Early in the 1950s, machine translation technology started to take off. And this innovation has come a long way since then. It still falls short of the ability and dexterity a human intellect can bring to the task of interpreting a paper, though.
The Initial Step
A collection of recommendations on how to make the concept of machine translation a reality was put together in 1949 by Warren Weaver of the Rockefeller Foundation. To make it possible for machines to translate across languages, he combined the ideas of natural language, information theory, and code-breaking techniques developed during the Second World War.
The Georgetown-IBM project was one of the first machine translation achievements. In 1954, IBM displayed a device that could translate Russian phrases into English at its New York location.
The world was thrilled by the concept even if the machine could only translate 250 words (into 49 phrases). Money was being poured into this new area of computer science as a result of interest in this method on a global scale. The Georgetown experiment researchers anticipated that machine translation will be perfected within three to five years, full of optimism after their initial accomplishment.
Its Directionless Blocks
Despite the early optimism, machine translation turned out to be far more challenging than experts had anticipated. This is supported by the fact that, more than 60 years later, it is still not fully grasped.
To improve the technology utilized in the Georgetown experiment, generative linguistics, transformational grammar, and bilingual dictionaries were employed. Semantic ambiguity, however, was rapidly recognized as a problem. How would a machine translate a word that has several meanings and determine which meaning was meant in the original language and, consequently, which word to translate it into?
Early machine translations were far from flawless, even if they were at a level of quality that allowed for a rudimentary grasp of the original text. The race to master machine translation was taking far longer than anticipated (mostly between the US and the Soviet Union). By stating that machine translation was fundamentally not worth the time or money in 1964, the US Automatic Language Processing Advisory Committee (ALPAC) struck a setback to the US’s efforts. To assist human translators in their job, it was advised that resources should instead concentrate on automated tools (such as dictionaries).
How Does It Develop Internationally?
Other nations persisted in their efforts despite the US losing interest in machine translation (except for one or two significant private firms). Canada created the METEO System in the 1970s to translate meteorological information between languages. The system was of sufficient quality and could translate 80,000 words per day that it was utilized from 1977 to 2001 before being replaced with a new system.
Globalization was increasing the demand for it in other fields like never before. To defeat translation by machine, France, Germany, the Soviet Union, and the UK all put in a lot of effort. The cost and time savings for translating papers would be astounding if the art of translating using computers could be mastered. This information encouraged several governments and commercial businesses to continue their efforts, but they were still unable to create the ideal machine translation system.
By the end of the 1990s, the cost of these efforts had significantly decreased due to the increasing availability (and power) of computers. Japan, in particular, was aiming to lead the charge in the 1980s and early 1990s.
Some of the biggest technology businesses in the world began concentrating even more on machine translation in the 2000s. Along with Japanese initiatives, Google and Microsoft made significant investments in statistical machine translation in the US. Later, in the pursuit of improved outcomes, such attempts included fusing statistical systems with syntactic and morphological understanding.
Today’s Machine Translation
Undoubtedly, the development of computer translation has made many human translators uneasy. Since the highly publicized success of the Georgetown experiment in 1954, this has remained the situation. Many translators were concerned at the time that they might lose their jobs in a few years. Today, a lot of translators share this sentiment.
Despite these worries, computer translation is still not capable of outperforming human translators. In February 2017, Sejong Cyber University and the International Interpretation and Translation Association of Korea held a grand tournament to put this to the test. Three machine translations (Google Translate, Systran’s translation tool, and Naver’s Papago app) and four human translators each participated.
The correctness, linguistic expression, logic, and arrangement of the translations were assessed by three (human) translators. The four test texts were translated more quickly by the robots, but the humans easily prevailed with a top score of 49 out of a potential 60 points. Google Translate had the highest machine score, which was 28. Particularly, the subtlety of expression and feeling was emphasized as being unreachable by computers.
Machine Translation’s Future
It would be simple to assume that robots will be able to translate as proficiently as humans in a few years given current advancements in MT technology. But the same prediction was made in the 1950s, and it still hasn’t materialized.
Despite the enormous capacity of contemporary computing systems, history has given us cause to distrust the ability of machines to translate. It can be a useful tool for the time being, but it must be used in conjunction with meticulous post-editing performed by a human linguist as a part of a strong quality-control procedure, as we do in our MT solution.
Although machine translation has advanced significantly since the 1950s, it still has a long way to go before it can match the level of linguistic nuance that the human brain can produce. Until then, our amazing staff of knowledgeable human translators is ready to provide translation of your papers at a caliber that no computer can equal.