A Scalable and Fault Tolerant Architecture to Provide Deaf Accessibility as a Service

Abstract

Deaf people face serious difficulties to access information. The fact is that they communicate naturally through sign languages, whereas, to most of them, the spoken languages are considered only a second language. When designed, Information and Communication Technologies (ICTs) rarely take into account the barriers that deaf people face. It is common that application developers do not hire sign languages interpreters to provide an accessible version of their app/site to deaf people. Currently, there are tools for automatic translation from sign languages to spoken languages, but, unfortunately, they are not available to third parties. To reduce these problems, it would be interesting if any automatic translation tool/service could be publicly available. This is the main goal of this work: use a preconceived machine translation from Portuguese Language to Brazilian Sign Language (LIBRAS) (named VLIBRAS) and provide Deaf Accessibility as a Service (DAaaS) publicly. The idea is to abstract inherent problems in the translation process between the Portuguese Language and LIBRAS by providing a service that performs the automatic translation of multimedia content to LIBRAS. VLIBRAS was primarily deployed as a centralized system, and this conventional architecture has some disadvantages when compared to distributed architectures. In this paper we propose two distributed architectures in order to provide a scalable service and achieve fault tolerance. For conception and availability of this service, it will be used the cloud computing paradigm to incorporate the following additional benefits: transparency, high availability, and efficient use of resources.