In an increasingly streamlined and automated world, many individuals and businesses turn to machine translation in search of a quick, easy and free translation service. Convenient and easily accessible on virtually any internet-connected device, programs like Google Translate can offer users ‘on-the-go’ translations for simple queries.
However, machine translation is not without its flaws. An automated translator will generally only provide a single result, irrespective of whether the word being translated could take either a masculine or a feminine form.
At first glance this may not seem like much of an issue. However, the machine-learning algorithms responsible for educating translation software can often have the unfortunate consequence of reinforcing potentially sexist stereotypes. Also, in many types of translation – such as legal or medical translation – an error as simple as a misplaced gender can drastically alter the meaning of the text.
Machines translators, such as Google Translate at its ilk, learn from pre-existing examples of translated text. Through various complicated technologies, huge quantities of information is processed in order to provide users with translations for their queries. This often results in gendered assumptions by the software. These assumptions will inadvertently replicate the gender biases present in the primary material from which the software has been educated. This sees the machines unintentionally learning to associate certain words with specific genders, thus perpetuating outdated gender stereotypes.
For example, prior to recent changes to Google Translate, the Turkish phrase “o bir doktor” would exclusively return the translated result of “he is a doctor,” irrespective of the gender-neutrality of the original phrase. Similar assumptions were seen in queries relating to other gender-stereotyped professions such as the assumption that a “nurse” would be female.
At the end of 2018, Google introduced some changes to its AI in order to “promote fairness and reduce bias in machine learning” as part of its ‘Responsible AI Practices.’ This has involved the addition of masculine and feminine translations for a number of gender-neutral queries when translating from English into French, Italian, Portuguese or Spanish, as well as from Turkish into English.
In an increasingly gender-conscious world, applying assumptions of gender and stereotyped gender associations in translations can be problematic. Not only can they be perceived as offensive, they can also be simply inaccurate. Nor does Google’s latest effort to rid its machine translation software of gender bias do anything about the increasing recognition of non-binary gender identities.
Recent changes in the law, such as the 2018 decision in Germany to recognise a third gender on official documents, raise a number of potential translation challenges, the likes for which machine translation software is not even slightly prepared.
A number of European and world languages – German included – consist of a grammatical gender pervasive throughout the vocabulary of the entire language. This can make gender-neutral use of the language virtually impossible. For example, in German, there is no gender-neutral word for “teacher” – the only words to describe the role are gendered: “Lehrer” (male teacher) or “Lehrerin” (female teacher). Greater acceptance of non-binary identities may result in the need for gender-neutral terminology – a change that will alter grammatically gendered languages at their core.
Gender identity can be an incredibly complex, personal and sensitive issue. Attitudes towards gender identity vary widely between different countries and cultures. This makes translating material related to non-binary gender identity something that must be done with great consideration and finesse. Due to the intensely personal and nuanced nature that certain words and phrases can have, even amongst those regularly participating in the discussion, a vast understanding of the context of the material is necessary in order to ensure that the true meaning of the source material is accurately represented during the translation process. Not only this, but translation of material of this kind often involves being careful about causing offence – or even legal problems – in countries that have strong opinions on the matter. For translations of this kind, machine translation will always pale in comparison to translation by a conscientious human being.
Despite it being a controversial subject in some parts of the world, there are a number of cultures that have long recognised a concept of non-binary gender identity. However, difficulties in understanding can occur when cultural meaning is lost in translation.
The Bugis people of South Sulawesi, for example, recognise five different genders – men, women, “calalai,” “calabai” and “bissu” – and have a cultural belief of harmonious co-existence between all five genders. While the “calalai” and “calabai” are generally equivalent to trans men and trans women respectively, the word “bissu” proves more complicated to translate.
The “bissu” – an androgynous gender recognised as possessing both male and female qualities – was originally translated into English as “transvestite priest.” Through translation into a foreign language, much of the deeper cultural meaning of the original word was lost. It was replaced instead by a poor substitute that fails to accurately reflect the meaning of the source material. It was only through study of the Bugis people and their culture that the more appropriate translation of “gender transcendent” was decided.
Although a machine translator is good for the odd word or phrase, technology has a long way to go before it is sophisticated enough to deal with all the nuances of language. It should go without saying that any translation of legitimate importance should be carried out by a certified translation professional. There is more to translation than the literal translation of the words themselves. A deeper understanding of the text as a whole, as well as its context, is necessary in order to ensure that a translation accurately reflects the meaning of the original source material.
Why not leave a comment below to share your thoughts on this topic?