Re: Semantics and Embedding Vectors

Am Di., 11. Okt. 2022 um 18:12 Uhr schrieb Adam Sobieski <
adamsobieski@hotmail.com>:

> Thank you for the information about FrAC. It appears that, with it, we can
> anchor concepts/items to static and contextualized embedding vectors.
>

It's meant to, at least. As FrAC development is coming to a close, I'd be
curious to learn whether people get along with the naming and modelling
decisions we agreed upon so far. FrAC development has been driven mostly by
people from lexicography, NLP and linguistics (and SW, of course), mostly
with an academic background, so if there are naming decisions that people
feel a mainstream developer might have issues with, we can still take that
into consideration.


>  Perhaps, someday, we’ll be able to do mathematics, e.g., dot products and
> lengths, with first-class vectors in SPARQL.
>

To some extent, that can already done with custom functions, and I remember
a biomed paper proposing exactly that a few years back. As for making
embeddings first-class citizens in SPARQL, I would love to see that -- and
a have a bit of a radical idea on that, but it's not fully worked out, yet.
Whenever I have the time to put it on paper and run a few tests, I'd like
to discuss it here.

Best,
Christian

>