Jaśkowski–Fitch Natural Deduction Proofs for Metadata Extraction Through Machine Learning Systems

In the present paper, I will focus on propositional knowledge, proof theory, and the deductive transmission of information in data mining. The application of proof theory to machine learning-based metadata extraction systems will allow us to intend metadata extraction as mathematical objects. More in detail, I will consider a recent first-order logic (FOL)-based applied ontology for metadata extraction, by the name MAke Decision for Metadata Extraction (MADME), and I will build a specific logical system for declarative sentences applied to the extraction of metadata through machine learning systems. MADME’s strength is that it bases its decision-making procedure on a rigorous logical structure. The objective of this paper is to demonstrate how the logic in MADME can be optimized using Jaśkowski–Fitch natural deduction proofs. Jaśkowski–Fitch systems will determine new operations among the best machine learning-based metadata extraction systems from heterogeneous documents sources and, consequently, will support the research community in knowledge extraction and management.