TLDR: The TPTP World, a key infrastructure for Automated Theorem Proving (ATP), has been extended to support non-classical logics. This includes new languages (NTF), a problem library, and tool support, enabling ATP systems to handle a wider range of real-world applications in AI, philosophy, and computer science by standardizing non-classical problem representation and solution reporting.
The TPTP World, a well-established framework for Automated Theorem Proving (ATP) systems, has expanded its capabilities to include non-classical logics. This significant development, detailed in a recent paper by Alexander Steen and Geoff Sutcliffe, marks a crucial step towards bridging the gap between classical and non-classical reasoning in practical applications.
For years, the TPTP World has been a cornerstone for researchers and developers in ATP, providing a comprehensive library of problems, solutions, and standards for various classical logics. However, many real-world applications, particularly in artificial intelligence, philosophy, natural language semantics, and computer science, require the nuances of non-classical logics. These include areas like knowledge representation, planning, multi-agent systems, formal ethics, and software verification. Historically, the development of ATP systems for non-classical logics faced challenges due to diverse input formats and inconsistent result reporting, hindering their widespread adoption and interoperability.
The extension of the TPTP infrastructure to support non-classical logics began around 2015. This initiative aimed to create a uniform and integrated environment for both classical and non-classical ATP. The result is a new family of languages, known as Non-classical Typed Form (NTF), which are built upon the existing classical TPTP languages (Typed Extended First-Order Form – TXF, and Typed Higher-Order Form – THF). These new languages, NXF and NHF, introduce non-classical connectives and a standardized way to specify the exact logic being used for a problem.
The NTF languages are designed to be syntactically consistent with their classical counterparts, offering a uniform syntax for a wide range of non-classical logics while requiring minimal changes to existing software. Non-classical connectives, such as those for necessity and possibility in modal logics, are enclosed in braces (e.g., {$box} for necessity or {$possible} for possibility). These can be simple unary connectives or more complex ones with parameters, allowing for flexibility in representing various logical systems like multi-modal logics or epistemic logics.
A key innovation is the “logic specification,” a meta-information tag within the TPTP file that explicitly defines the non-classical logic of a problem. This is crucial because the same language syntax can be used for different non-classical logics, and this specification ensures clarity and reproducibility. For instance, in quantified normal multi-modal logic, the specification can detail properties like domain constancy (whether individual domains are the same across different worlds), designation (rigid or flexible interpretation of symbols), term interpretation (local or global), and the specific modal axiom schemes (e.g., K, M, S4, S5) that characterize the modalities.
The TPTP problem library has been updated to include non-classical logic problems written in the NTF language. As of v9.1.0, there are over 200 NTF problems, covering various types of modal logics and their semantic properties. This expansion provides a rich set of benchmarks for ATP system development.
Several ATP systems have been integrated into the TPTP World to handle these new non-classical problems, including KSP, nanoCoP-M, MleanCoP, and Leo-III. Some of these systems use direct non-classical reasoning, while others, like Leo-III, translate non-classical problems into classical logic problems that can then be solved by powerful classical ATP systems like E and Vampire. This translation approach, facilitated by tools like ATFLET, demonstrates the versatility of the TPTP framework.
The TPTP World also provides extensive tool support for non-classical logics, including parsers, printers, derivation verifiers (GDV), and model verifiers (AGMV). These tools help in processing, checking, and visualizing both problems and their solutions, including Kripke interpretations that illustrate the semantics of modal logics. An interactive derivation viewer (IDV) and an interactive interpretation viewer (IIV) further enhance the user experience by allowing detailed examination of proofs and models.
Also Read:
- Streamlining Answer Set Programming with Constant Set Reduction
- Defining Data Preferences for Inconsistent Knowledge Bases
In summary, the integration of non-classical logics into the TPTP World represents a significant advancement for automated theorem proving. It provides a standardized, flexible, and comprehensive infrastructure that supports research, development, and deployment of ATP systems for a broader range of real-world applications. This work paves the way for further exploration and standardization of other non-classical logics, aiming for a more unified and accessible environment for automated reasoning. For more details, you can refer to the full research paper available here.
