A model artificial intelligence which can directly translate speech from one language to another has created a research team at the American Meta Technology Society.
Most existing translation systems with mechanical learning are text -oriented or include multiple steps, namely speech recognition, translation into text and conversion of the text into speech.
In addition, linguistic coverage in existing speech models for speech is short of covering text models for text.
In attempting to deal with these restrictions the new model, called SEAMLESSM4Tmakes direct translations for up to 101 languages and can pave the way for quick translations, according to Post to Nature Magazine.
Specifically it can make translation from speech to speech by recognizing 101 languages and translating to 36, translation from speech to text (101 languages to 96), translation from text to speech (96 languages in 36), translation from text to text (96 languages) and automatic speech recognition (96 languages).
According to the research team, SeamlessM4T translates to speech to speech to speech translates to up to 23% more accurate than existing systems.
In an accompanying article commenting on the research in the same magazine, the Associate Professor at Tallinn Technology University in Estonia, Tanel Almeme, notes that the biggest virtue of this model is the fact that all the data and the code for executing and optimizing technology is publicly available.
However, it discerns that there are some obstacles, such as limited language translation or the difficulty of translating talks into noisy parts or among people with strong accents, something that people translators handle more easily.
Alison Keneke, an assistant professor in the Department of Informatics at Cornell University, is very interesting to the fact that researchers quantified the toxic, harmful or offensive language that translations can introduce and searched for any translation.
“Although speech technologies can be more effective and cost -effective in transcribing and translating people (who are also prone to prejudices and errors), it is imperative to understand the ways in which these technologies fail, disproportionately for certain demographics.”