Semantic Translation

From Fmepedia

Traditionally spatial data translation was carried out by an application written for a specific purpose. For example, it would translate only from one format to another, it would work only on a limited set of geometries, and it would be assumed that the destination schema would match the source data.

Semantic Translation is the concept where a central processing engine and rich data model are provided to permit the user to expand on traditional methods. The rich (format neutral) data model allows any geometries to be used, whilst a centralised processing engine allows any number of formats to be applied, with schema mapping defining in which way the source data is mapped to the destination schema.

In other words, for the most part FME will automagically (http://en.wiktionary.org/wiki/automagical) translate data from one format to another with the minimum of user intervention - all of the basic rules for the translation (eg that Shape format requires upper case attribute names) is held by FME. Only where a complex destination schema requires careful mapping to the source data (eg when writing GDF) does FME require the user to be actively involved.

See this white paper (http://www.safe.com/solutions/whitepapers/semantic_translation.htm) for further information on the concept of Semantic Translation.