Interplay of Open Semantic Interchange and Open Data Contract Standard

[supertramp17]

Curious to know what the community thinks of how OSI and ODSC interplay might look like. While OSI is useful for business semantics and analytics, ODSC provides certain operational guarantees.

An enriched schema/metadata spec will help generate a good semantic model which can be represented in OSI format.

Does it make sense to support some form of integration between the two formats? Possible options could be

1. For a given semantic model, include a data contract field or object structure. This could include whether or not schema is ODCS compliant, version number etc.
2. Inherit schema from a published ODCS schema
3. Add a metadata contract section in OSI which could map to ODCS

[jochenchrist]

Jochen here, TSC Member at Bitol for ODCS and ODPS, and OSI working group member for ontologies.

Great question!
While it is somewhat overlapping, data contracts and semantic models are complementary standards that work together to create a complete, well-governed data ecosystem.

A data contract, defined with the Open Data Contract Standard (ODCS), specifies the agreement between a data producer and its consumers: the schema, data quality rules, SLAs, ownership, and terms of use, mostly from a trust and governance perspective, but also testable via quality engine. It answers: "What data is available, who owns it, and what guarantees come with it?" Data Contracts can be applied for any type of data products (including tables and schemas, semantic models, folders, binary files, kafka events, APIs.

A semantic model, defined with OSI, is mostly used in the context of BI applications, it adds business meaning on top of the raw data: metrics, relationships, dimensions, and AI context. It answers: "What does this data mean, how do tables relate, and how should metrics be calculated?" It does not include terms of use or SLAs.

Does this help to clarify?

[supertramp17]

Thanks @jochenchrist for the response.
Would it be good to have a "generated_by" field ? Given that it's common for AI Agents or workflows to auto generate files, it would be good to disclose how the files were generated and what inputs were used or considered. We could make the OSI file as self contained as it can be. This applies to odcs schema files too.
Wdyt ?