In this paper, we present a system that performs sign language synthesis in the framework of an educational platform for young deaf children. The proposed architecture is based on standardized virtual character animation concepts for the synthesis of sign sequences and lexicon-grammatical processing of Greek sign language (GSL) sequences. A major advantage of the proposed architecture is that it goes beyond the usual single-word approach which is linguistically incorrect, to provide tools to dynamically construct new sign representations from similar ones. Words and phrases are being processed and the resulting notation subset of a lexical database, HamNoSys (Hamburg Notation System), eventually transformed into GSL and animated on the clients’ side via an h-anim compliant avatar.