1 A very brief history of machine translation

Machine translation has developed dramatically over the last 70 years. Watch the following animation to learn about the history of machine translation, its development, and what it can and can’t do for language learners.

Download this video clip.Video player: Video 1

Show transcript | Hide transcript

Transcript: Video 1 A very brief history of machine translation

SPEAKER:

A very brief history of machine translation. Machine translation is an automated process to translate text or speech from one language to another. It is very easy to use, widely available online, and free 24/7, with countless applications just a search away.

But how did machine translation become what it is today? To trace its origins, we go back to 1949 when Warren Weaver, a mathematician at the Rockefeller Foundation, proposed the idea of a machine capable of translation. His vision sparked an enthusiasm and funding started flowing in.

Within a decade, the first machine was built. Initially, machine translation was rule-based, relying on grammar rules and dictionaries. However, computers struggled with understanding grammar and meaning. Results were disappointing, and so the project was abandoned in 1966.

In the 70s and 80s, researchers found that machines could translate repetitive texts using pre-translated sentences, a method called example-based translation. This approach had its first success translating weather forecasts between English and French in Canada, revitalising interest and funding in the 1980s. By the 90s, the proliferation of texts on the world wide web and the computational power of machines led to statistical machine translation.

This method used frequency and probability rather than rules or examples to produce translations. In 2006, Google Translate launched, bringing machine translation to anyone with an internet connection. Despite advancements, statistical methods had issues with fluency and accuracy.

The breakthrough came with artificial intelligence. In 2016, Google Translate adopted neural machine translation, NMT, which uses neural networks to process words as numbers and sentences as patterns. While NMT produces more human-like translations, it doesn’t understand emotions, metaphors, or context, often resulting in errors and biases.

Machine translation might seem easy and straightforward, but it has limitations. Understanding its history helps us make more informed choices. Machines calculate and match patterns, but do not understand or think. This is something that you must do.

[MUSIC PLAYING]

End transcript: Video 1 A very brief history of machine translation

Download

Video 1 A very brief history of machine translation

This short video offers a concise history of machine translation, tracing its origins from 1949 to modern-day Neural Machine Translation (NMT). It explains the evolution from rule-based, example-based, and statistical approaches, highlighting key milestones, such as the launch of Google Translate. Despite advancements, machine translation remains imperfect, requiring human understanding and intervention for the accurate interpretation of texts. The next section will discuss how to select machine translation tools appropriately and safely among the wide variety freely available online.