Text-linguistic evaluation of Twitter's auto-translation service: Donald Trump’s Tweets during 2020 USA elections

Abstract

Over the years, technology has revolutionized our daily lives and played an essential role in developing all work fields, including translation. However, some argue that depending on machine translation (MT) completely would produce a poor-quality translation. To illustrate, social media is a way to communicate across cultures, yet allowing users to access as much content as possible requires translation services. According to Twitter’s 20 Earnings Release, Twitter users in the 2020 1st quarter reached 166 million users worldwide, but not all of them speak the same language. Therefore, offering translations of tweets into as many languages as possible was necessary to assure a wide range of communication across different linguistic boundaries. In collaboration with a well-known MT software, namely, “Google Translate,” Twitter has launched an auto-translation service that translates all tweets immediately into the default language of the user’s account. Although on the face of it, MT saves time and effort, its outcome does not live up to expectations and may indeed hinder rather than help the process of communication. Therefore, research aiming to develop MT performance and outcome is essential to keep up with the ever-growing need for successful and timely communication nowadays. With this in mind, this research aims to evaluate the auto-translation (MT) of Donald Trump’s tweets during the US presidential elections in 2020, using text linguistic analysis model (Neubert and Shreve 1992), identifying and analyzing issues in the MT output, and suggesting alternative solutions to those issues in order to help produce better quality translations of these tweets. This textual product-oriented study finds out that although MT produces understandable translation, generally speaking, it is not ready yet to be relayed upon completely in translating Twitter’s political tweets, because of the recurring violations of the seven standards of textuality.