة الشمسية in Arab

This is thank to the fact that each of those items can have one main label and a series of aliases in each languages.

This disambiguation makes it a great base for a tagging system, the hard work of deduplicating concepts being done by the Wikidata community.

It also gives structured relationships between topics: see how Quora uses it

But building PeerTube tagging system on top of Wikidata would have another major virtue: it would link the tags to the Wikidata-aware web of data! We could for instance easily link to each others between PeerTube and Inventaire:

• books about London: https://inventaire.io/entity/wd:Q84 (API)
• books about Exoplanets: https://inventaire.io/entity/wd:Q44559 (API)
• page for the book "La Horde du Contrevent" and the people sharing a copy of it: https://inventaire.io/entity/wd:Q3209796

I would love to be able to suggest a list of videos on a given topic on those different Inventaire pages: if there was a way to query PeerTube instances APIs to aggregate such a list using Wikidata items ids, that would be crazily awesome :D

As a bonus (I don't see an application in PeerTube yet, but let's see), out of this possibility to link between systems using the Wikidata ids, Wikidata also provides you with identifiers in many databases: it can tell you that Q84 equivalent in OpenStreetMap is 65606 or 119861367 at BNF, etc. which allows to crosswalk from one identifier system to another:

• how to go from the BNF id to the Inventaire page: https://tools.wmflabs.org/hub/bnf:119861367?site=inventaire

As you can see, I'm quite enthusiastic about it, so feel free to ask for any further explanation ^^

Limits

The limit of this idea is that you depend on what is in Wikidata: for instance, you couldn't tag a video with a JDLL 2018 tag (there is JDLL (Q3187082) but not JDLL 2018) or a [insert your unknown association/school/new concept] tag (see Wikidata notability criteria). A suggestion to work around it: make the tag input autosuggest Wikidata items when typing, but if none is selected, still allow to create the tag: it will just be a plain string, not linked to anything but the other plain string tags.

Further readings

• Wikidata: The New Rosetta Stone

[maxlath]

Following up on the conversation with @Chocobozzz at the JDLL and on mastodon, here is how I think the task of deduplicating tags could be eased by the use of Wikidata items:

Each Wikidata item represents a concept: as a first approximation, 1 Wikipedia article => 1 Wikidata item.
Example:

• the city of London, a.k.a. Londres in French or Londra in Italian, is uniquely identified in Wikidata as Q84
• exoplanets, a.k.a. extrasolar planets, Экзапланета in Belarusian, or كوكب خارج المجموعة الشمسية in Arab have the unique id Q44559

This means that you can search for any of those terms in any of those languages and be gently redirected to the uniquely identified concept:

• search London in English
• search Londres in French
• search Londra in Italian
• search exoplanet in English
• search extrasolar planet in English
• search Экзапланета in Belarusian
• search كوكب خارج المجموعة الشمسية in Arab

This is thank to the fact that each of those items can have one main label and a series of aliases in each languages.

This disambiguation makes it a great base for a tagging system, the hard work of deduplicating concepts being done by the Wikidata community.

It also gives structured relationships between topics: see how Quora uses it

But building PeerTube tagging system on top of Wikidata would have another major virtue: it would link the tags to the Wikidata-aware web of data! We could for instance easily link to each others between PeerTube and Inventaire:

• books about London: https://inventaire.io/entity/wd:Q84 (API)
• books about Exoplanets: https://inventaire.io/entity/wd:Q44559 (API)
• page for the book "La Horde du Contrevent" and the people sharing a copy of it: https://inventaire.io/entity/wd:Q3209796

I would love to be able to suggest a list of videos on a given topic on those different Inventaire pages: if there was a way to query PeerTube instances APIs to aggregate such a list using Wikidata items ids, that would be crazily awesome :D

As a bonus (I don't see an application in PeerTube yet, but let's see), out of this possibility to link between systems using the Wikidata ids, Wikidata also provides you with identifiers in many databases: it can tell you that Q84 equivalent in OpenStreetMap is 65606 or 119861367 at BNF, etc. which allows to crosswalk from one identifier system to another:

• how to go from the BNF id to the Inventaire page: https://tools.wmflabs.org/hub/bnf:119861367?site=inventaire

As you can see, I'm quite enthusiastic about it, so feel free to ask for any further explanation ^^

Limits

The limit of this idea is that you depend on what is in Wikidata: for instance, you couldn't tag a video with a JDLL 2018 tag (there is JDLL (Q3187082) but not JDLL 2018) or a [insert your unknown association/school/new concept] tag (see Wikidata notability criteria). A suggestion to work around it: make the tag input autosuggest Wikidata items when typing, but if none is selected, still allow to create the tag: it will just be a plain string, not linked to anything but the other plain string tags.

Further readings

• Wikidata: The New Rosetta Stone

[picsi]

I second this, using Wikidata tags would greatly enrich the content metadata.

[picsi]

I second this, using Wikidata tags would greatly enrich the content metadata.