Modal Logic, Probability and Machine Learning Systems for Metadata Extraction - Ratio Mathematica 53 (2024)

Aug 2, 2026

Upload by Riley

Artificial intelligence since its inception has developed two major branches: logical reasoning and machine learning. Despite this, their interaction remains limited. The paper argues for tighter integration by using probabilistic modal logic and an infinite-valued Łukasiewicz component with a unary modality P to provide qualitative feedback on machine-extracted descriptive metadata, yielding a probabilistic modal logic system that assigns probability distributions to possible worlds of metadata truth values.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Abstract

Introduction

Background: AI, symbolic models, and probability

Motivation: probabilistic logic learning and data quality

Metadata in scientific and medical research

Precision medicine context and data quality needs

Ratio Mathematica Volume 53, 2024 Modal Logic, Probability and Machine Learning Systems for Metadata Extraction Simone Cuconato* Abstract Artificial intelligence (AI), since its inception, has had two major subfields, namely: logical reasoning and machine learning. Despite this, the interactions between these two fields have been relatively limited. In this paper, we highlight the need for closer integration of logical reasoning and machine learning. In our approach, logical reasoning tools such as probabilistic modal logic, are employed to provide qualitative feedback on the extracted descriptive metadata. The logical system we consider emerges from combining of S5 modal logic with the formulas of the infinite-valued Łukasiewicz logic and the unary modality P that describes the behaviour of probability functions. The result is a well-motivated system of probabilistic modal logic, that defines a probability distribution over possible worlds of the truth value of metadata extracted from precision medicine approach to Alzheimer’s disease articles through machine learning systems. By incorporating symbolic reasoning, we bridge the gap between the speed and efficiency of quantitative techniques and the need for precise, verifiable metadata in high-stakes environments such as medical research. Keywords: Modal logic; Probability; Logical reasoning; Machine learning systems; Metadata. 2020 AMS subject classifications: primary 03B45, secondary 03B52, 68T27 . 1 *University of Calabria, Department of Humanities, Via P. Bucci, Rende(CS) 87036, Italy; [simone.cuconato@unical.it. This](simone.cuconato@unical.it. This) article was written during my research stay as a Visiting Researcher at the Department of Computer Applications, Chitkara University, Punjab, India. 1 Received on April 10, 2024. Accepted on December26, 2024. Published on December31, 2024. DOI:10 .23755/rm.v53i0 .1592. ISSN: 1592-7415. eISSN: 2282-8214. ©Simone Cuconato. This paper is published under the CC-BY licence agreement. S. Cuconato 1 Introduction Symbolic and sub-symbolic represent the two main branches of artificial intelligence (AI) . AI has experienced rapid growth in recent years, thanks to the availability of powerful hardware, specific development platforms, and above alla massive amount of data. This rapid growth would never have been possible without a rigorous symbolic models based on articulated logical-mathematical systems [1, 2] . Logic and technology intertwine at all levels, from the lowest hardware level, governed by Boolean circuits and arithmetical operations in the stack memory; through the structure of assignment, sequencing, branching and iteration operations defining modern high-level programming languages; up to the equivalent abstract formulations of recursive definitions for algorithms [3] . Alongside logic, the branch of mathematics that plays a fundamental role in AI and data science today is undoubtedly statistics, and probability is the backbone of statistical inference and machine learning. For this reason, one of the central questions in data mining and AI concerns probabilistic logic learning, which involves the integration of relational or logical representations with probabilistic reasoning mechanisms and machine learning principles. Figure 1: Probabilistic Logic Learning as the intersection of the three domains. In this paper, we investigate the intersection of logic, probability, and machine learning by developing a symbolic approach based on probabilistic modal logic to qualitatively optimise the analysis of data extracted through machine learning systems [4] . Our approach aims to improve the consistency, accuracy, and completeness of the extracted data, ensuring higher data quality in critical applications. It is no coincidence that in recent years, the attention of computer scientists has shifted toward a complex and multidimensional concept: data quality [5] . Data quality is a multidimensional con

FAQs

What problem does the paper focus on?

It focuses on integrating logical reasoning with machine learning to improve the quality and trustworthiness of descriptive metadata extracted from precision medicine research.

How does the approach combine logic and probability?

It builds a probabilistic modal logic by combining S5 modal logic with infinite-valued Łukasiewicz logic and a unary modality P that captures the behavior of probability functions.

Why is metadata quality important in precision medicine?

In precision medicine, sharing high-quality data depends on linking the data with correct, verifiable, and complete metadata, which supports scientific information sharing and knowledge production.

Aug 2, 2026

Upload by Riley