Sound and Complete Query Answering in Intensional P2P Data Integration

Contemporary use of the term ‘intension’ derives from the traditional logical doctrine that an idea has both an extension and an intension. In this paper we introduce an intensional FOL (First-order-logic) for P2P systems by fusing the Bealer’s intensional algebraic FOL with the S5 possible-world semantics of the Montague, we define the intensional equivalence relation for this logic and the weak deductive inference for it. The notion of ontology has become widespread in semantic Web. The meaning of concepts and views defined over some database ontology can be considered as intensional objects which have particular extension in some possible world: for instance in the actual world. Thus, non invasive mapping between completely independent peer databases in a P2P systems can be naturally specified by the set of couples of intensionally equivalent views, which have the same meaning (intension), over two different peers. Such a kind of mapping has very different semantics from the standard view-based mappings based on the material implication commonly used for Data Integration. We show how a P2P database system may be embedded into this intensional modal FOL, and how we are able to obtain a weak non-omniscient inference, which can be effectively implemented. For a query answering we consider non omniscient query agents and we define object-oriented class for them which implements as method the query rewriting algorithm. Finally, we show that this query answering algorithm is sound and complete w.r.t. the weak deduction of the P2P intensional logic.

💡 Research Summary

The paper proposes a novel logical framework for peer‑to‑peer (P2P) data integration that is based on the notion of intensional meaning rather than extensional data transfer. By combining Bealer’s intensional algebra with Montague’s possible‑world semantics for S5 modal logic, the authors define an intensional first‑order logic, denoted Łω, which extends ordinary first‑order logic with an abstraction operator h Aᵢ that turns any formula A into a term representing “that A”. This allows predicates such as “believe(x, h Aᵢ)” to be expressed without resorting to higher‑order syntax.

In Łω, the domain is split into extensional objects (the usual data values) and intensional objects (universals, propositions, and n‑ary relations‑in‑intension). An intensional algebra (Alg_int) supplies conjunction, disjunction, implication, and negation on intensional objects, together with a projection operator. Crucially, intensional equivalence between two queries q₁ and q₂ is captured by the modal formula ♦ q₁ ≡ ♦ q₂, meaning that in every possible world the extensions of q₁ and q₂ coincide. This equivalence is the foundation of the paper’s mapping concept.

A P2P network N consists of independent peers P_i = (Ω_i, M_i). Ω_i is the peer’s ontology (global schema) expressed in Łω; M_i is a set of view pairs (q_i, q_j) that link a view defined over Ω_i with a view defined over another peer’s ontology Ω_j. Each pair is required to be intensional‑equivalent, i.e., ♦ q_i ≡ ♦ q_j. When a user poses a conjunctive query q over peer P_i, the system uses the view pairs in M_i to locate intensional‑equivalent views on other peers, rewrites q into queries q′ over those peers, and aggregates the answers.

Traditional view‑based mappings rely on material implication (K_i q_i → K_j q_j) and thus force the transfer of actual data from one peer to another, potentially causing inconsistency. The intensional mapping (K_i q_i ≈ K_j q_j) instead only asserts that the two peers share the same meaning; no data is transferred, preserving each peer’s autonomy and allowing ontologies to evolve independently.

Because the set of possible worlds (all evolutions of the P2P system) is infinite, a fully omniscient reasoning engine would be impractical. The authors therefore introduce a weak, non‑omniscient deduction system. It consists of two inference rules: (1) a modal transition rule K_i φ → φ, which allows reasoning from a peer’s knowledge to the current world, and (2) the intensional equivalence rule ♦ φ ≡ ♦ ψ, which permits substitution of intensional‑equivalent queries. Using only these rules, a query agent can reason about the current world without exploring all hypothetical evolutions.

The operational component is an object‑oriented class QAgent. Its principal method, rewrite(q, P_i), performs the following steps: (a) scans the local view set of P_i to find all (q_i, q_j) pairs where q_i matches the user query; (b) generates the corresponding query q′ over the target peer P_j; (c) dispatches q′ to P_j, collects the result set; and (d) merges all results into the final answer for the original query. The paper proves that this algorithm is sound (every returned tuple is a true answer in some possible world) and complete (every tuple that is true in some possible world will be returned) with respect to the weak deduction system defined for Łω.

The contributions can be summarized as follows:

1. A new intensional modal first‑order logic Łω that unifies algebraic intensional semantics and S5 possible‑world semantics, providing a formal basis for meaning‑based P2P mappings.
2. The definition of intensional equivalence as the core of non‑invasive peer mappings, eliminating the need for a global schema and preserving peer independence.
3. A weak, non‑omniscient inference mechanism and a concrete query‑rewriting algorithm implemented as a QAgent class, together with rigorous soundness and completeness proofs.

The work addresses a key challenge in the Semantic Web: how to align heterogeneous ontologies without a central mediator. By focusing on intensional (meaning) equivalence rather than extensional (data) inclusion, the framework supports dynamic addition of peers, autonomous evolution of ontologies, and scalable query answering. Although the paper does not present empirical performance results, its theoretical foundations open a promising direction for robust, meaning‑driven P2P data integration and for future systems that combine distributed databases with grid‑computing resources.