A descriptive type foundation for RDF schema

This paper provides a type theoretic foundation for descriptive types that appear in Linked Data. Linked Data is data published on the Web according to principles and standards supported by the W3C. Such Linked Data is inherently messy: this is due to the fact that instead of being assigned a strict a priori schema, the schema is inferred a posteriori. Moreover, such a posteriori schema consists of opaque names that guide programmers, without prescribing structure. We employ what we call a descriptive type system for Linked Data. This descriptive type system differs from a traditional type system in that it provides hints or warnings rather than errors and evolves to describe the data while Linked Data is discovered at runtime. We explain how our descriptive type system allows RDF Schema inference mechanisms to be tightly coupled with domain specific scripting languages for Linked Data, enabling an interactive feedback to Web developers.