In the past years, it has increased the demand for hypermedia applications, which relate objects of audio, video, text, images, etc. in time and space. A hypermedia application is formally specified in a document. It is usual to group a set of applications structurally and semantically similar to each other as a documents family. Conceptual models for hypermedia authoring are the basis for documents specification, but they are not satisfactory to define documents families. This work presents as contribution a method for hypermedia authoring based in the identification of these families and their instantiation in order to create new documents. This method addresses the hypermedia document authoring as a social practice, where specialist authors collaborate to help and to ease the task of less skilled authors. Another contribution of this work is to define a language for the specification of these documents families, named TAL (Template Authoring Language), which can be applied in the proposed method. TAL relies mainly on extending the concept of composition, usually present in hypermedia models. In TAL, compositions can be incomplete, with a few of its internal elements purposely left undefined and with restriction rules on how these missing elements may be included in the composition. In order to achieve this language, an empirical study was conducted investigating the learning and use of Nested Context Language (NCL). The choice of NCL relies on the fact that it is based in a compositional model that can be extended to specify documents families, and because it is used by various professional profiles and not just programmers in the hypermedia document authoring. Thus, a third contribution due to this work is to provide guidelines to NCL evolution, especially improving its usability.