As Speech Translation Advances, Barriers Continue to Fall

Article Featured Image

Speech translation was already a rapidly growing market, thanks to evolving technologies, growing international travel, and business globalization. The COVID-19 pandemic slowed international travelers’ need for translation capabilities, but that need is expected to rebound as the world re-emerges from the pandemic.

Market research firm Mordor Intelligence forecasts that the speech-to-speech translation market will grow at a compound annual rate of 9.4 percent between now and 2025 thanks to further globalization, which it says will pick up again.

Speech-to-speech translation helps break down language barriers, helping to foster international business and tourism and to equalize communication ability, the Mordor forecast says.

Since the growth of the tourism industry is fueling the demand for speech-to-speech translation solutions, vendors such as Raytheon, Lingmo, and LogBar are actively positioning their products along those lines.

Mordor Intelligence and others also point out that the technology behind speech-based translation has improved immensely in just the past year or two, sparking further demand and interest.

The most prolific advancements have been in the areas of artificial intelligence, neural processing, natural language understanding, speech recognition, and text-to-speech conversion, making speech translation much better than it was just a few years ago, says Sue Reager, chief technology officer at Language Preservation Technologies. “Translation has moved from good to wonderful. Translation is strong at a very high level.”

Greater computing power and the ability to house it on smaller devices have also had a dramatic impact on speech translation technology.

Additionally, the technology has benefited from better audio quality—thanks largely to noise reduction and echo cancellation—that is possible on phones, headsets, intelligent virtual assistants, and other devices involved in translation, according to Reager.

“If the audio was crap, the speech translation was crap,” Reager says, noting that those issues have mostly been resolved. “Now you have great technology and great translation.”

Reager says speech translation has evolved from “an insane focus on English” a few years ago to excellent translation between most major languages. Mordor’s research, for example, found a rising demand for Mandarin Chinese, which is not surprising given China’s rising position in the world, but translation of less common languages is also possible today.

Reager’s firm is working on translations to and from Cherokee and other Native American languages.

The ability to speed up and slow down outputs is also making translations better for use with animation. Reager notes that the popular children’s television show Sesame Street is now available in Cherokee. The translation, she says, appears to be natural because the sound and puppet movements are in synch, which wasn’t possible earlier.

Another major advancement is the ability to add emotion and colloquialisms in speech translation, Reager says, maintaining that these advances have taken speech translation far beyond the robotic-sounding monotony that plagued it in the past.

Several Market Participants

It’s not just firms like Reager’s that are leading the innovations in speech translation technology. Some of the top technology companies have recently advanced their language translation capabilities as well.

Google, for example, earlier this year launched Transcribe for Android, a feature of its Translate app that can convert live speech in one language into another language almost instantly. Google Transcribe rolled out with support for English, French, German, Hindi, Portuguese, Russian, Spanish, and Thai.

Reager also credits Google with offering translations in some less common languages where there is no return on investment. “That is heroism of the first magnitude,” she says.

Fellow tech giant Amazon, meanwhile, has started offering several translation services through its AWS Lambda function, which connects with Amazon’s three artificial intelligence-based language services, Amazon Transcribe, Amazon Translate, and Amazon Polly. Moreover, Amazon Alexa can now provide voice translation from English to as many as 48 other languages.

“Machine translation works quite well with major language pairs,” says Evgeny Matusov, lead science architect for machine translation at AppTek, a provider of artificial intelligence, speech recognition, neural machine translation, and natural language understanding technologies,

AppTek in early July launched two consumer speech translation applications, AppTek Speech Translate and AppTek Speech Transcribe. With both applications, users can communicate in real time with other people in 17 languages (English, Arabic, Portuguese, Mandarin Chinese, Dutch, French, German, Greek, Hebrew, Italian, Japanese, Korean, Pashto, Persian, Russian, Spanish, and Turkish). The applications also support a variety of dialects, including 12 Arabic, two French, three Spanish, five English, two German, and two Portuguese dialects.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues