[Pangyo Tech] Kakao Enterprise, a total of 25 papers registered in global academic conferences this year

Kakao Enterprise has been recognized for its global top-level AI research capabilities in 2021.

Kakao Enterprise announced on the 14th that it had registered two papers and made a session presentation at ‘NeurIPS’, the world’s best AI academic conference held from December 6th to 14th. As a result, the number of AI papers that Kakao Enterprise has registered in global academic societies this year reaches a total of 25.

‘Newlips’ is an academic society with the highest level of authority in the fields of artificial intelligence, machine learning, and brain science. It carefully selects and introduces only 30% of theses from excellent researchers, including leading groups in each field. Kakao Enterprise received the honor of an oral presentation given to the top 1% of researchers at Newlips in recognition of its research on methodologies that improved data bias issues in the computer vision field. The other paper is a paper proposing an algorithm that can defend against various attacks in the field of computer vision.

Kakao Enterprise stood out in the field of natural language processing AI this year.

In November, two papers were passed at EMNLP, the world’s top academic society in the field of natural language processing. One proposal proposes a model strategy and robustness evaluation dataset to implement a chatbot capable of natural human-like conversations. The dataset was released through GitHub, the world’s largest open source community. The other is a joint research project with Seoul National University and is a paper proposing a machine translation methodology using word alignment information.

The paper was also passed at two workshops hosted by ‘EMNLP’. First, at ‘WMT21’, the methodology that ranked first in a challenge to evaluate how effectively a new domain or a terminology dictionary of a specific domain is reflected in machine translation was revealed, and at another workshop ‘NewSum’, Korea University and Wise A joint research team, including Nut and Microsoft, was formed and presented a conversation summary methodology, which received favorable reviews.

Last October, we published a paper in IEEE’s SCI journal ‘Signal Processing Letters’ that presented a method to improve the learning word recognition rate and improved the word recognition rate by about 53% compared to the previous one, and in August, we published a paper in the renowned international academic journal ‘Signal Processing Letters’. A paper jointly researched by Kakao Enterprise and Sungkyunkwan University was published in ‘IEEE Access’. This is a paper on methodology to improve the performance of question-answering models. Using this technology, it took first place in the ‘Passage Ranking’ category of the global AI natural language understanding competition ‘MS MARCO’ in October.

Achievements in the computer vision field also stood out. In October, a paper was registered at the International Society for Computer Vision (ICCV), one of the highest-level academic societies in the computer vision field. He was recognized for proposing a new concept called ‘local logit’, which can reflect detailed spatial information compared to existing research. In the future, we plan to use this technology to efficiently redesign deep learning algorithms and advance facial recognition and computer vision-based services.

Kakao Enterprise stood out in global artificial intelligence challenges in various fields throughout the year, sweeping the top rankings a total of six times. In particular, it took first place in the ‘Passage Ranking’ field of the global AI natural language understanding contest ‘MS MARCO’ and the ‘terminology task’ field of the WMT21 Challenge, and the ICCV-MFR Challenge, which competes face recognition performance depending on whether or not a mask is worn. He also achieved second place and was recognized for his outstanding research capabilities.

Choi Dong-jin, head of Kakao Enterprise AI Lab, said, “We will continue to devote ourselves to AI research, share good technology cases, and create a virtuous cycle with more advanced services.” He added, “Ultimately, the goal is to accelerate the sharing and diffusion of AI technology and expand the technology ecosystem.” “It’s a goal,” he said.

Source:  Pangyo Techno Valley Official Newsroom

→ Go to ‘Asia Innovation Hub Pangyo Techno Valley 2021’ news