Kakao Enterprise announced that its self-developed natural language processing AI model ranked first in the performance evaluation of the Korean machine reading dataset ‘Korean Question Answering Dataset 2.0 (hereinafter referred to as KorQuAD 2.0).’
‘KorQuAD 2.0’ is a large-scale Korean question and answer dataset. Since ‘KorQuAD 1.0’ was released as open source in December 2018, the current 2.0 is △understanding HTML documents containing lists, tables, etc. △technology to quickly process massive documents. The level of difficulty has increased even further, with demands for backing. Evaluation is done by presenting content extracted from the data set as a question, and the AI model finds the correct answer.
Kakao Enterprise’s natural language processing model ‘LittleBird-large’ ranked first with an accuracy score of 90.22. Not only accuracy, but also processing speed was about 1.7 times faster than the existing No. 1 model. In addition, the ‘LittleBird-base’ model, which was developed together to increase inference speed and improve practicality by optimizing the model size, ranked 6th with an accuracy of 88.57 points, about 3.4 times faster than other companies’ models with similar performance. Achieved success in processing data.
Kakao Enterprise plans to continue to enhance the model and apply it to search, chatbots, and the AI contact center ‘Kakao i Connect Center’. For example, when applied to the ‘Kakao i Connect Center’, AI counselors will be able to provide more accurate answers to customers’ inquiries or requests by utilizing a knowledge search system that applies the machine reading model.
In 2019, Kakao’s own language model ranked first in ‘KorQuAD 1.0’. At the time, Kakao’s language model received 91.85 points, 0.65 points higher than the 91.20 points found to be the human performance level of the human being evaluated, marking the first case where AI scored higher than humans.
Choi Dong-jin, head of Kakao Enterprise’s AI Lab & Service, said, “Machine reading is a technology that goes beyond simply understanding documents and provides optimized answers through analysis and reasoning, and has a direct impact on actual services using AI.” He said, “Kakao Enterprise will continue to strengthen its research capabilities in various areas such as vision and voice as well as natural language processing.”
Source: Pangyo Techno Valley Official Newsroom
→ Go to ‘Asia Innovation Hub Pangyo Techno Valley 2021’ news