Almost four years ago, Google claimed that its machine translator had reached the level of human translators’ work. Since then, numerous studies have been conducted on texts written by ‘bots, and many of them indicate that maybe Google was right.
But a new article questions these findings. For one thing, author Marion Marking points out, the people who were selected to read and evaluate machine translations weren’t always professional translators. Often, they were amateurs found via crowdsourcing. This means that errors in spelling, syntax, and other areas, escaped their notice.
But the article isn’t just an exposé. Markle offers five suggestions for improving machine translation. These include hiring professional translators as evaluators, and evaluating fluency.
Read on to learn more about why many machine translation studies have been flawed, as well as ways to fix flaws in machine translation.