As Speech Translation Advances, Barriers Continue to Fall
AppTek Speech Translate is a two-way speech communication application that offers conversational real-time streaming speech-to-speech translation using the latest in artificial intelligence and machine learning. With it, users speak directly into their mobile devices to transcribe and simultaneously translate spoken content in AppTek’s 17 languages; the app delivers the translation as a natural-sounding spoken output and even provides access to 100 preloaded offline phrases to help in emergency, medical, transportation, and dining situations.
AppTek Speech Transcribe can convert voice memos, lectures, meetings, interviews, and other spoken audio content into text in real time across AppTek’s 17 languages. It can also be used as an accessibility aid when speaking with deaf or hard-of-hearing individuals.
“These simple-to-use and highly accurate applications enable users to speak fluently in multiple languages across a wide range of domains, including travel, healthcare, conversational, and more, greatly enhancing communications in a breadth of situations,” said AppTek CEO Mudar Yaghi in a statement at the time of the release.
Another tech leader, TransPerfect, is expanding the number of languages and dialects that its GlobalLink platform for transcription and subtitling supports through a partnership with AppTek.
Through the partnership, GlobalLink will be able to produce transcripts and subtitles among the 30 languages and dialects that AppTek currently supports. In addition, AppTek’s patented deep neural network-based intelligent line segmentation solution allows for the automated conversion of transcripts to subtitle files, with text broken up based on both syntax and semantics to reduce the amount of human effort needed to create subtitles.
More Important in Call Centers
While machine translation has many uses, it has become more important than ever in call centers since the onset of the COVID-19 pandemic, according to Jen Snell, vice president of product marketing for Verint Intelligent Self-Service. Call volume all over the world rose dramatically once government-mandated lockdowns began, particularly as people ramped up their online purchases and digital transactions. “The pandemic has provided us an opportunity to build our knowledge rapidly,” Snell relates.
Verint had already built its own speech recognition engine capable of supporting 70 languages for contact centers around the world. The engine offers flexibility with conversational AI, voicebot integrations, multiple speech recognizers, and semantic processors, as well as other Verint offerings.
In the contact center, Verint’s ASR helps improve operational efficiency with conversational AI to understand what callers are saying so that the system can provide the right self-service options or transfer calls to the right agents.
“We can offer faster out-of-the-box support,” Snell says of Verint’s machine translation capabilities. However, like other experts, she notes that machines can’t handle all translation tasks by themselves, particularly when the caller has a very heavy accent.
As translation technologies improve, Reager sees them moving into other verticals where they previously didn’t work well enough. One of the main ones is education.
Reager predicts that speech translation will become more embedded in the classroom to the point where students can take classes from several universities in different countries without language concerns. Online lectures, books, and other course materials could all be translated into several languages to enable students to assemble an education from anywhere.
Speech translation systems are also being used throughout the world in medical facilities, hotels, retail stores, factories, and police departments and are basically applicable anywhere that spoken language is used to communicate.
Speech translation can be used in so many places and for so many purposes now largely because of recent advancements in computing power that have enabled it to move to more devices and platforms, Reager and others contend.
Far from Perfect
However, despite all of the advancements, how much companies can rely strictly on machine translation and how much they need to augment machine capabilities with human interpreters depends on how precise the translation needs to be and the type of information being translated, according to AppTek’s Matusov.