The Yandex translator with the YandexGPT neuronal network became the leader in the quality of the translation of the English to Russian text according to the results of the Dibimt International Reference Point. In this category, the Russian service was around Google Translate and the GPT-4 model, which is used in Chatgpt.
Author:
https://rb.ru/author/tbatyrov/
Subscribe to RB.ru on Telegram
In addition to the Yandex translator, the five leaders included the GEMMA 2 Google, GPT-4 and GPT-3.5-5-Turbo of OpenAI, as well as the Deepl online translator. Google Translate took the seventh line of the qualification.
What about other languages? The Yandex service became the leader in translation from English to Russian for the second consecutive year. At the same time, last year also occupied the third place in terms of translation of English to Chinese, German, Italian and Spanish. This time in these categories, the Yandex translator was not included in the qualification. The company explained to RB.ru that “they did not request participating in translation nominations to other languages” to focus on the implementation of AI to improve English-Rusa translation, which is the greatest demand.
“Market [онлайн-перевода] Change quickly. The development of artificial intelligence opened new opportunities for players, and many large companies have introduced large language models in their products, ”said Peter Ermakov, responsible for the automatic learning brand in Yandex. According to him, the company preserved the leadership of the quality of Russian translation precisely thanks to the integration of Yandexgpt.
According to the company, the introduction of a large language model “allowed to double the precision of the translation during the year.” “Since the large language model works with the entire text at the same time, the translation of words that have several meanings is more precise, and the sentences are better agreed to each other. In addition, YandexGPT includes the style in which the text is written, ”said RB.RU in Yandex.
- The reference point Dibimt measures the percentage of errors in the translation, that is, the inappropriate interpretation of the meaning and meaning of the words. In total, 22 translation systems participated in the tests.
- The investigation is based on a database of several thousand prayers in English with homonyms, words that are written the same, but have different meanings. Models and services of translated offers, and reference algorithms and professional translators determined the proportion of words properly translated
Author:
Timur Bcyrov
Source: RB
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
