Semantic MediaWiki

You've described most of open source software development and yet people still write the code because they need the code and want to share the code with others because others shared with them. This isn't the problem you think it is. The problem is that most of the web's current funding model is broken and user hostile. This isn't a problem with semantic data or software development this is a problem with the capitalist model of post-scarcity information production, distribution, and control. That's what needs to change if we're to benefit in proportion to what we contribute instead of those who appropriate and exploit the most benefiting the most. Let the web, as it exists now, break - it hasn't been ours for a long time.

> But had they been adopted for what they are: well defined metadata according to vocabularies and ontologies that following accepted, interoperable standards, the development of ground truth and knowledge graphs would have been much more advanced that the current state.

Yes, this is very true.

As an addendum, I think many people in industry don't quite realise how prevalent RDF and RDF-based ontologies are in academia. They're one of the main ways researchers integrate data. No one cares about lofty ideals, it's about the usefulness of RDF graph model and the various associated standards built on top of it.

Hard to say.

Depending on the context, read / write model, and how inference costs go, inferring on edit to classify and tag the content could prove more expensive than having users do it client-side on read.

LLMs ain't gonna be adding trustworthy metadata to pages on my wiki that can be useful for filtering (which is the entire purpose of SMW for us).

My guess it that there is some cross-pollination and benefits from and to both sides. Machine Learning / LLMs benefit from having higher quality data that they can be trained from or that you can use to automatically add more relevant context behind the scenes.

Semantic technologies or let's say "structured data entry" can benefit from having better auto-suggestions and validations from machine-learning.

The part that I find real interesting is how we put the users in between, so they're in full control and can use AI as an "assistant", but improve on it. And that the "structured data entry" helps the users to give him structure (that will later be helpful to the end-consumers).

SemanticMediaWiki is much older than wikidata. Some of the people involved were involved in both.

Big difference is that wikidata is an entirely separate wiki from client wikis, where smw is much more closely integrated into "articles". This allows wikidata to abstract across multiple language wikipedias. In practise that means that wikidata is often used to store factual data outside of articles that than gets pulled in via templates. SMW tends to more get used as a backend to turn mediawiki into a user programmable CRUD/workflow app (particularly with PageForms extension)

SMW is less concerned about scalability and performance. Wikidata puts anything of questionable scalability separate from the system (e.g. sparql api is isolated from rest). SMW would be unlikely to scale to anywhere near wikipedia levels in its current form.

Perhaps surprisingly, it isn't particularly useful for general encyclopedias like Wikipedia. Real life is messy; it doesn't always reduce cleanly to a set of key/value pairs.

Where extensions like SMW really shine is for wikis which discuss highly structured content, like content in video games. Another extension which I've seen used in this space is Cargo [1].

[1]: https://www.mediawiki.org/wiki/Extension:Cargo

Around 10 years ago I experimented with this to try to create a geographical hierarchy of articles in a wiki.

It was, in a nutshell, not great: there's a fundamental contradiction between wikis, which are very loosely structured and edited by humans who do all sorts of wacky unpredictable shit, and hierarchies, which need to be carefully pruned and maintained.

So this use case didn't work, and I think it's telling that (AFAIK) not a single Wikimedia project has adopted it.

The guild wars 2 wiki makes great use of it imo

Check out WikiChip for an example of it. https://en.wikichip.org/wiki/WikiChip

We used SMW at Liquipedia, an esports wiki. The idea was that all the data about players and teams could be exposed through SMW. Unfortunately we're now in the process of transitioning away from it due to major performance issues and data inconsistencies. It might be ok on smaller wikis but for us it's become a source of a lot of problems.

I used it on a popular wiki for adding metadata that allowed for decently powerful search filterings. It was important to do this because mediawiki's default search is/was absolutely horrible.

We never bothered doing anything like concepts, detailed types, etc...we only used it for basically adding tags to pages that you could query as part of a search. It worked okay with that. We had some issues where we wished a property could be an enumeration or nullable, which I don't think ever happened in one of their many releases.

Adding the metadata represented a bit of a challenge. We eventually had it generated from templates that would usually get filled out anyway.

Cf Wikipedia, it seems that indeed the support still exists: https://en.m.wikipedia.org/wiki/SMW%2B

Eventually I presume it all goes down to Michael Erdmann [@emwiemaikel on Twitter] having the right to maintain the codebase.