Introducing Fuzzy Labels to Agent-Generated Textual Descriptions of Incomplete City-Traffic States

An aim of this research is to create methods for a provision of textual information to users of a distributed multi-agent information system. In particular we focus on a traffic information system where agents transform the numerical data about states of the city traffic obtained using a distributed sensor network into natural language summaries. The basis for the transformation from numerical data into a linguistic domain are zadehian fuzzy-linguistic models of concepts. Unlike in typical Natural Language Generation approaches, this paper focuses on the provision of summaries in situations where data is incomplete and on conveying this incompleteness to the user using belief-based language statements. We provide an algorithm based on a theory of grounding for an agent-based evaluation of local summaries with autoepistemic operators of possibility, belief, and knowledge. We also propose a method for an aggregation of summaries generated by local agents in order to obtain a textual summary of complex structures of the road network (e.g. areas, districts, precise routes).