PUBLICATIONS
The Latest Language News
Let us talk again about the latest language news. I have highlighted the news that I consider the most important.
SK Telecom launches A. Call Translator for real-time interpretation; Languages include Korean, English, Japanese, and Chinese
SK Telecom has announced the launch of A. Call Translator, a service that provides AI-based real-time interpretation during calls. Users of the service can activate live translation by tapping the service icon at the bottom of the phone’s dial pad and selecting Korean, English, Japanese, or Chinese. When users make a call using this service, the recipient will be informed that the conversation will be processed through interpretation. A. Call Translator can be used by SK Telecom subscribers on their iPhones through the A. app, and the person on the other end of the line does not need to be an iPhone user or A. app user. By combining telecommunications technology and AI, A. Call Translation is expected to enhance user convenience and ease language barriers.
Socotri language straddling survival amidst the absence of technology
https://globalvoices.org/2024/01/17/the-socotri-language-straddling-survival-amidst-the-absence-of-technologyThe ancient Socotri language is fading amid a surge of Arabic dominance, the war in Yemen, in progress since 2015, and the shifting winds of power in the region. The Socotra archipelago has long weathered the tides of exclusion and isolation amid shifting regimes in Yemen, and Socotra has emerged as a notable victim of political gambles, casting shadows over its people and language. The fog of marginalization has obscured Socotra’s cultural identity and hindered its linguistic progress.
The Socotri people are grappling with sudden technological openness. Despite the Socotri’s commitment to their native language, the absence of essential technological services like accessible internet restricts their expression, negatively impacting the survival of the Socotri language. Isolation has safeguarded the archipelago’s linguistic and cultural legacy for millennia, but a swift, unplanned technological leap driven by political shifts and the presence of the United Arab Emirates has become a double-edged sword for the Socotri, shrouding linguistic and cultural landmarks.
The Socotra archipelago, a living natural and historical museum, stands gracefully like a colossal ship in the Indian Ocean. Its strategic location connects the Middle East, Asia, and Africa and is a crucial passage for oil tankers. Comprised of four main islands adorned with rare fauna, the archipelago gained UNESCO World Heritage status in 2008. The Socotris, an indigenous ethnic group affiliated with the Mahra Qahatani tribe, proudly speak their ancestral language, Socotri, alongside Arabic.
As of the start of 2022, the internet penetration rate in Yemen was 26.7 percent, meaning 22.59 million of the 33 million people in Yemen lack internet access. The current state of fixed internet service in Socotra inadequately serves the Socotri people, lagging far behind the global information and technology revolution. Socotri faces extinction but thrives on its people’s determination. In the delicate balance between tradition and modernity, Socotri’s fate is uncertain. The narrative in Socotri society emphasizes the need for recognition, preservation, and adaptation, echoing the resilient spirit of a community committed to ensuring Socotri’s unique melody endures through time.
Unsupervised speech-to-speech translation from monolingual data
https://blog.research.google/2023/12/unsupervised-speech-to-speech.html
Speech-to-speech translation is a type of machine translation that converts spoken words from one language to another. The technology has the potential to break down language barriers and facilitate communication between people from different cultures and backgrounds. Translatotron 3, a novel unsupervised speech-to-speech translation architecture, shows that it is possible to learn speech-to-speech translation from monolingual data alone. This method opens the door not only to translation between more language pairs but also towards translation of non-textual attributes of speech such as pauses, speaking rates, and speaker identity. The method does not include any direct supervision to the target languages, and it is therefore the right method for the preservation of the paralinguistic characteristics of the source speech in the translation. Back-translation is used to enable speech-to-speech translation.
The quality of translation is measured using BLEU on automatic speech recognition transcriptions of the translated speech compared to the corresponding reference translation text. Speech quality is measured using the MOS score. Translatotron 3 outperforms the baseline by large margins in every aspect measured: translation quality, speaker similarity, and speech quality. It particularly excels on the conversational corpus and achieves speech naturalness similar to that of the ground truth audio samples.
Machine Translation Specialist, R&D
Artem Eidel
Artem Eidel is a Machine Translation Specialist in R&D. His responsibilities include creating and maintaining MT models, evaluating MT quality, and selecting suitable MT providers for different language pairs and subjects.
