The 2017 Speech Industry Star Performers: AppTek

AppTek Makes Captions Clearer

Applications Technology (AppTek) already had one of the most reliable captioning engines on the market, and it continues to improve it. The McLean, Va.–based company just last month added speaker change detection in the latest release of its automatic closed captioning platform, AppTek Caption. The release also includes automated punctuation technology and updates to its proprietary automated speech recognition (ASR) engine with specific models for news, entertainment, and general use. The end result, the company says, is increased accuracy and readability of captions produced for either live or prerecorded content.

“Our research has shown that the hearing-impaired community finds the viewing experience of our closed captioning platform preferable to the electronic newsroom techniques and on a par with human transcription. With the inclusion of our speaker change identification and the latest improvements to our automated punctuation, we continue to make our captioning platforms more accurate and engaging for the audiences we serve,” stated Adam Sutherland, AppTek’s CEO, in a statement. “The AppTek team’s deep expertise in neural networks, machine learning, and artificial intelligence allow us to continue to make breakthroughs in ASR technology as well as commercial products based on that technology.”

And to make that technology more available, AppTek, which was founded in 1990, acquired in 1998 by the embattled Lernout & Hauspie, and then went private again in 2001 after a management buyout, this spring launched a streaming audio application programming interface (API) for companies needing live transcription of audio streams. The API allows for instant transcription using automatic speech recognition.

“AppTek has been providing live and offline transcription for several years,” Sutherland said in a statement at the time. “Our Live Closed Captioning Appliance provides live transcription to newsrooms and other live venues globally, and our Media Workbench and RESTful API service deliver offline processing for multiple media and call center customers. Expanding to an instant, cloud-based API offering expands the companies and industries with whom we can partner.”

Jintao Jiang, AppTek’s chief scientist, worked with the technical and scientific team to adapt AppTek’s proprietary ASR technology to the live environment. “Tailored for the mobile environment, our English, Spanish, and Arabic API results are excellent. With up to 92 percent accuracy and less-than-one-second response time, we are very happy with this offering and are looking to expand the live API to the 11 other languages our ASR supports,” he said in a statement.

And some of those languages are not typical. AppTek’s hybrid machine translation engine covers many unconventional language pairs, including English with Arabic, Chinese, Dutch, Persian, French, German, Hebrew, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Turkish, and Ukrainian. Its unidirectional pairs go from Indonesian, Pashto, Tagalog, and Urdu into English.

Translating between those languages was made even easier this year with the latest version of AppTek’s Talk2Me speech-to-speech translation applications, released in the spring. The app leverages AppTek’s proprietary automatic speech recognition (ASR) technology and its real-time streaming API infrastructure for faster and more accurate translations.

Some of the companies already benefiting from the technology are Emergent, an IT solutions provider, and AD Digital, a Brazilian company specializing in media solutions for commercial and government sectors. AppTek also this past year inked a deal to provide the captioning for Sky News Arabia, representing the first time an Arabic-language news broadcaster is providing live closed-captioning to its broadcasts.