Meta says it has broken new ground with its Universal Speech Translator, which can now translate Hokkien. For those who are not aware, this specific Chinese dialect is mostly spoken rather than written, making it one of the more challenging languages for artificial intelligence (AI) systems to learn and translate to other languages.
The corporation has made the decision to train its Universal Speech Translator AI using speech-to-speech for this purpose as well as for other comparable languages that lack a written form. You may visit the firm’s website for a more thorough explanation, but the news release from the company gets to to the point about how this is accomplished. But if you simply want to understand the basics, keep reading because we’ll try to offer you a crash course on Meta’s AI prowess.
The Universal Speech Translator AI, according to the news release, first converts Hokkien into a series of acoustic noises, which are then utilised to produce the language’s waveforms, each uttered word having its own distinct waveform. These terms are then compared to their corresponding Mandarin equivalents (which have been predetermined by the system’s creators) and used as the foundation for translating the text into other languages and back again.
As the AI can only translate one sentence at a time, Meta claims that the Hokkien translator is still a work in progress. In other words, it will be some time before you can actually perform back-and-forth translation, much alone utilise the Universal Speech Translator to impress anybody who speaks the dialect. On the other hand, the business is making the technology open-source so that other researchers may improve upon it.
The AI of the Universal Speech Translator can currently support just the unwritten language of Hokkien, but as research moves forward, more are likely to be added. The project’s stated goal is to “eventually allow real-time speech-to-speech translation between all existent languages, even principal spoken ones,” according to Meta. The technology will let everyone communicate more easily by removing language barriers, but the business also made it clear that it would eventually play a role in its big Metaverse vision.