A Graph Retrieval Architecture and System for Online Sign Language Dictionary- with an application to Taiwanese Sign Language

Abstract

Emerging applications for an online sign language dictionary require that retrieval systems retrieve a target vocabulary through visual symbols. However, when people encounter an unknown vocabulary in sign language during communication, they require the online dictionary to retrieve the vocabulary with higher recall-rate and smaller-sized graph through a mobile device. Still, three situations show that the current online dictionary needs an extension. First, previous works lack of retrieving the target graph of a vocabulary through its complete visual symbol-portfolio. Secondly, they often respond a large number of possible images; however, their precisions and recall rates remain very low. Thirdly, previous works of sign language gloves can convert the visual symbols into the graphic features, but only part of the symbols, ignoring the symbols of expression and relative direction. Therefore, the aim of this study is, based on Taiwanese Sign Language, to design a new graph retrieval architecture for sign-language (GRAS), and to implement a new graph retrieval system for sign-language (GRSS) based on this architecture. Finally, we invite users to evaluate GRSS. The experimental results show that GRSS gets convincing performance. And, GRSS adopting RDF technology can improve the performance of GRSS without adopting RDF technology.