QF experts discuss whether AI can replace human translators

  • 2 months   ago

Members from HBKU entities underline future of translation profession, and role of machine translation

With the growing presence of artificial intelligence in the job market, it is predicted that future jobs will be significantly impacted by automation. Many of today’s jobs, across various industries, may get replaced by artificial intelligence (AI) tools – which may result in existing professions ceasing to exist.

In the translation industry, there is no doubt that AI technologies have increasingly evolved the way professional translators work. There has also been an increasing reliance on machine translation to facilitate communication between people. But does this mean that at some point in the future human translators will not be needed?

Machine vs human

From a technical perspective, Dr. Ahmed Abdelali, Senior Software Engineer at Qatar Computing Research Institute, a part of Qatar Foundation’s (QF) Hamad Bin Khalifa University (HBKU), explains that while early machine translation relied heavily on resources curated by professionals in terms of bi-lingual lexicons and syntactic transfer rules, modern machine translation uses resources that are not specifically designed or intended for it.

Dr. Abdelali says: “Mining parallel corpora and extracting phrases from semi-parallel data is the new trend. And after acquiring a large collection of parallel data, researchers also explored using back-translation as a method to further expand their data.

“The recent methodologies combined with the advances in computing power allowed researchers to digest large amounts of data in the scale of billions of words.”

Texts produced by machine translation systems are typically very ‘fluent’, but the ‘adequacy’ in certain cases can be very low. - Dr. Ahmed Abdelali

The availability of resources in certain domains has made translation easier for machines to grasp, especially that certain areas of human communication do not require complex composition and high creativity. This made it feasible for machines to produce quality translations that are as good as human translations.

However, Dr. Abdelali explains that machine translation is still facing challenges to produce quality translation in certain types of text, such as creative texts.

“Texts produced by machine translation systems are typically very ‘fluent’, but the ‘adequacy’ in certain cases can be very low. It is possible that machine translation systems can learn clues from the context, but still there are more challenges in certain types of human language where sarcasm, cynicism, and irony is used.”

Machine and emotions

Although machine translations and automated captions have made great progress over the years, it still lacks absolute accuracy. According to Nada Al-Mahmeed, Director of Translation and Training Center at the Translation and Interpreting Institute (TII) at HBKU, one of the major challenges facing machine translation is the subjective nature of language.

AI does not yet detect humor or differences in voice tones, and does not analyze body language - Nada Al-Mahmeed

“Natural languages are subjective, and they are constantly evolving with time, and that is a major challenge that faces machine translation. While AI excels at tasks that are objective, natural languages are subjective ideas and communication between humans; and the languages we speak are grounded only in convention, not objective, reality,” she says.

Another major challenge that faces machine translation is that it is yet not able to capture feelings without human involvement.

“AI does not yet detect humor or differences in voice tones, and does not analyze body language. In the translation and interpreting field, all of these factors affect the translation, interpreting, and conveying of a message from one language to another.

“This, in turn, will always require human interaction, validation, and input. The level of human involvement, however, may always change.”

Upgrading the profession

The time it will take for AI technologies to become completely independent and no longer require human intervention is still unclear. But for the translation field, Al-Mahmeed says that AI will be helping in advancing the translation and interpreting field, rather than it is taking over. It will also create new and more important roles for translators.

Even if AI, at some point in the distant future, was able to develop a human-like command of our languages, that does not mean that it will dominate the translation field or leave human translators jobless. Nada Al-Mahmeed

“Even if AI, at some point in the distant future, was able to develop a human-like command of our languages, that does not mean that it will dominate the translation field or leave human translators jobless,” Al-Mahmeed says.

Translators will have bigger and more important roles, as they will be using machine translation output, and act as revisors and post-editors to ensure quality and adequacy of the text produced, Al-Mahmeed says. “In order to be an excelled reviser and post-editor, you would need to be an excellent professional translator.

“There is also a potential that the need for professional translators will be focused more on specialized areas such as legal and medical texts – where there is an extensive use of jargons and specialized terminology, and errors are not tolerated.”

As a leading institute in the field of translation, TII has been pursuing efforts to ensure that technology and AI are utilized to help translators and interpreters excel further at their jobs, while building local capacity – those who master AI and various other technologies to support the needs of the community.

“At the Translation and Training Center, which is part of TII, we expose our students to the professional use of the most advanced Computer Assisted Translation tools, and we offer training workshops for students and the wider community of professional translators.

“We also continue to build trusted translation memories and terminology databases in various fields, including medical, legal, and sports. Such continuous building of memories and databases will help translators complete a given task in less time and with increased quality.

 

“This will not only contribute to speeding up the translation process for professionals within the field, but will also ensure a consistent style, voice, and establish a ‘common language’ used in spoken and written content related to accessible and assistive technology.” Al-Mahmeed says.

Source: Qatar Foundation

Comments