NoSQL systems have gained popularity in recent years due to their capability to deal with huge amounts of information, even in an unstructured way. The schemaless nature of NoSQL databases leads to the widespread opinion that no modeling activity is required for them. However, in the literature, many researchers recognized the need for the classical three levels of modeling also in this kind of system. This paper proposes a solution for the conceptual and logical modeling of a document-based database, which starts from the well-known ER conceptual model and automatically translates it into a set of document collections through predefined labeling and translation rules. The usage of a standard conceptual model, like the ER one, and the JSON Schema language as a document-based logical model, promotes the usage of the proposed rules, as well as the modeling activity in the NoSQL world.