Introduction to Artificial Intelligence: does AI disrupt human translation and interpreting services? and why is it important?

Today, all document translations and interpreting from a source language to a target language can be done by Artificial Intelligence. Artificial Intelligence is primordial for every sector, in which the technology integrates with translating programs or a part of research motors. Yet, matching a source to a target language does not mean replacing a human-made work. Human needs to intervene to render a fluid and coherent piece of translation or interpretation. This article gives you insight and the importance of Artificial Intelligence.

Reading this article helps to answer these questions:

• What is Artificial Intelligence in translation and interpreting?
• How AI function performs its role and helps human?
• Where can AI be found?
• Does AI cause the disruption of human translation?

© 2021 Sorada T.T. All rights reserved

Artificial Intelligence with languages

According to IBM, Artificial Intelligence refers to “the ability of a computer or machine to mimic the capabilities of the human mind-learning from examples and experience, recognizing objects, understanding and responding to language, making decisions, solving problems, and combining these and other capabilities to perform functions a human might perform¹.”

AI applies to diverse technologies. In translation and interpreting, the two most functional AI components are Speech to speech translation and Natural Language Processing(NLP).

Natural Language Processing(NLP)

Natural Language Processing is a part of AI and is integrated into computer software or research motor to process human languages, such as Google Translation, Grammarly, and Bing Microsoft Translator². This technology translates written or verbal languages by matching them up between a source and a target language. The most traditional machines for translation or Computer-Assisted Translation(CAT) tools with NLP are SDL Trados Studio, MemoQ, Wordfast, and Déjà Vu.

Speech-To-Speech Translation

Speech-to-speech translation is also a part of AI. This technique has been developed for decades to break the language barrier. Data scientists designed it for people who came from different backgrounds, cultures, and languages. It processes speech recognition, translates into a target language, and generates speech in the target language³.

Nowadays, developers have mixed the STS system and created numerous head-gear gadgets to interpret a real-time conversation. The most well-known trademarks on the market are Google Earbuds, Timekettle, and Waverly Labs.

Technologies for machine translation and conference interpretation

Machine Translation

Machine translation came from the set of algorithms and statistics on computer systems. The computer acts as an automated translation machine that may use the rules-based system, statistical system, neural machine translation, or a hybrid combination⁴. Google Translate is an example of Machine Translation.

Contrary to the CAT tools, the software translates the source text from translation memories, a terminology database, and glossaries⁵. Within a second, users can get work done with the help of human-made data.

How? The term database in the software was already prepared, revised, and approved by linguists, translators, or professionals who have been brainstorming, discussing, and translating words for years. Thus, the professional got every perspective of how one word can be interpreted in a different meaning.

Machine translation vs. CAT Tools

Machine Translation and CAT Tools are two distinct systems. Using MT technology is quick and instantaneous, only in a few seconds. Yet, the quality of machine translation is dubious. The machine does not recognize the cultural and rhetorical aspects of the words. It only matches and translates texts from one to another language that are existed on the system.

For a simple translation, Machine Translation can do the work perfectly and accurately. For an elegant translation, such as an eloquent text, intricate poem, or a literature translation, human redaction or CAT Tools may be more suitable choices.

Conference interpretation, language technology, and hardware

Technology for a conference or simultaneous interpretation is approximately the same thing as translation. Today’s technology, AI is now operating with earbuds, headphones, or earphones. When combined with 4G/5G, internet, and translation engines, the hardware acts as a tool for interpreting a conversation or business conference in real-time.

The first example, Timekettle. The brand has invented its interpreting earbuds product using only the world’s top engines, such as Microsoft, Google Translate, iFlytek, DeepL, AmiVoice, and Hoya⁶.

Also, Google Pixel Buds uses Google Translate engine to interpret only a simple phrase. However, the earbuds version of 2017 did neither succeed in translating a complex sentence nor interpret a simultaneous conversation. Watch a crash-test video from Tech Insider to understand its efficiency.

Technology of translation engines for translation and interpreting

As explained above, all translation engines are using AI technology to interpret a conversation and translate a document from the source to the target language.

© 2021 Sorada T.T. All rights reserved

Application Programming Interface

Each translation engine comprises a different AI technology depending on its purpose and usage.

The first example, Microsoft is using the cognitive Application Programming Interfaces(APIs), in which Azure Cognitive Services and Azure Speech Service are parts of the Microsoft cloud family. Plus, the core of cognitive APIs is the Neural Machine Translation(NMT), which, as Microsoft has claimed, is more accurate, fluent, and human⁷.

The second example, the DeepL is also a translating API where it collects words and texts on its servers. The company claims that its AI technology comprises unique neural translation networks, which render a better translation quality than Microsoft and Google⁸.

Is AI going to replace a translator, linguist, or terminologist?

The answer is simple. Not that much. It depends on languages’ evolution because some of them are rather evolutive and growing as well.

New word’s definition in pop culture

The Thai language is among them. If it were 50 years ago, many words had only one plain simple meaning. Now, they evolve and have multiple meanings and usage. Eventually, they become modern-day slang, and only the natives would understand.

In Thai pop culture, this phenomenon occurs all the time. Celebrities or idols like to define and invent something catchy. They tend to explain a situation or a feeling most uniquely and fabulously. Thus, the words or phrases became popular and went viral.

Modern-day culture with MT & CAT tools

Understanding modern-day slang is complicated. Machine translation will not process correctly. Because it is too intricate, and it needs more background and history of words. CAT tools will be able to perform correctly if terminologists register it on their glossaries or software data. Unless terminologists are not educated themselves with new slang, idiom, or upcoming trends.

AI technologies for conference or conversational interpreting can neither understand all slang nor interpret pop culture references. AI must learn numerous elements: latest trends on social media, culture roots, communities, generations, and surroundings.

Theoretically, if AI can achieve those perspectives like a human being, there is a possibility that the human translation will be less significant. However, human-made work( i.e., glossary and terminology) needs real humans to accomplish and train AI.

Punctuation and grammatical errors

Not all AI technologies can 100% detect punctuation and grammatical errors on documents. But, Grammarly and Hemingway did a decent job for simple tasks.

Grammarly adopts hybrid AI technologies, such as machine learning, deep learning, high-quality training data, and natural language processing⁹. The application works fine to avoid silly errors. But, for a complicated grammar solution, users have to pay for a premium plan to get a human assistant.

Also, punctuation rules are much advanced than grammatical errors or misspellings. Misplacing a comma in one sentence may change the whole meaning. That’s the reason why AI is far from perfect in correcting texts.

Summary

AI technologies are here to do a massive load of translation. For example, the translation task of 100,000 words in a few days deadline, from the source to target language, is possible with the help of CAT tools or MT. Nevertheless, proofreading gig needs a real human being to reread and improve the quality of the translation. Saying that AI will disrupt human work may sound exaggerating.

© 2021 Sorada T.T. All rights reserved

References

1. IBM Cloud Education, “Articifial intelligence”, 3 June 2020, accessed 22 april 2021, https://www.ibm.com/cloud/learn/what-is-artificial-intelligence#toc-what-is-ar-DhYPPT4m
2. Diego Yse, “Your Guide to Natural Language Processing(NLP)”, 15 January 2019, accessed 24 april 2021, https://towardsdatascience.com/your-guide-to-natural-language-processing-nlp-48ea2511f6e1; IBM Cloud Education, “Natural Language Processing”, 2 July 2020, accessed 24 april 2021, https://www.ibm.com/cloud/learn/natural-language-processing
3. Ye J. and Ron W., “Introducing Translatotron: An End-to-End Speech-to-Speech Translation Model”, 15 May 2019, accessed 25 april 2021, https://ai.googleblog.com/2019/05/introducing-translatotron-end-to-end.html
4. LingoHub Academy, “What is Machine Translation(MT)?”, accessed 29 april 2021, https://lingohub.com/academy/glossary/machine-translation; Trados, “What is machine translation?”, accessed 29 april 2021, https://www.trados.com/solutions/machine-translation/; European Union Institutions, “New Technologies and Artificial Intelligence in the field of language and conference services”, 29 april 2019, https://ec.europa.eu/info/sites/default/files/final_host_paper_iamladp2019_en_version.pdf
5. Ibid.
6. Takashi Yamada, “These real-time in-ear translator earphones help you fluently speak in as many as 40 different languages”, 13 april 2021, accessed 29 april 2021, https://www.timekettle.co/blogs/news/these-real-time-in-ear-translator-earphones-help-you-fluently-speak-in-as-many-as-40-different-languages
7. https://docs.microsoft.com/en-us/azure/cognitive-services/translator/translator-info-overview#microsoft-translator-neural-machine-translation
8. DeepL, “Another breakthrough in AI translation quality”, 6 February 2020, accessed 30 april 2021, https://www.deepl.com/blog/20200206/
9. Grammarly, “How We Use AI to Enhance Your Writing I Grammarly Spotlight”, 17 may 2019, accessed 30 april 2021, https://www.grammarly.com/blog/how-grammarly-uses-ai/