Measuring the Impact of Automatic Speech Recognition on Number Rendition in Simultaneous Interpreting
Elisabetta Pisani, Claudio Fantinuoli
Abstract
Automatic speech recognition (ASR) has been proposed as a tool to augment interpreters’ performance in simultaneous interpretation. In particular, ASR could offer support during the translation of several problem triggers identified in literature, such as numbers, terminology, and named entities. Because of the novelty of this technology, little is known about its real potential in the booth. The effect on the interpreters’ performance of manual information lookup with computer-assisted interpreting tools and of ASR simulated systems recently have been in the focus of empirical research. The present experimental study adds to this body of evidence by analysing the impact of a real-life ASR tool on the simultaneous interpretation of numbers. The results show a significant reduction of error rate, which drops from 39.8% without any support to 14.8% with the use of the ASR support.