Abstract:We present a novel approach to querying classical inconsistent description logic (DL) knowledge bases by adopting a~paraconsistent semantics with the four Belnapian values: exactly true ($\mathbf{T}$), exactly false ($\mathbf{F}$), both ($\mathbf{B}$), and neither ($\mathbf{N}$). In contrast to prior studies on paraconsistent DLs, we allow truth value operators in the query language, which can be used to differentiate between answers having contradictory evidence and those having only positive evidence. We present a reduction to classical DL query answering that allows us to pinpoint the precise combined and data complexity of answering queries with values in paraconsistent $\mathcal{ALCHI}$ and its sublogics. Notably, we show that tractable data complexity is retained for Horn DLs. We present a comparison with repair-based inconsistency-tolerant semantics, showing that the two approaches are incomparable.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the problem of handling contradictory information encountered in the query of classical inconsistent description logic (DL) knowledge bases. Specifically, the article introduces a new method to query classical DL knowledge bases containing contradictory information. By adopting four - valued intuitionistic semantics (namely exactly true \(T\), exactly false \(F\), both true and false \(B\), and neither true nor false \(N\)), and allowing the use of truth - value operators in the query language to distinguish answers with contradictory evidence and those with only positive evidence.
### Main Problems and Challenges
1. **The Fragility of Traditional OMQA**:
- Traditional Ontology - Mediated Query Answering (OMQA) lacks robustness in the face of contradictory information because every Boolean query will draw conclusions from an inconsistent knowledge base.
2. **The Limitations of Repair Semantics**:
- Although repair semantics (such as AR semantics, brave semantics, and IAR semantics) can handle data inconsistency, they are defined based on consistent subsets and cannot fully utilize the advantages of four - valued logic.
3. **The Deficiencies of Existing Paraconsistent DL Query Languages**:
- Existing Paraconsistent DL query languages (for example, the work of Nguyen and Szałas (2012) and Zhou et al. (2012)) have a drawback: given a knowledge base \(K\) and a concept \(A\), it is not possible to write a query \(q\) such that the answer set contains only those individuals \(a\) for which \(A(a)\) is exactly true.
### Main Contributions of the Paper
1. **Introduction of a New Query Language**:
- A new query language is introduced, which extends the existing query language by adding truth - value operators and can distinguish query answers of "at least true" and "exactly true".
2. **Computational Complexity Analysis**:
- The computational properties of this new query language are explored, and by transforming it into classical OMQA, it is proved that the data and combined complexity of Paraconsistent query answering in Horn description logic ontologies are the same as those of classical OMQA.
3. **Comparison with Repair Semantics**:
- The Paraconsistent OMQA semantics and repair semantics are compared, showing the incomparability of the two methods. In particular, for queries with the \(T\) operator (i.e., exactly true), it neither over - approximates IAR semantics nor underestimates brave semantics and CAR semantics.
### Summary
This paper aims to improve the ability to handle DL knowledge bases containing contradictory information by introducing four - valued Paraconsistent semantics and a new query language, and provides theoretical support and computational complexity analysis for this new method. In addition, it also explores the relationship between Paraconsistent semantics and repair semantics, revealing their different characteristics in handling contradictory information.