Naver Clova and Papago Present 100 AI Papers at Global Conference

[Asia Economy Reporter Yuri Choi] Naver Clova and Papago announced on the 4th that they published 100 regular papers at global artificial intelligence (AI) conferences last year. This shows growth in research capabilities compared to 69 papers in 2021.

Notably, a study on methods to efficiently apply ultra-large language models to services was accepted at EMNLP 2022, the most prestigious conference in the field of natural language processing. In this research, Naver proposed a quantization-based additional training technique that minimizes the degradation of generation and understanding quality of the model while reducing memory usage and increasing speed when applying ultra-large AI to services. This is a key technology that can reduce the cost of utilizing large-scale AI technology in services.

A research paper on technology that enables AI to remember and manage previous conversation information during long-term dialogues with AI was also presented at EMNLP 2022. This technology has been applied to Naver’s AI check-in call service for elderly people living alone, called ‘Clova Care Call,’ providing a more personalized empathetic conversation experience to service recipients. At the global speech signal processing conference Interspeech 2022, studies were presented on methods to easily apply emotions to speech synthesis and improve speaker recognition performance, contributing to the competitiveness of voice AI services such as Clova Dubbing and Clova Note.

In addition, achievements were made in machine learning and computer vision. At ICLR 2022, the most prestigious conference in the field of machine learning, Naver became the first Korean company to publish double-digit papers (13 papers). At the two major global conferences in computer vision, CVPR 2022 (14 papers) and ECCV 2022 (13 papers), Naver also became the first Korean company to publish double-digit regular papers.

Discussions on AI ethics are also being continuously led. In June last year, Naver held a workshop on ultra-large AI ethics at ACM FAccT, the world’s top conference in computer science fairness, sharing insights with leading experts in the field. More than 100 researchers, including those from Google, OpenAI, and DeepMind, participated, showing high interest.