Evolution of Machine Translation

By: Sheekha Sanghvi

By: Klaidi Shehi

Introduction

Humans have spent millennia trying to translate information across more than 7000 languages that exist around the globe. In fact, humanity could preserve a large portion of the ancient world’s knowledge thanks to the Rosetta Stone. She assisted historians in deciphering Egyptian hieroglyphs and the Bagdad House of Wisdom, where scholars translated thousands of scientific, historical, and religious texts into Arabic.

Translation has always been an essential tool in aiding the transmission of knowledge. As such, people began to inventing ways to make translation faster, more efficient, and more accurate over time. This need for translation and advancing of our technology has resulted in the development of intelligent computer systems like machine translation, or MT. MT essentially translates words much faster than humanly possible with an ever-decreasing margin of error. Here is a summary of Machine Translations’ humble beginnings and how it has developed into what it is today.

The Early 1950s

In 1949, Warren Weaver, a researcher at the Rockefeller Organization, proposed using machines to aid in the translation based on information theory and using machines to break codes during World War 2. This research  came to fruition in 1954 when Georgetown University and IBM demonstrated the first example of MT, where the system could translate 250 words and 50 carefully selected Russian sentences into English.

MT throughout the 1960s to 1990s

Despite setbacks from the Association for Machine Translation and Computational Linguistics (ALPAC) finding that MT could not compete with human translation in terms of speed and quality in the 1960s, MT research continued, especially in Canada, the United Kingdom and Russia. In the US, MT was used for governmental agency intelligence for organizations like the US Department of Defense and the Air Force. Going into the 70s to the 80s, demand for commercial translation began growing as commerce began globalizing and the need for translation for international companies and supply chains grew. This resulted in several MT companies launching, such as Trados, and several large organizations developing their MT systems, such as IBM, Toshiba, Mitsubishi, and Panasonic, to name a few.

By the 90s, there was a significant increase in the use of Machine Translation due to of the growing power of computers and the falling manufacturing costs  to make them. In addition, MT’s popularity grew as consumers began to adopt personal computers and move from mainframes allowing more consumers to access the technology.

The 2000s: MT and the Internet

MT experienced it’s most significant boom in adoption as several online Machine Translation services began popping up, such as Babel Fish Altavista, which could translate texts into several languages once people started using the internet. In order to better forecast the intended translations, these systems employed statistical machine translation to gather data from the most prevalent previously completed translations to predict the most accurate results. This further fueled MT adoption, and combined with the connectivity and convenience of the internet, MT user numbers skyrocketed.

MT Today

Accurate translation, at least on a smaller scale, has never been as accessible as it is today. Simple translations like “How to say have a good night in Turkish” can be done in seconds with a relatively high degree of accuracy through Google Searches. MT has also allowed for a reduced cost for translation services allowing for large documents that would have cost significantly more to now be priced at much more affordable rates. This helped increase the public’s  access to knowledge and information. The development of new MT systems such as Neural Machine Translation has also begun by many researchers and tech giants such as Google. This method works similarly to how the human brain constantly looks for suitable patterns and decision-making processes to develop a sequence of words. In 2020, this system could instantly translate texts with 60%-90% accuracy indicating that human editing and quality assurance are still necessary.

MCIS’ Machine Translation Services

MCIS has recognized the rise of MT and its demand, and we have made strides to develop our MT systems. In the near future, we want to provide our clients with machine translation services that will quickly and affordably translate various documents.  We will offer MT just on its own, MT with post-editing by one of our qualified and trained translators and even post-editing of MT from a client’s previous MT outputs. MT services are pretty affordable, with the cost being on a per word bases and depending on the level of services that are needed. It will vary on the quality of output desired but still be cheaper than traditional translation. Stay tuned for our MT services by subscribing to our newsletter and following us on social media like Instagram @mcislanguagesolutions.

Conclusion

Overall, Machine Translation has come a long way and is on the cusp of becoming not only the fastest way to translate languages but also the most accurate. MCIS is just as excited as you are to dive into this new tech to help everyone, no matter the language and aid in sharing information as efficiently as possible.

Citations

Austin Ringer. (2012). A (Brief) History of Machine Translation – Smartling. Smartling.com. https://www.smartling.com/resources/101/a-brief-history-of-machine-translation/

Nicole Post. (2018). The Evolution Of Machine Translation – Localize Blog. Localize Blog. https://localizejs.com/articles/the-evolution-of-machine-translation/

Summa Linguæ. (2021) A Brief History of Machine Translation: When did it start? Summa Linguae. https://summalinguae.com/language-technology/a-brief-history-of-machine-translation-when-did-it-start/