Machine Translation Technology – It’s just not that simple

Jan 19, 2022 | Language

A Changing World – but should it be?

We live in a world where, in the modern day, it makes so much sense for many companies to use machines rather than human workers for day-to-day tasks.  This can range from large scale manufacturing to ordering screens in fast food restaurants. The objective for businesses seems clear – the use of machines and computer software can speed up the process while cutting the cost of employing large numbers of workers.  However, anyone who has ever sheepishly had to wait in a supermarket for an employee to enter a passcode into a self-checkout till will know that it is never really that simple.

The same outcome is inevitable in the language translation industry.  More and more, we are finding that some businesses are understandably sold with the promise that machine translation technology will save them time and money by cutting out a third party.

However, are these businesses able to get the same quality from machine translation output as they would with a professional, human linguist?

Increasingly, we are discovering this to be a problem for so many of our clients – which is why they continue to come back to us for high quality, timely translations.  So, why is this?

Honest Mistakes

One of our medical clients asked us to translate an IFU for a medical device, with the language pairs of English into Spanish and French.  They had trouble using machine translation – specifically Google Translate.  When attempting to translate the word “foot”, as in a table or chair foot, the machine translated this as a human foot – an error only a human would be able to notice and was not picked up by the translation software.

Another regular example we have found in the medical industry is that a medicine tablet is no longer the most popular type of tablet.  Machine translation software regularly translates this as a device, like an iPad, as opposed to the more traditional medical tablet.

Very recently, we noticed an error when translating a series of email conversations with the language pair of German and French. When attempting to sign off with “best wishes” with French as the target language, the person had used machine translation; thus literally offering best wishes. Only human translation could help avoid this mistake and translate the sign off correctly as “cordialement”.

These examples strongly convey how easy it is to miss small mistakes caused by machine translation systems.  A mistranslation can completely change the meaning of a sentence and often an entire text or passage.  Only the trained eye of a native target language translator can provide the correct linguistic and cultural insight to guarantee an excellent quality of translation.

Businesses often believe that the use of machine translation will save them time.  However, it is clear from these examples that the opposite is so often true as these small, hidden mistakes take considerable time to find and rectify. Machine translation quality remains flawed and significantly inferior to any professional human linguist.

The rapid advances in machine translation only highlight the need for human professionals

There can be no doubt that machine translation has advanced quite considerably in recent years. The evolution from statistical machine translation to neural machine translation NMT has unquestionably improved user experience and allowed businesses to perform basic, understandable translations.

However, this intense development has still fallen evidently short of human translation as, quite simply, patterns and statistics will only take you so far. Issues such as long sentences continue to cause issues for users. As well as this, such an advanced attempt at performing natural human duties requires significant system and software training.

This level of machine translation requires a large amount of time and money and the end product is still likely to have major inconsistencies and flaws.

Google Translate has become a verb

Whether you work regularly with different languages or need a quick solution to an international communication issue, Google Translate is so often the go-to machine translation tool for just about everyone. The appeal is completely understandable: an automatic translation available in a matter of seconds. However, there is a reason it is not the world’s exclusive resource in language conversion.

Google translate, along with all types of automated translation tools, is full of inconsistencies and limitations. For example, how often have you put a sentence in one language through a machine translation system and then put the translated version back through? The final version is often a laughable attempt that would never be deemed acceptable in a professional capacity. Machine translation work often results in a risky game of Chinese Whispers where the end result is a non-sensical mess in the target language.

For businesses, the idea that machine translation technology can produce a 20,000 word project in a second for free where it would take a human linguist over a week naturally seems enticing. However, any business with experience in using this type of resource for good quality translations will tell you that they now only use the help of human translators. The level of quality provided by a human being cannot be replicated by machine translation systems. Our linguists use their specific expertise to ensure an accurate translation with every project.

Idioms – fun but dangerous

Languages are wonderful, unique aspects of life. The multiple phrases and sayings that are exclusive to all different languages should be enjoyed and celebrated.

However, there is a special level of care that must be taken when translating idioms which, in the human translation vs machine translation debate, advantages the former.

An unavoidable and common downfall of machine translation output is how literal the translation appears in the target languages. Despite the advances of machine translation capabilities, there are niche aspects of certain languages that are virtually impossible for language translation technology to pick up on its own.

For example, the Czech way to say “to beat around the bush” is “chodit kolem horké kaši”. However, machine translation would translate this as “Walk around in hot porridge”, (see here) which would leave anyone reading in the target language scratching their head.

In French, “avoir le cafard” literally translates as “to have the cockroach”. However, the intended meaning is “to feel down in the dumps”.

The Thai phrase “ชาติหน้าตอนบ่าย ๆ” would be considered an equivalent to the English idiom “it’ll be a cold day in hell”. Machine translation would produce this as “one afternoon in your next reincarnation”.

These are huge aspects to consider when using machine translation. When translating large texts, it is extremely easy to miss these errors. A mistranslated idiom is a sure sign of a translation that has not been cared for and appears unprofessional and often embarrassing. Idioms can add a colourful, even fun aspect to a text and, really, they require the hands-on expertise of a human to provide better translations than a machine. Machine translation quality simply is not at a level to produce such advanced level of language.

Gender Bias in Machine Translation

Another major disadvantage of using machine translation is that it generally demonstrates blatant gender bias. The grammatical rules of many other languages differ from English in that nouns are either masculine or feminine.

For example, in a translation with a language pair of English and French, it is crucial to put the correct gender for even simple things like the names of countries.

However, the major issue is that machine translation almost always assumes that the subject in question is male if it is not made abundantly clear otherwise.

This problem is caused by an error in the algorithm that forces incorrect assumptions. Not only is this an irritating problem to fix, but could easily cause offence as it favours one gender and could appear to ignore another; hence another inevitable issue with the output quality of machine translation.

Taken out of context

Context is a huge aspect of translation which only a human linguist can understand and perfect. Machine translation technology will always translate sentence by sentence meaning an entire paragraph could have multiple different meanings. Even rule based machine translation output will be dictated exclusively by each different sentence and cannot truly relate to the text translation around it.

Human translators naturally understand the importance of bringing a text together. When assigning projects to our linguists, we look for those who have experience in the relevant fields and ensure that they do their research before starting to make sure they are able to provide a professional translation that makes sense for the target language reader.

Cultural Issues

The issues with the exclusive use of machine translation does not stop with incorrect translations of words.  There are inevitable cultural limitations that a machine can have when translating texts.  A machine could, for example, translate every word correctly and be understandable to the target reader.  However, businesses who have tried machine translation have often found their target texts to be clunky and appear poorly written as the machine translation naturally copies the grammar and punctuation of the source.

For example, it is well known in translation that Spanish sentences are often long with in-depth descriptions and lists taking up to 4-5 lines.  In English, these would, if translated correctly, be split into 3-4 different sentences.  However, there is no way for machine translation software to know this and the results that it produces often come across as unprofessional.

Another example of machine translation proving unable to provide the cultural touch of a human linguist is when translating places, names and, crucially in this case, specific groups.  In a recent French into English translation project, we noticed that while France uses the term “LGBTI” as the name for the well-known inclusivity group, the name in English in fact extends to “LGBTQI+”.  Without the correct cultural awareness and research from a human translator, an entire community could have been left out of the text.  Again, this is remarkably easy to miss and there is no way machine translation could be expected to translate this properly without a human touch.

Many language pairs cause similar cultural mistranslations on machine translation services. Even the most state of the art artificial intelligence will lack the cultural input of a human. Machine learning in translation is limited to very basic grammatical adaptations but tends to remain literal and culture blind. Existing human translations prove that natural language processing and translation from a person is hugely superior when dealing with cultural text translation.

Machine translation does not work out as cheap as you think

As alluded to, a major draw for businesses to use machine translation technology is that it appears to be a cheap alternative to using human translation. If the idea is to have a messy, unverified, literal mirroring of the source language, then technically this is true.

However, a significant point to consider is that if you wish to have your machine translation project checked by a human translator, they will charge more than they would to check a human translation. The reason for this is that translators know that checking machine translations will take more time and effort because it will inevitably be full of grammatical and structural errors. Depending on the number of errors and how understandable the translation is, this method can often completely backfire meaning a whole new translation is necessary.

The only way to guarantee high quality translation is with a professional, human linguist and proof-reader who are native in the target language and fully proficient in the source language.

It’s a Team Game

At My Language Connection, we make sure to leave no stone unturned with our translation services to guarantee that all linguistic and cultural mistakes or potential hiccups are rectified; thus able to provide the highest quality of translation. Our linguists are all certified with years of experience in their offered services and we only ask them to translate and proofread into their native language.  Together with these experts, we use CAT Tools to ensure maximum linguistic and cultural quality as well as consistency in every project using Translation Memories and glossaries.  From the beginning until the end of the translation process, we will help you with every query and specification you have.

We use every resource available to guarantee customer satisfaction from research to actual translation and delivery.

Our hands-on, friendly and timely service that machine translation cannot provide is the reason that our stock continues to rise and our clients continue to return to us.

Contact us today for the best translation results around by calling 0330 058 0251 or by emailing info@mylanguageconnection.com.

Get a Quote