Companies are dealing with many cognitive changes with the introduction of the Industry 4.0 paradigm. In this constantly changing environment, knowledge management is a key factor. Dialog systems, being able to hold a conversation with humans, could support the knowledge management in business environment. Although, these systems are currently hand-coded and need the intervention of a human being in writing all the possible questions and answers, and then planning the interactions. This process, besides being time-consuming, is not scalable. Conversely, a dialog system, also referred to as chatbot, can be built from scratch by simply extracting rules from technical documentation. So, the goal of this research is designing a methodology for automatic building of human-machine conversational system, able to interact in an industrial environment. An initial taxonomy, containing entities expected to be found in maintenance manuals, is used to identify the relevant sentences of a manual provided by the company BOBST SA and applying text mining techniques, it is automatically expanded. The final result is a taxonomy network representing the entities and their relation, that will be used in future works for managing the interactions of a maintenance chatbot.