They following is an essay I wrote for a class while on exchange in Montreal.
Google translate is a free to use translation service that currently supports 103 languages (Languages, 2016). It is available online, offline, on mobile devices, and is embedded into the Google Chrome internet browser (About Google Translate, 2016). It is of common opinion that Google Translate has transformed many aspects of everyday life, from tourism, to education, to even medicine. In this essay, it will be argued that google translate utilises a myriad of technologies in a way that would not have been possible until very recently but it carries with it a number of social, ethical and political concerns.
Google Translate launched in April 2006, but in order to understand the historical context of its launch we must look back to the 1950’s. Starting around 1950, researchers began to develop a computer system to translate one language into another. In order for the computer to do this, researchers imputed grammar and rules of language. Unfortunately, this system did not work, as human language proved to be too complex for the computer systems to understand (Nielsen, 2011). Towards the 1990’s, researchers tried an entirely new approach for translation; they stopped using rules of grammar and language and instead fed into the computer enormous amounts of already translated documents for the computer to analyse. This method involved the computer analysing the documents for patterns and reoccurrences and then working out the probability of two words having the same meaning. The system eventually worked so well it could even understand when words changed their place in a sentence between languages (Nielsen, 2011). This method is called statistical machine translation, and is what Google Translate used when it was first launched.
Google’s statistical machine translation system was extremely effective, with the company winning every category of a translation evaluation in 2005 (NIST). This system served as Google’s main program for translation until as recently as November 2016, when Google announced they would be switching to neural machine translation (Turovsky, 2016). Currently functioning with only eight language pairs, google hopes to eventually implement the system into all 103 of their supported languages. Neural machine translation, works by translating whole sentences at a time “which it then rearranges and adjusts to be more like a human speaking with proper grammar” (Turovsky, 2016). This new machine also “learns over time to create better, more natural translations” (Turovsky, 2016). In the words of Google, “Google Translate is improving more in a single leap than we’ve seen in the past ten years combined.” (Turovsky, 2016). In another effort to make their service feel more human, Google released the Crowdspeak app in August 2016, which asked users to perform quick, simple tasks, such identifying handwriting and approving translations (Shankland, 2016).
Another huge leap forward for the technology was Google Translate becoming embedded in the Google Chrome internet browser. In 2010, this allowed users to translate entire webpages (Brinkmann, 2010). Although not particularly accurate, it now allowed users to access webpages that before they would have had to manually enter each individual sentence into a translator. In 2014, Google also announced a browser extension which would allow users of other browsers to utilise Google Translate, as well as allowing user to translated only the selected text on a web pages instead of all of it (Tian, 2014).
Another form of Google Translate that wouldn’t have been anywhere near as advanced even a few years ago, is the Google Translate app. In its current iteration, the Google Translate app is capable of real time visual translation, through the use of a mobile devices camera. The Google Translate app uses deep neural nets, which is a form of deep learning, that allows the system to identify images, or in this case words and letters, and ‘understand’ them (Nielsen, 2016). In order for the app to identify letters as they are in the real world, Google fed ‘dirty letters’ into its system that had deliberately been distorted in order for real world distortions to be catered to (Good, 2016). The app then forms the letters into words which it then translates into the desired language. Finally, the app renders the translation into the image, to replace the words in their original language.
When Google first fed the ‘dirty letters’ into the system they made sure they had the exact right amount that would allow the app to function but also wouldn’t take up too much data. This means that the app can be used in areas of low or slow internet, and even offline if the data has been already downloaded (Good, 2016). Another prominent feature of the app is its voice conversation mode, which uses much the same process as the real time visual translation, except instead uses the devices microphone, and voice analysis instead of image analysis.
From the web based applications of Google Translate, to the mobile apps, google translate utilises technology that even as recently as twenty years ago would have been almost inconceivable. Not only this, but Google Translate uses them together in such a way that many have argued has revolutionised language translation. These days, Google Translate can be seen to have an effect on many lives and industries, such as tourism, professional language translation, education, medicine, and even entertainment.
Google Translate is a free service that is accessible anywhere in the world; this makes it seemingly ideal for international travel. Google markets its translation software as breaking down language barriers and bringing the world closer together. It is the poster product for globalisation, and to a certain extent it serves this purpose well. If you wish to ask for the bathroom or order food in a foreign country, you will probably be successful but when you require large amounts of information such as a hotels entire website you may struggle (Suau, 2015). Another major concern when travelling is if you suddenly require medical assistance.
Language barriers have long been an issue in the medical profession. Whether tourists or immigrants, medical professionals are bound to encounter people who do not speak their language. With professional translators being expensive and often difficult to coordinate, Google Translate becomes an obvious tool to use. However, it would seem the use of Google Translate in this context could potentially cause more problems than it would solve. A 2014 study by Patil and Davies, found that Google Translate struggles to accurately translate medical jargon. For example, “Your husband had a cardiac arrest” was translated to “Your husband’s heart was imprisoned”, and “Your wife is stable” translated to “Your wife cannot fall over” (Patil and Davies, 2014). While comical to look at now, in a life-threatening situation, these incorrect translations could have dire consequences. This also leads to issues of consent, as a patient can’t consent to something they done fully understand.
Google Translate has also had a major impact on professional translators. While many thought that Google Translate and other such technologies would lead to unemployment for professional translator’s, studies have found that these technologies have in fact aided them. A 2015 study, found that Google search and Google Translate were the most used online tools of professional translators (Alonso, 2015).
Google Translate has also had a large effect on language education. While technologies such as spellcheck and Google have become commonplace in the field of education, it would seem there is more backlash against Google Translate in the classroom. A wide held assumption is that Google Translate is so inaccurate, that instead of helping students it would in fact teach them the wrong thing, especially with Google Translate’s difficulty with grammar and sentence structure. However, studies have found that while Google Translate is “far from able to produce error-free text… judging in relation to international testing standards, the level of accuracy is approaching the minimum needed for university admission at many institutions.” (Groves & Mundt, 2015). Used in tandem with other grammar and translation services, Google Translate could be a great help to students learning language.
Another, somewhat surprising use of Google Translate is in the form of entertainment. Recently, many videos have become quite popular wherein people perform popular songs but with lyrics that have been put through several layers of Google Translate, resulting in lyrics quite different from the original. One of the most successful people to do this is Malinda Kathleen Reese, who’s popular ‘Google Translate Sings’ series has generated over 30 million views and her first video in the series has gained over 9 million views (Malinda Kathleen Reese, 2014).
While Google Translate has proved it can be a valuable tool, it is important to consider who it really serves. Although accessible anywhere in the world, Google Translate contains a bias towards certain languages. The language Google Translate is best at translating to and from is English, followed by mostly major western European languages such as French and Spanish (Barré, 2011). This bias is probably due to the original corpus of text that Google Translate started with. The original text was made up of millions of translated United Nations documents (Nielsen, 2011) which lead to not only a language bias, but also influenced the formality and tone of the translations. This extremely evident in one of the ‘Google Translate Sings’ videos mentioned earlier. The parody of the popular Adele song ‘Hello’ makes it very evident where the translation originated from. In the video, the phrase, “There is such a difference between us” becomes “There is a conflict between the United States”, and “It’s no secret that the both of us” becomes “but it’s defence secretary saying our two”. These skewed translations indicate the type of text that the translations are based upon. Google admitted to this strange bias towards formal language when they first announced Google Translate in 2006, stating, “Our system works better for some types of text (e.g. news) than for others (e.g. novels) — and you probably should not try to translate poetry.” (Och, 2006).
Another issue with Google Translate is that it will always give you a result. While this might sound good, in actual fact it means that you will be provided with an answer regardless of if it is correct or not.
Another extremely important factor that is often not considered, is the issues of privacy and confidentially that come with using a public service like Google Translate. Google’s terms of service agreement states, “When you upload, submit, store, send or receive content to or through our Services, you give Google (and those we work with) a worldwide license to use, host, store, reproduce, modify, create derivative works communicate, publish, publicly perform, publicly display and distribute such content.” (“Google Terms of Service – Privacy & Terms – Google,” 2014). This means that everything you type into Google Translate, Google has the right to not only read but to also take for themselves and use however they wish. This may not be a concern if your translating your Spanish homework or having fun with song lyrics but many people do not realise that Google can store and use whatever they enter, and because of this may upload sensitive material, purely for the sake of saving time and money (Blake, 2015).
Overall, Google Translate is a successful culmination of many of the latest technologies resulting in a product that is for the most part extremely functional and useful, however, users of this technology should also consider the social, ethical and political concerns that are attached to using such technologies. It will be interesting to see how the technology will continue to advance, and to see what new concerns this advancement will bring.
About Google Translate. (n.d.). Retrieved 2 December 2016, from https://translate.google.com/about/intl/en_ALL/
Alonso, E. (2015). Google and Wikipedia in the Professional Translation Process: A Qualitative Work. Procedia – Social and Behavioral Sciences, 173, 312–317. https://doi.org/10.1016/j.sbspro.2015.02.071
Barré, J. (2011, October 31). PROMT, SYSTRAN, GOOGLE, BING – Has the age of machine translation finally arrived? » bab.la blog [bab.la blog]. Retrieved from http://blog.bab.la/sprache/promt-systran-google-bing-%e2%80%93-has-the-age-of-machine-translation-finally-arrived/
Blake, M. (2015, January 5). Man vs. Machine: Google Translate Jeopardizes Client Confidentiality, eDiscovery. Retrieved from http://abovethelaw.com/2015/01/man-vs-machine-google-translate-jeopardizes-client-confidentiality-ediscovery/
Brinkmann, M. (2010, February 7). Google Translate Integrated In Google Chrome 5 – gHacks Tech News. Retrieved 3 December 2016, from http://www.ghacks.net/2010/02/07/google-translate-integrated-in-google-chrome-5/
Good, O. (2015, July 29). How Google Translate squeezes deep learning onto a phone. Retrieved from https://research.googleblog.com/2015/07/how-google-translate-squeezes-deep.html
Google Terms of Service – Privacy & Terms – Google. (2014, April 14). Retrieved 2 December 2016, from https://www.google.com/policies/terms/
Groves, M., & Mundt, K. (2015). Friend or foe? Google Translate in language for academic purposes. English for Specific Purposes, 37, 112–121. https://doi.org/10.1016/j.esp.2014.09.001
Languages. (n.d.). Retrieved 3 December 2016, from https://translate.google.com/about/intl/en_ALL/languages.html
Malinda Kathleen Reese. (2014). ‘Let It Go’ from Frozen according to Google Translate (PARODY). Retrieved from https://www.youtube.com/watch?v=2bVAoVlFYf0
Malinda Kathleen Reese. (2016). Google Translate Sings: ‘Hello’ by Adele. Retrieved from https://www.youtube.com/watch?v=GMi4MtyDg40
Nielsen, M. (2011). Reinventing Discovery : The New Era of Networked Science. Princeton, US: Princeton University Press. Retrieved from http://site.ebrary.com/lib/alltitles/docDetail.action?docID=10498997
Nielsen, M. A. (2015). Neural Networks and Deep Learning. Retrieved from http://neuralnetworksanddeeplearning.com
NIST 2005 Machine Translation Evaluation Results. (n.d.). Retrieved 3 December 2016, from http://www.itl.nist.gov/iad/mig//tests/mt/2005/doc/mt05eval_official_results_release_20050801_v3.html
Och, F. (2006, April 28). Statistical machine translation live. Retrieved from https://research.googleblog.com/2006/04/statistical-machine-translation-live.html
Patil, S., & Davies, P. (2014). Use of Google Translate in medical communication: evaluation of accuracy. BMJ (Clinical Research Ed.), 349, g7392–g7392. https://doi.org/10.1136/bmj.g7392
Shankland, S. (2016, August 29). New Crowdsource app lets you work for Google for free. Retrieved 4 December 2016, from https://www.cnet.com/news/new-crowdsource-app-lets-you-work-for-google-for-free/
Suau Jiménez, F. (2015). Quality translation of hotel websites: interpersonal discourse and customer’s engagement. Onomázein, (32), 152–170. https://doi.org/10.7764/onomazein.32.8
Tian, C. (2014, October 15). Translate web pages more easily with the new Translate Chrome Extension. Retrieved from https://translate.googleblog.com/2014/10/translate-web-pages-more-easily-with_15.html
Turovsky, B. (2015, July 29). See the world in your language with Google Translate. Retrieved from https://googleblog.blogspot.com/2015/07/see-world-in-your-language-with-google.html
Turovsky, B. (2016, November 15). Found in translation: More accurate, fluent sentences in Google Translate. Retrieved 3 December 2016, from https://www.blog.google/products/translate/found-translation-more-accurate-fluent-sentences-google-translate/