Shapes from natural language in Verbal Image