About 30 years ago Donald Knuth proposed the the term of "literate programming" ( http://www.literateprogramming.com/ ): Knuth suggested to extract programming code from it's own documentation. The idea was to priorize documenting a program's functions over coding the functions itself.
The reason was simple: You cannot really understand a program without having it's documentation available, for programming code is to formal, to abstract, not redundant enough to work properly as a human communication language. If software engineer wants to produce high quality results it is not surprisingly herself who benefits most from writing the documentation properly. Same is true for data modelling, i.e. ontology engineering. Experience shows that a data model (often outlined UML) or an ontology (typically written in OWL, RDFS or F-Logic) much better can be understood if it is appropriately documented.
semAuth revives Donald Knuth's idea of literacy. The idea is to structure and to tag prose text appropriately such that we can extract a data model or a semantic representation (i.e. an ontology) automatically on a well defined and easily comprehensible basis. Today, in the world of the semantic web, we propose the term "semantic authoring".