Temporal Description Logic for Ontology-Based Data Access

Alessandro Artale, Roman Kontchakov, Frank Wolter, Michael ZakharyaschevIJCAI 2013

TL;DR

This paper introduces TQL, a novel temporal description logic designed for ontology-based data access (OBDA). TQL extends the OWL 2 QL standard to handle temporal data and conceptual modeling, crucially maintaining the ability to rewrite conjunctive queries into first-order queries for efficient evaluation over databases.

What Problem Does This Paper Address?

Ontology-Based Data Access (OBDA) is a powerful paradigm for querying data sources using ontologies to enrich conceptual understanding. While OWL 2 QL is the current standard for OBDA, it lacks the ability to model temporal aspects of data, such as events with validity times (e.g., employment contracts). Existing temporal description logics often sacrifice the crucial property of first-order rewritability, making efficient query answering difficult. This paper bridges this gap by proposing TQL, a temporal extension of OWL 2 QL that supports temporal conceptual modeling while preserving first-order rewritability for conjunctive queries.

Methods

The core contribution is the design of TQL, a temporal description logic that extends OWL 2 QL. TQL interprets ontologies over sequences of temporal interpretations, allowing concepts and roles to evolve over time. It introduces temporal operators 'sometime in the past' (◊p) and 'sometime in the future' (◊F) for temporal conceptual modeling. A key technical achievement is the development of a method to rewrite conjunctive queries (CQs) over TQL knowledge bases into unions of CQs (UCQs) that can be evaluated over a temporalized ABox (instance data augmented with temporal precedence relations). This rewriting process involves constructing "canonical models" and leveraging monotonicity lemmas for temporal reasoning. The paper also addresses the challenge of arbitrary TBoxes by reducing them to a "normal form" and extending the rewriting techniques.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Bridging the Temporal Gap in OBDA: The paper successfully integrates temporal reasoning into the OWL 2 QL OBDA framework, enabling richer modeling of real-world scenarios where data is time-dependent.
  • Preserving First-Order Rewritability: A significant technical feat is maintaining the first-order rewritability of queries, which is essential for efficient query processing in practical OBDA systems. This is achieved through careful design of the logic and the rewriting process.
  • Temporal Conceptual Modeling: TQL provides expressive means to model temporal aspects of concepts and roles, such as rigidity (time-independence), persistence, and instantaneousness, going beyond simple time-stamping of facts.
  • Canonical Models as a Tool: The concept of canonical models plays a crucial role in proving the correctness of the rewriting procedure, providing a structured representation of the entailed knowledge.
  • Monotonicity is Key: The paper highlights the importance of temporal monotonicity properties for proving the correctness of rewritings and for simplifying the temporal reasoning process.

Key Results

The paper presents a main theorem (Theorem 2) stating that for any TQL TBox T and CQ q, a UCQ q' can be constructed such that answering q over a TQL KB (T, A) is equivalent to evaluating q' over A augmented with temporal precedence relations. This demonstrates the feasibility of efficient query answering for temporal OBDA. The paper also establishes that certain extensions to TQL, such as next/previous time operators or certain types of temporal axioms, would lead to non-rewritability, suggesting TQL strikes a good balance between expressivity and tractability.

Limitations and Future Work

The authors acknowledge that their focus was on the existence of rewritings, not on their efficiency. Future work includes investigating efficient rewriting techniques, exploring temporal intervals in data representation, and extending the expressive power of TQL. Specifically, they suggest that adding next/previous time operators (that do not form cycles) might still preserve rewritability, and allowing arithmetic operations on temporal terms could enhance query expressivity.

Summary updated

BibTeX
@inproceedings{artale2013temporal,
  author = {Alessandro Artale and Roman Kontchakov and Frank Wolter and Michael Zakharyaschev},
  title = {{Temporal Description Logic for Ontology-Based Data Access}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2013},
}
Figure