As I briefly reviewed in the previous post, neural network machine translation is a relatively new approach that appeared in the 2010s. Improvements in computer performance, advances in deep learning technology, and, crucially, the ability to build vast amounts of language data have brought about tremendous changes. Looking at the explosive growth of big data today, I'm excited to see how many more amazing innovations will occur in the future and transform human communication.
Principles of neural machine translation
Neural network machine translation is similar to a statistics-based approach in that it trains computers with data. However, there is a fundamental difference in that it uses much larger amounts of data and deep learning to build a kind of artificial neural network. Artificial neural networks resemble the brains of higher organisms made up of associations of numerous nerve cells.
Recent artificial intelligence technology has evolved by creating artificial neural networks to solve various problems and connecting them to build huge and complex network structures. Neural network machine translation can also be seen as one of these various artificial neural networks. The fact that it has evolved at a tremendous speed through various stages of development in a short period of time is reminiscent of the evolutionary process of our human brain.
Neural machine translation develops in a way similar to how a baby learns a language. The computer gives a lot of data to learn, makes it learn on its own, and develops a translator so that it can translate based on this. Of course, we make mistakes at first, but we keep improving as we remember them and try again over and over again.
Beyond the limits of human ability?
Modern systems say there's no need for humans to tell us what to look for in data. It's a method where artificial intelligence analyzes itself, finds patterns, and directly solves problems. Looking at this, expectations are bound to grow even greater due to the limitless possibilities that artificial intelligence has.
Machine translation continues to push the limits of performance through deep learning and big data. Humans learn vast amounts of information that cannot be acquired even if it takes a lifetime, and they show a problem-solving ability that surpasses humans. Basically, we just give the machine refined data to learn, and the machine can handle the rest of the work. This makes it difficult for us to grasp all the complex processing processes that take place deep within neural networks.
However, despite this, machines will eventually have no choice but to evolve through algorithms and learning data created by humans. If so, I think it's more accurate to think that machines enhance human abilities rather than go beyond the limits of human abilities.
As a tool for expanding human capabilities
However, it doesn't seem easy for machine translation to surpass the level of human experts right now. This is because all languages have developed complex and unique characteristics over the course of thousands of years of evolution. Therefore, in order to make an accurate translation, it is necessary to understand the culture of the relevant language area, and an advanced thought process of understanding the overall context and making inferences based on appropriate background information is required. In other words, it will still take a little longer for an artificial intelligence translator that boasts high-quality automatic translation capabilities to come out.
If so, why not consider machine translation as an efficient tool to help humans with translation work for a while? In addition to language differences, texts are more diverse than expected, such as colloquial language, written language, general documents, and specialized documents, so differences in translation quality are bound to occur depending on their characteristics. Instead, documents in a specific field of expertise can obtain highly accurate translation results if only the data, or corpus, is well-structured.
As an example, Twigfarm was able to significantly improve translation quality, especially for documents in specialized fields such as law, finance, machinery, and medicine by using translation dictionaries, which are databases that extract commonly used words and sentences in specific fields. In other words, rather than replacing humans, we are overcoming the limitations of general-purpose translators with specialized translation systems and making translation work more accurate and efficient by preventing human errors such as typos.
To the future of human communication
Even if you just look at the Tower of Babel stories, it seems that a world without language barriers is a long-standing dream for humanity. Even the 17th century philosopher Descartes, who couldn't imagine computers, proposed a kind of machine translation model. However, even today, when physical boundaries are disappearing due to the development of technology, communication difficulties due to language barriers still exist.
In this reality, machine translation breaks down language barriers that prevent communication between people, and can be a great tool for everyone who wants to communicate and understand each other. Advances in machine translation and artificial intelligence, which are becoming more accurate with artificial neural networks, are making humankind's old imagination a reality. Machine translators are already helpful enough as a means of communication in many ways.
Of course, if you expect a translation of the deep meaning or literary expression implicit in the sentence, it may still be lacking a lot. However, with the development of artificial intelligence technology, an era of strong artificial intelligence called AI (Artificial General Intelligence) comes, the limitations and problems of current machine translation may be completely solved.
It is impossible for us humans, who have a limited lifespan, to master all languages and knowledge. However, I believe that if we make good use of machine translation technology along with human delicate communication skills, even more wonderful things will happen in the future.
From free communication between diverse people speaking different languages, to the exchange of vast amounts of content from around the world, and even future communication between humans and artificial intelligence.
References
[1] Deep learning and machine translation https://terms.naver.com/entry.naver?docId=3580263&cid=59088&categoryId=59096
[2] Artificial neural network https://ko.wikipedia.org/wiki/인공_신경망
[3] brain https://ko.wikipedia.org/wiki/뇌
[4] Kang Byeong-gyu Lee Ji-eun, “How Neural Network Machine Translation Works and Translation Accuracy - Using Chinese Translation as an Example” (Chinese Language and Literature Society, 2018, 253-295)
[5] [Market Outlook] Evolution of AI interpretation and translation https://www.itdaily.kr/news/articleView.html?idxno=83486
[6] Kim Sun-mi, Coexistence of Human Translation and Machine (NMT) Translation in the AI Era - Focusing on the 'Augmentation (Augmentation) Strategy' in Business Administration (Korean Society of Interpreters and Translators, 2018, 1-32)
Good content to watch together
[AI Story] How Machine Translation Meets Artificial IntelligenceWhy does LETR, a language processing engine, focus on text languages?
