About

Yeon-Soo Lim

WORK EXPERIENCE

• AIR Lab@42dot(Hyundai Global Software Center) (2022.09 ~ )
• AIR Lab@Hyundai Motor Company (2022.03 ~ 2022.09)

Education

• M.S., in Computer Engineering, Kyung Hee University (2020.03 ~ 2022.02)
• B.S., in Computer Engineering, Kyung Hee University (2014.03 ~ 2020.02)

Technical Skills

• Hand-on experience in applying PyTorch, NLP tools (NLTK, etc.) and open libraries (huggingface, fairseq, etc.) for Korean NLP tasks
• Experience in several NLP tasks with Non-autoregressive decoding, text summarization and Natural Language Understanding
• Build ML pipeline system use Airflow, MLflow, Docker and kubernetes
• Python: Skillful, C++: Middle level

Research Experience

Non-Autoregressive Machine Translation (2020.10 ~ 2022.01)

• Research Non-Autoregressive Decoding technology for machine translation
• Proposes a novel machine translation model which changes the length of a target sentence iteratively to an optimal length
• Propsed model expand or delete token to find optimal target sentence length

Korean single and multi-document summary (2020.03 ~ 2021.12)

• Develop a topic-centric text summarization model with statistical attribute model. Apply PPLM to the text summarization for generate a topic-centric summary.
• Develop a text summarization model using BART
• Propose post-processing method to merge semantically similar summaries for multi-document summarization.

Korean dialogue model based on deep learning (2020.03 ~ 2020.12)

• Research on dialogue generation technology in the form of a single model for multiple domains. Using Continual Learning, a single conversational model can respond to multiple domains.
• Develop a conversation model using seq2seq structure and GPT2 that reflects the keyword score of tokens measured by the keyword extractor.

Publications

International Journal Articles

• Yeon-Soo Lim, Eun-Ju Park, Hyun-Je Song and Seong-Bae Park. A Non-Autoregressive Neural Machine Translation Model with Iterative Length Update of Target Sentence, IEEE Access, vol. 10, pp.43341-43350, 2022.
• So-Eon Kim, Yeon-Soo Lim and Seong-Bae Parkk. Strong Influence of Responses in Training Dialogue Response Generator, Applied Sciences, 11(16), 7415, 2021.

Domestic Conference Papers

• Yeon-Soo Lim, BongMin Kim, Choong Seon Hong, and Seong-Bae Park. 2021. Hate Speech Detection Model using Noise Self-training. In Proceedings of the Korea Computer Congress 2021, pp. 376-378, 2021.
• Sunggoo Kwon, Yeon-Soo Lim, and Seong-Bae Park. The Study on Korean Single Document Summarization Using Pre-trained Language Models. In Proceedings of the Korea Computer Congress 2021, pp. 379-381, 2021.
• Yeon-Soo Lim, So-Eon Kim, Bong-Min Kim, Heejae Jung, and Seong-Bae Park. A Query-aware Dialog Model for Open-domain Dialog. In Proceedings of the 32th Annual Conference on Human and Cognitive Language Technology, pp. 274-279, 2020.

Patents

Domestic Patent (in Korean)

• Cheoneum Park, Yeonsoo Lim. Method, Apparatus, and Computer-readable Medium for Error Correction on Document Summarization. 2023/03, (application umber) 10-2023-0033520.