Bridging the Gap Between AI and Reality • Rhodes, Greece
Time: Tuesday, 4.11
Room: Room A
Authors: Marco Krumrey
Abstract: Active Automata Learning is increasingly applied in software quality assurance, yet its practical adoption is often hindered by the challenge of defining input alphabets with adequate abstraction. To address this obstacle, prior work introduced ALEX (Automata Learning EXperience), a tool supporting the streamlined creation and management of symbols and learning setups for web applications. Initially, alphabet symbols were modeled by QA teams; this process was later shifted to web developers through iHTML (instrumented HTML), a domain-specific language that augments HTML elements with instructions for deriving alphabet symbols (e.g., associating buttons with click events and identifiers). Building on this foundation, the current project explores the use of large language models to automatically identify frontend user interactions and generate corresponding abstract and concrete alphabet symbols for ALEX. The method follows a two-step process: natural language descriptions of possible interactions are first presented to the user, and subsequently refined into formal symbol definitions, enabling more efficient and accurate modeling.