What is ontology-based data access in knowledge graphs?

Ontology-Based Data Access (OBDA) is a method for querying data through a conceptual layer defined by an ontology, which acts as a bridge between user queries and the underlying data sources. In knowledge graphs, which organize data as interconnected entities and relationships, an ontology provides a formal model of the domain—defining classes (categories), properties (relationships), and rules (e.g., inheritance, constraints). OBDA lets developers interact with the data using high-level terms from the ontology, while the system handles the translation of these queries into queries executable on the raw data (e.g., relational databases, CSV files). This approach decouples the way users think about the data (via the ontology) from how it is physically stored.

A typical OBDA system has three components: the ontology, mappings, and data sources. The ontology, often written in OWL or RDFS, defines concepts like “Employee” or “Department” and relationships like “worksIn.” Mappings link these concepts to the actual data structures—for example, connecting the ontology’s “worksIn” property to a SQL database’s employee.department_id column. When a developer writes a SPARQL query (e.g., “List all employees in the IT department”), the OBDA engine uses the mappings to generate an optimized SQL query, joins tables, and returns results. Tools like Ontop or Stardog automate this process, enabling SPARQL-to-SQL translation. This abstraction is particularly useful when data is distributed across multiple formats (e.g., some in PostgreSQL, some in MongoDB), as the ontology unifies these sources under a single schema.

The key benefit of OBDA is that it simplifies querying for developers, especially in complex, heterogeneous environments. For instance, if an ontology defines “Manager” as a subclass of “Employee,” a query for “all Employees” automatically includes Managers, even if they’re stored in a separate database table. This reduces the need for manual query rewriting when schemas change. However, performance can be a challenge: complex mappings or large datasets may require optimization (e.g., caching frequent queries). OBDA is widely used in enterprise knowledge graphs, where integrating legacy systems or evolving data structures is common. For example, a healthcare project might use OBDA to query patient data scattered across EHR systems, lab databases, and research repositories, using a unified ontology for terms like “Diagnosis” or “Treatment.” Developers can focus on domain logic instead of data plumbing.