From language to action: Extraction and disambiguation of affordances in ModelAct

In this paper we focus on how information about concrete actions performed on food should be provided to IoT devices in terms of affordances extracted from corpora. Natural language processing has a role in defining which kind of knowledge devices interacting with machines and appliances should handle when humans send requests through natural language interfaces. We propose a model for the extraction of affordances of food from corpora and their role in sequences of procedural (sub)actions. The food processor of the future can find helpful this knowledge to interact with users suggesting alternatives in food processing in recipes steps and basic reasoning about preconditions and consequences in making meals.