First Place Award at the 2020 NAVER AI Rush

in Natural Language Processing

▲ (From the left) Ahn Hwijeen and Jeong Yeong-hun, Master's students

in the Dept. of Computer Science and Engineering

Ahn Hwijeen and Jeong Yeong-hun, Master's students in the Department of Computer Science and Engineering at Sogang (Supervisor Prof. Seo Jungyun) achieved an excellent score at the 2020 NAVER AI RUSH. Hosted by Naver Corporation, the 2020 AI RUSH, which took place over a period of two months beginning in July, was an AI project to resolve problems in diversified fields, including natural language processing, image analysis, and voice recognition. Hwijeen and Yeong-hun won the first prize in “Spam Detection” in natural language processing, and took third place in “Grammar Check,” winning an KRW 18 million prize and the qualification to be exempted from application document review and practice tests if they apply to Naver for employment.

Their award attracted special attention because of their unusual background. Both of them majored in American Culture and Chinese Culture respectively at the School of Humanities and International Cultures as undergraduates. American Culture major Hwijeen said, "Thanks to Sogang's flexible curricular system which allows extensive opportunities for non-majors in computer engineering, I was able to learn the basics of computer engineering." In his senior year, he developed a deeper interest in computer science and took related courses. Yeong-hun explained his motive behind joining the AI RUSH, saying "For the 2020 competition, AI RUSH was for big data collected from Naver portal, and I was intrigued by the idea that the AI system of my own making might be applied to an actual service."

The team used the “Ensemble” method to combine multiple deep learning models to solve the spam detection problem. Using CNN models to detect patterns in sentences and RNN models to reflect contextual information, they achieved high functionality. For the Grammar Check category, which aimed to develop an AI system to correct errors in sentences such as punctuation, they proposed the Data Augmentation Methodology to add errors to grammatically correct sentences to resolve the problem of insufficient learning data.

Ahn Hwijeen said, "It seems that most data are in English, and so is the research in natural language processing techniques. I wanted to help a wider range of people access knowledge by lowering the barriers of language and lingual technology." Jeong Yeong-hun expressed his ambition by saying, "If AI just follows the social prejudice and stereotypes as they are, it is inevitable that someone will be discriminated against because of technology. Therefore, I want to do research in ways that remove unnecessary prejudices from AI so that anyone can use value-neutral technology.“  

▲ (From the left) Ahn Hwijeen, Master's student in the Dept. of Computer Science and Engineering,

Prof. Seo Jungyun, and Jeong Yeong-hun, Master's student