텍스트 데이터 감성 분류 개선을 위한 Fine-tuning 방법 제안

Abstract

Research on efficient fine-tuning methods for Korean sentiment analysis Park, Jung Il Advisor : Prof. Kim, Pan Koo Ph.D. Department of Software Convergence Engineering Graduate School of Industrial Technology and Entrepreneurship, Chosun University Modern enterprises are increasingly adopting sentiment analysis as a crucial task, highlighting the accurate understanding of consumer opinions across various domains such as social media, product reviews, and customer feedback as a key challenge for success in competition. Sentiment analysis helps enhance products and services by grasping diverse opinions and sentiments of consumers. The utilization of large-scale text data collection and fine-tuning with pre-trained language models plays a vital role in improving performance. With recent technological advancements, sentiment analysis models demonstrate high performance, and the ELECTRA model, in particular, provides outstanding results through efficient learning methods and fewer computational resources. The purpose of this study is to optimize the performance of the model through efficient fine-tuning on various datasets using the KoELECTRA model that learned Korean in ELECTRA. An AI model to classify human sentiments as positive and negative and predict them. Through this, we expect to be able to achieve higher and faster performance in NLP tasks by efficiently fine-tuning various datasets suitable for the purpose.