Requirements specification is an early stage of system design. It consists of rephrasing and documenting stakeholders' explanations and needs in the form of clear and coherent requirements. However, these requirements are often expressed in natural language since it is the easiest communication method. Researchers have proposed semi-structured natural language templates or boilerplates for specifying functional and nonfunctional requirements, which consider security requirements. This seeks to enhance the quality of the requirements specifications and simplify their transformation to system models. However, it is still unknown what concepts, quality attributes, and good practices should be considered to specify requirements in a semi-structured natural language and how that information has been considered in the existing templates. In this paper, we aim to determine how templates are related among them and what are the implications (e.g., complexity, completeness, time) of using one or another. In this paper, we identify each template's concepts, quality attributes, and good practices by studying the template's aspects and then using a running example to formulate requirements using these templates. We also identify the aspects repeated or inherited from one template to another. This paper puts forward a taxonomy of requirements specification templates that categorize and specify the sources of the templates, which helps determine what template considers the aspects of another.