Space, metaphor and schematization in sign: sign language translation in the ZARDOZ system

Space, Metaphor and Schematization in Sign: Sign Language Translation in the ZARDOZ System Tony Vealel, Bróna Collins2 1 School of Computer Applications, Dublin City University, Glasnevin, Ireland. 2 Dept. Of Computer Science, TrinityCollege, Dublin 2, Ireland Email: tonyv@compapp.dcu.ie Abstract: The sign languages used by deaf communities around the world represent a linguistic challenge that natural language researchers in A.I. have only recently begun to take up. ZARDOZ is a system which tackles the cross-modal machine-translation problem, translating speech and text into fluid sign language. This paper presents an architectural overview of Zardoz, describing its central blackboard organization and the nature of its Interlingual representation, while discussing the deep conceptual issues raised by such a system, in particular, the role of schematization or concept chunking in translation, the coherent allocation of spatial indices in the communication of explicitly-realised spatial descriptions, and the creative contributions to the translation process that can be made by an underlying theory of metaphor. 1. Introduction Recent years have seen the acceptance of sign-language by the linguistics community as a fully- featured, first-class natural language, one that exhibits the full range of traditional linguistic phenomena, as well as a host of expressive powers unique to gestural communication. This paper describes the architecture and methodology of the ZARDOZ multilingual sign translation system, designed to translate spoken language (ostensibly English) into a variety of sign-language variants, in particular ASL (American), ISL (Irish) and JSL (Japanese). This goal of fluid articulation of sign language gestures from English language input embodies the unique linguistic challenge of cross- modal translation, possessing significant social, commercial and theoretical implications. For there exists a sizeable body of sign language users world-wide, for which such technology will provide valuable educational tools: the technology will not replace, but empower and educate new sign interpreters. Indeed, contrary to the perceived A.I. goal of humanizing machinery, sign translation systems are tailor-made for social situations where an obviously non-human translator is required, for a machine will not violate the doctor/patient and lawyer/client confidentiality expected by a signer. From a linguistic and cognitive perspective, the pursuit of cross-modal translation further challenges our preconceptions about what constitute language universals. And from a pragmatic A.I. perspective, sign-language translation is a unifying goal which provides an ideal opportunity for the synthesis of existing A.I. theories and techniques into a workable and socially-relevant application in the short term. As there exists an ever-growing body of research concerning the structural properties of sign-languages, for instance the treatment of ASL due to Liddell (1980), this paper complements this work in discussing the purely AI considerations of sign communication. Given these pragmatic and theoretical goals, the rest of this paper assumes the following structure: section two introduces the sign language medium, which serves to place the contents of the paper in some focus. Section three then presents an overview of the system architecture of ZARDOZ, which is conceived and implemented around the blackboard control metaphor. Section four discusses the issue of schematization in sign-languages, that is, the mechanism …

Content truncated to 500 words