Machine Translation for Languages of Afghanistan
Machine translation is a new phenomenon in the content localization landscape and it makes sense to discuss this topic in the context of languages of Afghanistan. In a previous post, four years ago we discussed what machine translation (MT) and how it fits with Dari language translation. Today we will discuss the latest developments in the field of MT and what it means for Afghanistan.
Machine Translation – A Quick Intro
From the beginning of time, humans have sought to use technology to automate things. We have used the flow of water to crush wheat into flours in our mills. We have used wind or ocean to generate hydro energy. And now, we have managed to leverage the power and intelligence of computers to translate text from one language to another.
While the concept of machine translation (MT) has been around for several decades in a variety of forms, it has only recently gained traction with the advent in artificial intelligence. Google, Amazon, Facebook, Microsoft and a number of other innovators are heavily invested in development of artificial intelligence and machine translation has been a natural application of this technology.
What Does Machine Translation Mean for Afghanistan
Afghanistan is a country with two official languages, Dari and Pashto, with a myriad of other languages that are spoken across this country. Numerous dialects and sub-dialects of each language are in use in villages and urban areas. Additionally, language has always been a contentious issue among a few groups. Communication is normally delivered in both official languages, but Dari typically gets associated with the language of spoken by people in the central government. Pashto is equally important and large swaths of population speak a multitude of dialects across several provinces.
Given the amount of content generated by Afghans for local consumption, translation from Dari into Pashto, and vice-versa, has been a logistical nightmare. Throw in English into the mix, given Afghanistan’s prominence on the world stage, things get even more complicated. The end result is a country with a large number of citizens who may be deprived of critical knowledge due to linguistic barriers.
Automating low priority communication through the help of machine translation can boost the delivery of services. As more and more Afghans join the global internet citizenship, it only makes sense to deliver important and relevant information to them as fast as possible in a manner that is understandable and of interest to them. Machine translation can help make that possible.
Machine Translation vs. Human Translator
Despite the significant strides that the global scientific community has made in the past century, we are still way off from achieving intelligence that comes with intuition and consciousness. An Afghan translator translating a page of legal text from English into Pashto will still deliver a far more interesting and accurate text than any machine translation engine in existence today. However, we are still limited to the speed at which the translator thinks and types. A machine translation engine can be useful to translate large volumes of less critical texts, such as facebook posts or emails, into Pashto in a matter of seconds or minutes.
A machine translation engine, just like a human translator, will need to be trained in a specific domain or subject matter. In order to develop an engine for automotive manuals, large volumes of translated data from car manufacturers, parts makers, technical product manuals etc. will be needed to feed through the engine as it will develop its engine. Additional technical intervention may be required by system administrators to provide specific instructions that may be unique to Dari or Pashto languages.
Over a period of six months to one year of continuous use of the machine translation engine, it will be consistently tweaked and its translation errors corrected by Afghan language experts to fine tune the final product. It is fair to say that by the end of the first year the machine translation engine will be in a position to delivery 80 to 90% of human quality translation. An editor will always be required to sift through the produced translation for correcting errors.
Do You Really Require Machine Translation for Your Project?
That depends on your particular business case. If you work with critical patient information in a hospital in Mazar e Sharif, regardless of the volume of data that you process, you are better off to stick with human translation service. However, if you are handling massive amount of financial information for your local investment website, we recommend developing a machine translation engine that would turn the English data from New York Stock Exchange into readily usable and meaningful information for someone sitting in Wazir Akbar Khan.
Do you gain anything by switching your Dari or Pashto translators with a machine translation engine? Yes, you do. You gain speed and capacity and cost savings. But you will not gain any quality improvements, in fact you will lose quality if you are not deploying a human validator (editor) in the process. Your translation costs will go down, but you have other costs such as developing and maintaining a translation engine with upfront investment. Depending on your situation, these costs could be significant but still lower than a manual process given your gains with speed and capacity.
As discussed, Afghanistan could use machine translation to ensure the population receive state or foreign information in their desired language. The Government of the Islamic Republic of Afghanistan could take advantage of a fully trained machine translation engine to translate legal, financial, legislative and state related content into both official languages as well as all other languages spoken in the country. The private sector could take advantage of such as system to save costs and deliver their products and services in local languages, improving their bottom line and efficiency.
If you have specific questions about using machine translation for Dari or Pashto, or in general, we would be happy to address them and connect you with our technical experts.