The linguistic ingredients of recipes: aspect, causation, and argument structure

Recipes outline goal-oriented procedures involving sequential actions with specific durations aimed at achieving precise outcomes. Ingredients are moved into different containers and undergo various processes, ultimately turning into new entities. The structure and complexity of recipes make them an exciting domain for AI and Natural Language Understanding research, for the challenges they pose and their potential for establishing benchmarks. Accurate understanding of procedures is paramount and involves correct interpretation of the sequence and duration of the various processes, as well as of the causal relationships expressed in each step. This paper aims to provide a thorough linguistic and textual analysis of recipe texts, exploring their nature, structure, and genre variations. It provides an in-depth examination of their primary linguistic functions, adopting a multi-language approach that involves analysing authentic recipes in typologically distinct languages such as English, Italian, and Chinese. The analysis consists of a function-form study of four main semantic frames, aimed at determining how different languages encode temporal and aspectual information, causal relationships such as caused motion and caused change of state, and resultant objects.