Feature-based natural language processing for GSL synthesis

The work reported in this study is based on research that has been carried out while developing a sign synthesis system for Greek Sign Language (GSL): theoretical linguistic analysis as well as lexicon and grammar resources derived from this analysis. We focus on the organisation of linguistic knowledge that initiates the multi-functional processing required to achieve sign generation performed by a virtual signer. In this context, structure rules and lexical coding support sign synthesis of GSL utterances, by exploitation of avatar technologies for the representation of the linguistic message. Sign generation involves two subsystems: a Greek-to-GSL conversion subsystem and a sign performance subsystem. The conversion subsystem matches input strings of written Greek-to-GSL structure patterns, exploiting Natural Language Processing (NLP) mechanisms. The sign performance subsystem uses parsed output of GSL structure patterns, enriched with sign-specific information, to activate a virtual signer for the performance of properly coded linguistic messages. Both the conversion and the synthesis procedure are based on adequately constructed electronic linguistic resources. Applicability of sign synthesis is demonstrated with the example of a Web-based prototype environment for GSL grammar teaching.