Career crisis for translators in the AI era

World
Europe

The growing use of Artificial Intelligence (AI) has had a dramatic impact on several industries, and professional translation is no exception.

Companies from Baidu and NetEase to Google and Facebook are constantly updating their online translation functions to be more user-friendly and, more importantly, more accurate.

"language" in global languages /Photo credit: By M. Adiputra

AI translation in development

According to Duan Xin, a Chinese software engineer working at Microsoft, traditional AI translation both in China and in other countries was limited to the phrase-based model in 2010 – researchers would collect words and phrases and the program would then arrange them into paragraphs and then texts.

Since 2016, Google AI translation has updated its phrase-based model to neural machine translation (NMT). This year, Facebook launched its automated translation system, which is designed to "take into account the entire context of the source sentence and everything generated so far, to create more accurate and fluent translations," wrote Facebook's engineers Juan Miguel Pino, Alexander Sidorov and Necip Fazil Ayan on the company's developers blog.

The aim is for the AI software to imitate the operation of the human brain when it learns a new language. The updated model provides more accurate translation, especially in the fields of law and finance.

A comparison of translation models based on quality /Photo via Google

AI translation challenges

But despite training, testing and constant analysis, one problem remains unresolved when it comes to AI doing translation work – low accuracy.

First, the phrase-based model does not work well when the translation is done between languages in which the order of words varies significantly because the technique relies on breaking down the sentences into parts.

Also, in a narrow view focusing on words themselves, a one-to-multiple relationship based on polysemy in linguistics often exists in the mutual translation, which makes AI difficult to target the most precise meaning in the given context.

Take the following Chinese proverb as an example. In "女子无才便是德", the Chinese character “德” has several meanings. In the proverb, it means virtue or merit but Google Translate mixed it up with its other meaning: Germany.

Screenshot of a Chinese proverb's translation via Google

Screenshot of a Chinese proverb's translation via Baidu

Besides low accuracy, the data pool, which the AI relies on to perform translation work, needs to be updated and expanded. The dataset at present, according to Duan, comes from a collection by humans and AI is learning through its logic.

User feedback and AI training are needed, which is time-consuming. The changing nature of languages, with new words and phrases introduced on a regular basis while others discarded, doesn't make the task any easier.

Undoubtedly, the development of AI translation technology has led to a career crisis among some translators and interpreters. However, many believe that the technology is a tool for humans rather than a replacement of human thinking.

'Humans should not be replaced by tech'

Sally Zheng, a freelance translator and interpreter working in Shenzhen and Hong Kong, usually tackles translations of law contracts, subtitles for trailers and company promo videos. She also occasionally handles literature released by Chinese publishers.

She uses computer-assisted translation (CAT) software such as MemoQ and Trados to handle a volume of documents which sees many terms repeated throughout.

"Computer-aided translation will impact the translation a lot, especially in the fields of daily talk, finance and law with internal rules involved,” she said.

Asked if she ever fears being replaced by a machine, she answers with an emphatic "no", saying that translation is a process of re-creation based on an understanding of the source language.

"I’m not worried about losing my job at all. Translation is not just about the languages, but deals with the culture and thoughts behind it," she explained.

"It’s a process of cultural exchange and collision. The professionals and non-professionals will provide entirely different translations. I think in future there will be more translators.”

"An author lives in my heart. I would love to do literary translation if it is not just low-paid work in China. Literature offers translators more freedom in dealing with words and we can be more subjective.”

A man writes the word "Peace" in different languages at a makeshift memorial at Place de la Bourse following attacks in Brussels on March 22, 2016. /AFP Photo

The invention of 'transcreation'

“Transcreation" is also a growing trend, especially in advertising, which sees a message adapted from one language to another, while aiming to maintain its original meaning, style and tone.

According to Lionbridge, the world's largest provider of localization solutions, transcreation aims to effectively preserve the message's creative and emotional intent while making it resonate in other languages and cultures.

The process depends very much on a human understanding of the nuances and vagaries of a language and is a process that AI currently can't tackle.

To conclude, with the thriving tech development currently, AI translation indeed can replace some translators and interpreters with low level. However, languages and words wane and wax. After all, they're as dynamic as human beings.