Logic-Based Approaches to Scheduling and Their Enhancement through Controlled Natural Languages

In the healthcare domain, the effective management of healthcare operations, such as patient scheduling, staff rostering, and resource allocation, remains a persistent challenge due to complex constraints and dynamic environments. Artificial Intelligence offers promising methods for optimizing these processes, particularly through Knowledge Representation and Reasoning (KRR) approaches that ensure explainable and logically grounded solutions. Among these, Answer Set Programming (ASP), also thanks to the presence of efficient solvers, has proven to be particularly suitable for modeling complex healthcare scheduling problems. Thus, in this thesis, we present solutions to two healthcare scheduling problems: the Pre-Operative Assessment Clinic scheduling problem, which involves assigning patients to days for examination and surgical preparation, and the scheduling of Periodic Treatments problem, where patients must follow predetermined treatment plans over a period of several weeks. The use of formal KRR languages like ASP, however, often requires specialized expertise, limiting accessibility for domain experts. To bridge this gap, Controlled Natural Languages (CNLs) have emerged as a means to express formal logic in a more intuitive and human-readable format. CNLs can enhance readability, reduce development time, and facilitate communication between technical and non-technical stakeholders. In this thesis, we propose a set of tools in this regard. First, we introduce CNL2ASP, which is able to convert sentences written in CNL into an ASP encoding. Then, the second tool is CNL2TEL. It extends CNL2ASP with temporal operators, and is able to convert CNL specifications into the TELINGO input language. Similarly, SBVR2ASP is a tool that translates specifications written in the Semantics of Business Vocabulary and Business Rules (SBVR) into ASP. A recent survey has highlighted several limitations in current approaches to SBVR conflict detection and analysis, and in our approach based on ASP, we address most of these limitations. Finally, the last tool we propose is CNLWizard, a framework that generates grammars for target representation languages, enabling the translation of problems stated in CNL into formal representations. This tool offers a flexible, high-level approach to defining desired grammars, significantly reducing the time and effort needed to create custom grammars.