Language-integrated semantic queries (LISeQ)

• Applicants: Professor Dr. Ralf Lämmel; Professor Dr. Steffen Staab
• Subject Area: Software Engineering and Programming Languages ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: from 2017 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 394523062

Graph-based data models allow for flexible data representation as needed for data exchange and integration. A graph-based data model includes structural information and it may include also, in the case of semantic data, schematic information as part of the data. The flexibility of graph-based data models makes them attractive, but programming with them is error-prone. A major reason is lack of type-safe integration into programming languages.Schema descriptions which are formally represented as logic-based data descriptions in the graph data itself are difficult to represent in programming languages. Problems arise due to multiple inheritance, mixture of structural and nominal typing, incomplete knowledge and sheer number of types in typical data sources. Type-safe integration must cover both representation of data descriptions and type-safe, expressive data access such that each query is viewed as defining implicitly a new type.Overall, type-safe integration requires: (R1) type-checked querying; (R2) an integration of queries in the programming language including polymorphism and type-inference to deal with the multitude of types; (R3) constraints for addressing potential null exceptions due the underlying open-world assumption for graph data in programs; (R4) an empirical validation regarding the effectiveness of the corresponding programming language extensions.Overall, the LISeQ project aims at the type-safe integration of graph-based data models into programming. In order to respond to the above requirements, we pursue the following objectives in LISeQ: (O1) typing of SPARQL queries including implicitly defined query types and leveraging schematic descriptions in the given graph data; (O2) static typing including type inference and polymorphism while meta-properties for the type system, such as type-safety, are shown in a research language; (O3) an empirical evaluation of LISeQ based on an open-source Scala compiler extension.We will formalize the schema of the graph data models in OWL, a standardized family of languages based on description logics. The considerable formal foundation and expressiveness of OWL lets us generalize over many simpler schema languages and it leads to expressive formalization of implicitly defined query types. Thus, OWL will be the basis for embedding semantic graph queries in programming languages. For the embedding, we build upon an established approach in the field of programming language design. Starting from a basic lambda calculus, we extend it to respond to the variety of challenges that arise when combining functional languages with complex logic-based data descriptions.LISeQ is an interdisciplinary research effort with WeST being renowned for its research in Semantic Web and Softlang being renowned in Language Design and Analysis. LISeQ will exploit methods from both fields of research and empirically evaluate these new ideas in a Scala-based extension.

• DFG Programme: Research Grants