Creating a MongoDB-backed corpus reader