Re: Supporting Arrays and Tensors

Piotr,

Thank you. I added a hyperlink to and example from your specification document.

Time-series (prediction) are very interesting use cases. Another use case, defining terminology and concepts while being able to include embedding vectors from one or more AI models, is already possible with your specification. It also defines SPARQL similarity functions: "tensor:cosineSimilarity" and "tensor:euclideanDistance".


Best regards,
Adam


________________________________
From: Piotr Sowiński <psowinski17@gmail.com>
Sent: Monday, February 16, 2026 6:41 AM
To: semantic-web@w3.org <semantic-web@w3.org>
Cc: nikita@neverblink.eu <nikita@neverblink.eu>
Subject: Re: Supporting Arrays and Tensors


Hi Adam,


You are welcome to contribute to the existing implementation of tensors in RDF: https://w3id.org/rdf-tensor/

GitHub repo: https://github.com/NeverBlink-OSS/rdf-tensor


If it does not address your use case, or you see room for improvement, please file an issue or open a PR.


We are in the process of rewriting the spec to be much more portable and testable, so there will be quite a few improvements in the next few weeks.


Recently, I also talked to some time series folks who want to use this work for their use cases. We are adapting the spec to make it easier for them to handle tensors of varying lengths. So – we are open to new use cases and new ideas :) you are welcome to express yours as well.


Piotr Sowiński
On 2/16/26 03:45, Adam Sobieski wrote:
Semantic Web Interest Group,

A document collecting together ideas for representing arrays and tensors (e.g., numerical vectors and matrices) in RDF languages is now available at:

https://github.com/AdamSobieski/Narratology/blob/main/Content/The%20Semantic%20Web,%20Arrays,%20and%20Tensors.md

Thank you for those ideas and approaches already shared, in this mailing list, and for any other ideas about how one might go about representing arrays and tensors.


Best regards,
Adam Sobieski