This paper reports on research and development work for the implementation of a Greek to Greek Sign Language (GSL) conversion tool, integrated into an educational platform that addresses needs of teaching GSL grammar, exploiting avatar technologies for the representation of linguistic message articulated in the 3D space. The Greek to GSL conversion engine that supports the platform is composed by two subsystems: one that matches input strings of written Greek with GSL structures, exploiting Natural Language Processing (NLP) mechanisms, which expand from tagging and lemmatizing to parsing for syntactic chunk formation and Machine Translation (MT), and a sign performance subsystem that uses NLP output to activate a virtual signer. The whole procedure is based upon adequately constructed linguistic resources of GSL. Here we focus on design and implementation issues of the structure mapping module that except from the coded NL knowledge used, is also described as to methodological and application development issues related to the implemented prototype.