Artificial Intelligence and Computer-Assisted Interpreting: Which Future for Professional Live Reporters and Subtitlers?

The Origins of Language Technologies

The developments of artificial intelligence and computational linguistics in the field of written language (machine translation or MT), which started in the 1950s, predate those of oral language. Together with natural language processing (NLP), they form the basis of current speech-to-speech and speech-to-text technologies (Checcarelli, 2009). The former are meant for automatic oral translation (machine interpreting) whereas the latter are meant for written transcription or translation of oral language (intralingual or interlingual live reporting and subtitling).

Why did technology first make its way into the world of translation and then into the world of interpreting, reporting and subtitling? Because these are market niches that are in high demand but are also difficult to scale to meet the growing demand for multilingual communication (Slaughter Olsen, 2023) and information processing.

Computer-Assisted Technologies for Interpreters, Live Reporters and Subtitlers

However, this does not mean that all the technologies that have been deployed are intended to replace human language professionals. In fact, some of them can be classified as computer-assisted interpreting (CAI) technologies, which come to the aid of both conference interpreters and live reporting or subtitling professionals.

Think, for instance, of the pandemic period, when already-existing technologies were perfected and new features were added to remote or hybrid videoconferencing platforms to support interpreters and live reporters and subtitlers in their work. (Slaughter Olsen, 2023) These setting-oriented technologies—such as Zoom, Kudo and Interprefy—were then joined by process-orientated computer-assisted technologies such as Interplex, Terminus, Interpreters’ Help, LookUp, Dolterm, Sketch Engine, Readwise Reader, Gaby-T, Intragloss and InterpretBank. (Fantinuoli, 2018; 2023). Moreover, among the latter, ChatGPT is a multifunctional chatbot which language professionals can currently use in combination with traditional tools to prepare for conferences, research and define terminology and idiomatic phrases, produce glossaries, transpose text into a different register, translate common languages and so on.

Computer-assisted technologies for consecutive interpreting already existed in the 1990s. The SimConsec system, which was devised by EU interpreters, consisted of using a digital voice recorder that recorded the original speech. The speech was then listened to again by the interpreters, who then did simultaneous interpreting. Another tool was the Consecutive Pen, introduced in 2014 and with several subsequent modifications. It was a digital pen used to take notes on a special paper that allowed consecutive interpreters to return to a word and listen to the corresponding audio in case of possible misunderstandings. The SightConsec system is still in use today and is based on automatic speech recognition; it basically consists of sight translation of oral text, which is transcribed while the speaker is speaking in real time by automatic speech recognition technology. It is a system that works, although there are clearly some speech recognition errors which the interpreter can compensate for in this case (Checcarelli, 2021).

Regarding computer-assisted technologies for simultaneous interpreting, there is a very recent project, from the universities of Ghent and Mainz, Ergonomics for the Artificial Booth Mate (EABM). It is an augmented interpreter, which consists of the possibility for the simultaneous interpreter in the booth to read the real-time transcription or translation of the speech on the screen by means of speech recognition, in order easily to recognise figures, proper names, technical terms and other such details.

Speech recognition is also used by intralingual and interlingual respeakers to produce, respectively, live captions or live subtitles of conference speeches, as well as live reports into the same or a different language.

Machine Interpreting, Reporting and Subtitling Technologies

In recent years, especially since the pandemic, developments in computer-assisted technologies have not come to a halt. In addition to Google and Meta (Santos, 2023), technology companies that had invested in improving videoconferencing platforms for use by interpreters and live reporters and subtitlers further developed speech recognition technologies combined with machine translation, to create automatic multilingual subtitles or to translate the speaker’s speech directly via video as if the technology itself acted as an interpreter (KUDO 2022).

The point is that automatic “interpretation” does not exist. Machine interpreters are speech translation systems in which translations are pre-loaded in order to function only in specific domains. Moreover, machines cannot feel emotions, do not know the context, do not understand the intentions of the speaker and reactions of the listener and sometimes cannot distinguish words, acronyms and jargon. Automatic simultaneous interpretation is even more complicated, because what is required is to reproduce a continuous flow of speech, without interruptions, with adequate contextual knowledge and the ability to predict what the speaker will say next (Fantinuoli, 2018).

In live reporting or subtitling, there is always a need for human intervention: when the machine makes speech recognition or translation mistakes, the respeaker or editor can correct them in real time, by interpreting the context and the speaker’s intentions, conveying the real meaning and intent of the original message and making the written outcome readable and accessible to a deaf or a foreign audience.

Which Future for Language Professionals?

All this leads us to say that technology is certainly of great help to professionals, now more than ever, because as the contexts of intervention change, as the market changes, greater specialisation is also required. Clearly, automated technologies will not be able to replace human interpreters or live reporters and subtitlers, especially in more delicate environments such as diplomatic and military and in medical environments, where people’s health and safety are at risk and good and accurate communication is fundamental to avoiding misunderstandings (Jo Kent, 2023a, b).

The pioneer of machine translation, the Israeli philosopher and mathematician Yehoshua Bar-Hillel, said in 1967 that machines would be intelligent if they could do three things, which they are still unable to do today—manipulate language, have knowledge of the world and be capable of reasoning and calculation (Press, 2023a, b), as well as have social skills, cultural understanding, empathy and intuition, be capable of producing a true equivalent effect from one language and culture to another and mimic the complexity of human reasoning.

In conclusion, it is computer-assisted technologies that are the future, because language professionals also need technology to make their job easier, to improve performance and to increase productivity. The future will not be for automatic interpreting, reporting or subtitling because the human aspect is irreplaceable. To maintain this, we must first recognise our value and our humanity, because interpersonal exchange, human interpretation and mutual understanding are the true bases of our professions.

Alessandra Checcarelli is a professional conference interpreter, translator and respeaker for Italian, English, German and French languages. Since 2015 she has been a member of  onA.I.R. Intersteno Italy and jury member of Intersteno international competitions in Berlin and Cagliari.

References: 

Checcarelli, A. (2009). Interpretazione automatica o assistita? Il rapporto tra l’interprete e l’intelligenza artificiale. Unpublished MA dissertation.                                                                                                                                                     

Checcarelli, A. (2021). L’interpretazione di conferenza: dalle origini alle moderne tecnologie CAI. –  InterGlobArte blog. URL: https://interglobarte.wordpress.com/2021/05/15/linterpretazione-di-conferenza-dalle-origini-alle-moderne-tecnologie-cai/                                                                                                                                                     

Fantinuoli, C. (2018). Computer-assisted interpretation: challenges and future perspectives. In Corpas Pastor, G. & I. Durán-Muños (eds) Trends in E-Tools and Resources for Translators and Interpreters. Brill.

Fantinuoli, C. (2023). Machine Interpreting and Translation Universals. – claudiofantinuoli.org. URL:  https://claudiofantinuoli.org/blog/2023/10/02/large-language-models-and-translation-universals/ 

KUDO (2022). “KUDO Unveils KUDO AI; the World’s First Fully Integrated Artificial Intelligence Speech Translator”. KUDO Newsroom. URL: https://kudoway.com/press-release/kudo-unveils-kudo-ai-the-worlds-first-fully-integrated-artificial-intelligence-speech-translator/

Ergonomics for the Artificial Booth Mate (EABM). Ghent University. URL: https://www.eabm.ugent.be/eabm/

Press, G. (2023a). Lessons learned from computer conversations and taming AI 70 years ago. – Forbes. URL:  https://www.forbes.com/sites/gilpress/2023/07/28/demonstrating-why-ai-cant-do-high-quality-translation/                                                                            

Press, G (2023b). Demonstrating why AI can’t do high-quality translations. – Forbes. URL:  https://www.forbes.com/sites/gilpress/2023/07/28/demonstrating-why-ai-cant-do-high-quality-translation/

Jo Kent, S. (2023a). Part 1: Why Conference Interpreting is the Wrong Model for AI and The World. – Medium. URL: https://stephjoke.medium.com/why-conference-interpreting-is-the-wrong-model-for-ai-and-the-world-part-1-5a9fd0fa5abf

Jo Kent, S. (2023b). Part 2: Why Conference Interpreting is the Wrong Model for AI and The World. – Medium. URL: https://stephjoke.medium.com/part-2-why-conference-interpreting-is-the-wrong-model-for-ai-and-the-world-558164f0f8fd

Slaughter Olsen, B. (2023). Interpreting and Technology – The Revolution at our Doorstep. – Multilingual. URL:   https://multilingual.com/issues/august-2023/interpreting-and-technology/

Santos, E. (2023). How A.I. is gaining ground in simultaneous interpretation. – El Paìs. URL:  https://english.elpais.com/science-tech/2023-09-03/how-ai-is-gaining-ground-in-simultaneous-interpretation.html