NCSOFT publishes three AI academic papers at INTERSPEECH 2022 

South Korea’s gaming giant NCSOFT Corp. will present three academic papers at INTERSPEECH 2022, the world’s largest and most comprehensive conference on the science and technology of spoken language processing, the backbone of voice artificial intelligence (AI).

This year’s conference is being held at Incheon Songdo ConvensiA through Thursday. 

NCSOFT operates the largest AI center among gaming companies in the country. 

The company’s Speech AI Lab operated by the AI ​​Center has been conducting research on speech synthesis and sharing the results with academic societies. 

The three thesis topics accepted this year include a study on conducting separation modeling on voice color and pitch based on adversarial multi-task learning; a study on improving speech quality using a pre-learned neural vocoder; and a study on improving the naturalness of synthetic sounds.

“NCSOFT’s voice AI technology is already looking at the next step beyond commercialization,” Speech AI Lab’s manager Kim Hee-man said. “We plan to continue to share our progress.”

Through the latest research, NCSOFT claims it improved speech synthesis (TTS) and singing voice synthesis (SVS). 

Speech synthesis consists of the artificial production of human voices, with the main use being the ability to translate text into spoken speech automatically. 

Singing Voice Synthesis is a generative task that produces acoustic waveforms of singing given lyrics and melody input.

At INTERSPEECH, visitors can check out how far NCSOFT’s voice AI technology has come and consult on job vacancies in the industry.

Write to Jee Abbey Lee at jal@hankyung.com