Category Theory for Semantic Data Interoperability

Présenté par : Ryan Wisnesky  Kenny MacKenzie  

Every engineering organization struggles to exchange and consolidate data reliably across their many data sources. Attempting to build integrations between all relevant systems is impractical with current mainstream practices, because of the quadratic relationship between applications and the number of integrations between them. To avoid the exploding cost of scaling ad-hoc integrations, organizations may attempt to standardize the application semantics. This approach hasn’t worked in practice because it's impractical to get many people from diverse domains to agree on a single perspective, and nuanced domain-specific meaning is always sacrificed.

Category Theory is a mathematical language for encoding and computing semantic structures across contexts, and has been used by mathematicians and scientists to formally communicate meaning across domains. Recent developments out of MIT have worked out how to apply the approach to data schemas and are leading to a paradigm shift in semantic data interoperability. There are several benefits, but two key benefits are compositionality and machine verification. Integrations can be added together and checked for integrity, which means that a quadratic number of integrations can be inferred from a linear input with mathematical guarantees that data won’t be corrupted. A third benefit is that the approach is agnostic to any specific data structure and can interoperate across all of them (SQL, XML, Json, Graph, RDF, etc). In this talk we will outline a classic challenge facing industrial engineering and how approaches built from Category Theory by Conexus can offer a new paradigm of solutions.

references

Presentation Video (Similar to proposed)

Presentation Slides (Similar to proposed)

Research Papers