The management of natural resources is becoming increasingly relevant due to its direct implication in society's life. Thus, individuals must make decisions based on environmental and social aspects. This work uses a chatbot to support users' decisions through an RPG scenario based on the participatory management of resources in the Lagoa Mirim Watershed and Canal São Gonçalo Basin. In this context, in addition to the chatbot, this study presents a pollution predictor to support decision-making, with a determination coefficient of 0.99, constructed using random forest. Also, we present five Word Embeddings models to expand the natural language understanding, based on a corpus of about 700 thousand sentences, capable of identifying relations between words.