Assessing the quality of English-to-Thai machine interpreting : translating Google Translate

Abstract

This research aims to assess the quality and potential of Google Translate when used as an English-to-Thai Machine Interpreting (MI) tool. The paper explores MI’s distinct definitions and terminology as well as its relationship to other relevant technologies, i.e. Machine Translation (MT), Automatic Speech Recognition (ASR), and Text-to-Speech (TTS). Four five-minute English-language recordings from the European Commission's Directorate General for Interpretation Speech Repository were selected as samples for the MI to interpret. Human evaluation was used to evaluate the Comprehensibility, Accuracy, and Usability of MI with five final-year interpreting students and six professional interpreters performing as judges. Results show that MI is still far from perfect with problems in every step of the MI process from ASR, MT, to TTS. Discussion and analysis further reveals how and why current MI fails in its processing of English into Thai, as well as presents contemplations on what should be done to enhance MI performance. The subtle decisions chosen, and mistakes made, by the MI tool when encountering uncertainty suggests that MI exhibit comparable error symptoms and display similar coping strategies to human interpreters. Thus, it is paramount that professional interpreters be involved the development and improvement of MI technology going forward.