Support for Vectors and Matrices

[AdamSobieski]

Hello. I would like to request a feature for SPARQL 1.2: support for linear algebraic datatypes and operations. This would include support for built-in datatypes for vectors and matrices and support, through language syntax or built-in functions, for linear algebraic operations.

See also

• Easier addition of support for custom datatypes to SPARQL endpoints #130
• Representing and Describing Vector Embeddings and Models onnx/onnx#4318
• Semantics and Embedding Vectors (https://lists.w3.org/Archives/Public/semantic-web/2022Oct/0028.html)
• SPARQL Function: LDScript: a Linked Data Script Language (http://ns.inria.fr/sparql-extension/)
• Scientific SPARQL: Semantic Web Queries over Scientific Data (https://ieeexplore.ieee.org/document/6313648)
• Vec2SPARQL: integrating SPARQL queries and knowledge graph embeddings (https://www.biorxiv.org/content/10.1101/463778v1)
• Embeddings for the Lexicon: Modelling and Representation (https://aclanthology.org/2021.semdeep-1.3.pdf)

[VladimirAlexiev]

• add an ML Model type to schema.org, likely under /Dataset, ... and potentially other major types (SingleTabularDataset, MultipleTabularDataset?) schemaorg/schemaorg#3140 for adding "schema:MLDataset"
• https://mlcommons.org/working-groups/data/croissant for a similar effort by MLcommons that's already brought fruit
• specialized stores like TensorStore
• binary formats like HDF5. And there's SPARQL over HDF5
• Data Abstraction Layers (eg GDAL) to isolate data access from the specific binary format/storage used.