Naver's 7 Papers Including Core Research on 'HyperCLOVA' Accepted at 'EMNLP 2021'

[Asia Economy Reporter Kang Nahum] Naver will share its research achievements on ultra-large-scale artificial intelligence (AI) at ‘EMNLP 2021,’ the world’s most prestigious conference in the field of natural language processing.

The Clova and AI Lab research teams announced on the 3rd that they plan to present a total of seven papers at this conference, including the core research paper on the ultra-large-scale AI ‘HyperCLOVA.’

EMNLP, with a 25-year tradition, is regarded as the top AI conference in the field of natural language processing. This year, it will be held in the Dominican Republic from November 7 to 11 and will also be conducted online simultaneously.

Notably, this time, the core research paper on Naver’s ultra-large-scale AI, HyperCLOVA, was selected for presentation at the main conference. HyperCLOVA is the first ultra-large-scale AI publicly revealed by a domestic company in Korea last May, and it is currently the largest Korean ultra-large-scale language model in the world among publicly available single models, having been trained on Korean data over 6,500 times more than GPT-3.

This research introduces the Korean ultra-large-scale language model ‘HyperCLOVA’ and the data used for its training, and verifies the performance of models of various sizes. The paper demonstrated that HyperCLOVA shows excellent learning performance on various Korean language tasks even with limited examples.

Additionally, it explains the innovative changes in AI service development methodologies brought by ultra-large-scale AI such as ‘No Code AI.’ A total of 37 authors participated in this research, including researchers from Naver’s Clova and AI Lab, as well as software platform engineers, search engineers, and members from various teams.

Ha Jung-woo, head of Naver AI Lab, emphasized, “This paper is a very meaningful achievement in that the global natural language processing academic community recognized the value of Korean AI beyond the existing English-centered language model research.”

Furthermore, Naver plans to introduce research on techniques that efficiently augment and distill data by automatically generating and labeling data using ultra-large language models related to HyperCLOVA.

Meanwhile, Naver will also participate as a silver-level sponsor at EMNLP 2021 together with Naver Labs Europe. The company plans to actively exchange with global IT companies and academia in the future and contribute to the advancement of AI technology.

This content was produced with the assistance of AI translation services.