Mastodawn

Michael Roberts

Hey @doriantaylor, I have some more dumb semantic web questions.

1. If I *don't* want to use a full existing ontology, because I intentionally want to have a simpler model during early phases of a project, is there a .. I guess a "sanctioned way" to enhance my ontology as time goes on without losing fidelity?

I know OWL's SameAs can identify individual resources as the same as some other resource, but this would kind of be that, only at the ontological level, maybe.

#SemanticWeb #RDF

Michael Roberts Jul 12

@doriantaylor
2. The entire point is to be able to merge disparate data sources into one graph, right? So let's say I have one database with records describing my print books, and I have some manner of dump from Calibre of my ebooks, and I have an additional table of information about PDFs in my possession - what's the "normal people" way of building a semantic graph with those three data sources?

Trying to articulate these questions in a way that I can even start searching for answers.

Michael Roberts Jul 12

@doriantaylor So like, in SQL I might have those three data sources in separate tables, but I could define a union view to query all three at once. I'm vaguely aware of seeing something like this in SPARQL somewhere (like, incorporating Wikidata into a query, maybe?) but I'm still working on integrating my picture of how it fits together.

Michael Roberts Jul 12

@doriantaylor I may have answered my own question. It's a federated query, I think.

Michael Roberts Jul 12

@doriantaylor How come everybody isn't using this stuff for everything? What have I been doing with my life?

@vivtek a lot of it is still really frustratingly missing

Michael Roberts Jul 12

@doriantaylor For real. Also *incredibly* badly documented.

Michael Roberts Jul 12

@doriantaylor In the sense of "everybody presents a single graph of five nodes and never explains anything interesting". I'll find better books soon enough, I'm sure.

@vivtek yeah you can pull in remote sources with federated queries but you can also have named graphs that are local

Michael Roberts Jul 12

@doriantaylor What I really want to be able to do is have a conventional SQL table of some stuff managed by regular CRUD actions but still have it mergeable with other, more free-form data in a separate graph.

@vivtek go see r2rml https://www.w3.org/TR/r2rml/

R2RML: RDB to RDF Mapping Language

Michael Roberts Jul 13

@doriantaylor Yeah, I found d2rq, too, but it was mothballed.

@vivtek so many vocabs/tools for rdf are abandonware, like most of them imo

@vivtek aside: i wrote about the lack of tooling a while back https://doriantaylor.com/there-is-no-sqlite-for-rdf — but it turns out there is actually this: https://github.com/oxigraph/oxigraph

Michael Roberts Jul 14

@doriantaylor You last mentioned Oxigraph to me in 2022.

@vivtek ah, i still haven't used it lol (i'm actually gonna make my own rdf store)

Michael Roberts Jul 14

@doriantaylor Oh, wait, no, my notes are muddy on chronology. In 2022 you posted that there was no SQLite for RDF, but in 2024 you mentioned Oxigraph.

Michael Roberts Jul 14

@doriantaylor Anyway, I'm drifting in the direction of building my own store as well.

@vivtek i'm just gonna prototype one in ruby and if it works out i'll maybe try writing it in rust or something i dunno

Michael Roberts Jul 14

@doriantaylor I'm drifting towards writing a store, too, and an Attean driver for it. But despite the fact that it seems to have captured the full attention of the Muse the past couple of weeks, semantic data really is a lower priority for me. So God knows what I'll actually do.

@vivtek one thing i am emphatically not interested in though is writing a sparql processor if i can get away with it; i'm mostly interested in stuff like transaction history and being able to have multiuser quad stores that can share space, as well as some kind of basic inferencing layer

@vivtek to add: the thing about sparql that's hard is a cost-based optimizer. right now putting a network-addressable sparql endpoint up is begging to get dosed because it's not only possible but easy to craft ultra-expensive queries, and you don't want to find this out by running them first, like you wanna be able to say "403: haha dick i'm not executing that"