NMT/ Neural Machine Translation vs. Human Translation Services- Know the Real Facts

Logo Authur
Dr. Dipankar Kundu
Dar Al Marjaan Translation Services
NMT/ Neural Machine Translation vs. Human Translation Services- Know the Real Facts

NMT (Neural Machine Translation) may be defined as AI-powered software that works as a language translator viz. Google AI Translation System, Quartz, ZD Net and many others. The subject of discussion is – ‘ Whether NMT can overpower and supersede Human Translation Industry or not? To summarize NMTs can definitely help Human Translators but can never supersede them. Present-day NMTs are not as quite as good as they have been purported to be. Let’s imagine that NMTs will become completely developed with time.

Even in such a situation in the future will it be able to handle translation the way Human translators do handle? If not, then what are the reasons? Is whatever said today of the prospect of Neural Machine Translators & Artificial Intelligence is a myth?

Here in this article, a sincere effort has been made to enumerate the reasons behind NMTs not being able to over-rule Human translations even at any point in time in the future. Lets us proceed towards each of these reasons one by one below:

  1. Neural Machine Translations are based upon neural networks. The neural network is a system that can be trained to recognize patterns in data thereby transforming some input data in one language into the desired output in another language. Before any machine translation can happen human engineers need to decide the architecture of the network. This requires significant expertise, a lot of experience and a fair amount of guess. These are the root causes for the failure of NMTs to overpower Human Translation services. Human engineers are technology masters but are not masters in languages and translation therefore the patterns that they produce are not devoid of faults. Also, the guessing of input data to output data is based upon guess rather than facts and context and hence leads to erroneous translations. This is why human translation services will overpower translations rendered by Neural Machines always and ever.
  2. Generally in every country and particularly in the United Arab Emirates, nearly 80% of translation assignments are done for legal conformity and to meet legal formalities, be it for submission in courts, ministries, embassies or consulates. These translations are known as legal translation, certified translation, sworn-in-translation. Human translators approved by and holding a license from the Ministry of Justice of the respective country are the only authorized persons to do such translations, without their written oath and certification these translations are not worthy of use and are not acceptable by the judiciary. A machine cannot be licensed or approved from Ministry of Justice, A machine cannot take an oath in front of law and even if the technology makes it possible law will never record the oath and witness of a machine as a machine can never be punished by law, a machine has no emotions and feelings. Thus, NMT s’ role in LEGAL translation can be completely ruled out and in all situations and it is confirmed that only human translation services are needed for LEGAL translation and not machine translation.
  3. Another wide field of human translation services is the translation of old and ancient manuscripts. The study of old manuscripts has made available to humankind a lot of beneficial information about the past be it in the field of science, technology, history, medicines, civilization, architecture and others. Since all the data collected by AI-powered software is from the web, Neural Machines cannot understand these documents. NMT will completely fail to translate old manuscripts due to a) it being handwritten documents, b) its primitive script, c) unfamiliar grammar, syntax, semantics & vocabulary d) not enough data available on the web to support machine translation of such a language and text of such nature. You may read an article Portuguese to English Translation of Ancient Manuscripts – A Guide for All Stakeholders to understand it better.
  4. Another important field is the translation of literature like poems, novels, short stories, books & magazines. In writing of literature the language used is not the language of daily life. An author often chooses words and phrases that fit exactly to a particular situation; these words and phrases are generally uncommon and very rarely found to have been used by others thus not available on the web as precedence. In addition to this, literature is an ornamental form of writing characterized by extensive use of idioms, phrases & figures of speech. The literal meaning of an idiom is definite when it is discrete but gives different connotations depending on its contextual use. Whereas, a figure of speech needs to be interpreted according to the storyline and the context. NMTs will never have the ability to understand the context and thus will fail in translating an idiom or a particular figure of speech correctly inside a literary content. Moreover, the authors frequently innovate and use new metaphors and similes. NMTs will never be able to recognize these data due to a lack of precedence and matching. Thus, in the field of translation of literary work we can say with confidence that machine translation will not be able to do justice to the job as it will be given by human translation services.
  5. Every day there are new inventions and discoveries in the field of science. Scientists work incessantly to find a new information and their literature is generally documented in their native languages. In the field of translation of new scientific data, NMTs will always fail, as NMT is an AI-powered software and can only identify and translate data that has been fed in. No engineer can apprehend the works of scientists in advance and hence, translation of scientific literature with new data and inventions will fall outside the ambit of NMT. In other words, machine translation will never be trusted as human translation services are trusted when it comes to translation in the field of science.
  6. The translations made by NMT will be devoid of all the major aspects of a human intellect that reflects in his writings for example feelings, emotions, perceptions, emphasis and human ideas. It also will fail to make value judgments. This makes machine translation weaker and human translation services more acceptable.
  7. NMTs lack the authority that a human translator possesses. Machine Translations by NMTs will look like a collage. This is because NMT systems are characterized by acute short-term memory loss. The translations by NMTs are done one by one sentences and they forget information gained from prior experiences. Thus, it gives an impression that different translator without co-ordination with each other has translated each sentence of the document. This leads to loss of contextual importance of the document full of flaws; lack of continuity and in most cases makes it utterly offensive and unintelligible. This is where human translation services will be an ever thriving industry.
  8. In each & every language, there are intricacies that cannot be tackled by NMTs. For eg. in Arabic, we have Ahrab (in English known as Declination) and three cases viz. Accusative Case, Nominative Case & Genitive Case. Arabic is not written fully voweled and thus a human who knows grammar can read a word with proper declination understanding the case. Correct reading gives the correct meaning. Also, in Arabic morphology, a word though looks the same in two different cases but needs to be read with different declinations according to applicable cases and this applies to almost each and every word in Arabic. In addition, there are verbs, which can be read in more than 2 ways, and in each way the meaning is different. Similarly, every language will have its own complex elements of reading, writing, and understanding. Machine Translation is therefore not much promising in comparison to human translation services.
  9. In every language there are dialects, for eg. Arabic has more than 30 dialects. In French, there are 28 different dialects. Similarly, in every language, there are a number of dialects. While making a translation a translator needs to know the target readers of the translated text and must have knowledge about their culture. Machine translation cannot achieve this objective. Thus, while translating from one language to other it will not consider the dialect and the culture of the target audience. Thus, in such a scenario it will fail to serve the purpose of the translation whereas human translation services meet this very fundamental objective. In other words, it fails and will fail miserably in Localization. To know about website localization you may read the article English vs. Local Language| Why You Need to Localize and Translate Your Website in Multiple Languages.
  10. The picturesqueness of creative writings cannot be achieved by the use of machine translation. The capacity to give picturesqueness and translate creative writings can only be handled by human translation services. Picturesqueness is a very important component of creativity and sometimes in writings related to business and commerce as well.
  11. Translation needs common sense in many instances; Human translation services are an industry that needs to be recognized as superior to machine translation when it comes to using common sense and intellect.


The above points are just a few and not exhaustive in nature. NMTs can be best used for translating the language of daily life communication between two individuals for the purpose of normal communication but it can never be elevated to a level where it can take the place of written translation rendered by human translators.

Get a Quote
  • Name*full name
  • Email*
  • Phone number*
  • Your company*
  • Your message here...*something more
  • 5