This article describes how technologies needed to fully automate inter- pretation work. It assesses the performance of existing technologies and compa- res them to the work of human conference interpreters. The article goes on to evaluate the problems to be encountered at every step of a spoken language translation system: speech recognition, machine translation and speech synthe- sis. The article closes with an overview of existing products and tries to define areas where current technology can improve interlingual communication. This article argues that a 'universal translator' as depicted in science fiction will be hard to implement with statistical machine translation at its core
© 2001-2025 Fundación Dialnet · Todos los derechos reservados