경량화 언어 모델 기반 비정형 물류 주문 정보 자동 추출에 관한 연구
A Study on Automatic Extraction of Unstructured Logistics Order Information based on a Lightweight Language Model
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Recently, Large Language Models (LLMs) have made revolutionary advancements in the field of natural language processing, playing a significant role in the processing and analysis of unstructured data across various industries. Particularly in the logistics industry, efficiently handling unstructured data such as customer order information has become extremely important to maximize operational efficiency. However, traditional natural language processing techniques, while processing text as unstructured data, have limitations in understanding complex contexts or extracting nuanced meanings. This study proposes an approach to processing unstructured data through a lightweight language model that can run on a PC. Specifically, it addresses the limitations of existing statistical methods or rule-based approaches that cannot fully comprehend complex meanings and contexts. We believe that processing unstructured data in the logistics industry will serve as foundational data to systematize manual tasks and enhance efficiency. As a result of experiments using a model refined based on actual E-mail data, we achieved 100% accuracy for structured data. Among the main fields of unstructured data, three achieved 100% accuracy, the fourth field achieved 80%, and the last field showed a figure of about 1%. These results suggest the potential to dramatically increase operational efficiency in the logistics industry, especially for small and medium-sized enterprises, by introducing lightweight models to process unstructured data.
Abstract
Recently, Large Language Models (LLMs) have made revolutionary advancements in the field of natural language processing, playing a significant role in the processing and analysis of unstructured data across various industries. Particularly in the logistics industry, efficiently handling unstructured data such as customer order information has become extremely important to maximize operational efficiency. However, traditional natural language processing techniques, while processing text as unstructured data, have limitations in understanding complex contexts or extracting nuanced meanings. This study proposes an approach to processing unstructured data through a lightweight language model that can run on a PC. Specifically, it addresses the limitations of existing statistical methods or rule-based approaches that cannot fully comprehend complex meanings and contexts. We believe that processing unstructured data in the logistics industry will serve as foundational data to systematize manual tasks and enhance efficiency. As a result of experiments using a model refined based on actual E-mail data, we achieved 100% accuracy for structured data. Among the main fields of unstructured data, three achieved 100% accuracy, the fourth field achieved 80%, and the last field showed a figure of about 1%. These results suggest the potential to dramatically increase operational efficiency in the logistics industry, especially for small and medium-sized enterprises, by introducing lightweight models to process unstructured data.
참고문헌 (22)
- 김경아, "토픽 모델링을 활용한 물류 분야로봇 활용 관련 연구 이슈와 동향 분석", 로지스틱스연구, 31(2), pp.87~99 (2023)
- 나진성, "친환경 공급사슬 관련 국내 연구에 대한네트워크 텍스트 분석", 로지스틱스연구, 24(1), pp.63~80 (2016)
- 왕치아, "콜드 체인 전자 상거래 물류 서비스에 대한 소비자 인식 변화 분석 : 냉동 해산물을 중심으로", 로지스틱스연구, 31(3), pp.39~55 (2023)
- Berechman, J., "Economies of scale in bus transit", Transportation, 12(2), pp.313~332 (1985)
- Garg, R., "Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough: Latest Trends in AI, Volume 2", pp.157~191 (2021), Springer
- Kang, Y., "Natural language processing(NLP)in management research : A literature review", Journal of Management Analytics, 7(2), pp.139~172 (2020)
- Kang, D., "Knowledge distillation for lightweight RoBERTa of Korean", Proceedings of the Korean Institute of Information Scientists and Engineers (KIISE), pp.31~34 (2020)
- Kim, D., "Spam Message Filtering Using Naive Bayes and Lifelong Machine Learning(LLML)", Journal of the Korean Institute of Information Scientists and Engineers : Data and Information Processing, 26(3), pp.147~156 (2019)
- Lee, S., "A Study on the Effect of Hybrid Text Analysis with LSP and Machine Learning for Logistics Efficiency: Focusing on Automatic Response Systems in Call Centers", Master, pp.2017
- Miwa, M., "End-to-end relation extraction using LSTMs on sequences and tree structures", Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, 1, pp.1105~1116 (2016)
- Sullivan, M., "The Digital Transformation of Logistics: Demystifying Impacts of the Fourth Industrial Revolution", pp.61~78 (2021), Wiley
- Swain, A. D., "Handbook of human reliability analysis with emphasis on nuclear power plant applications (NUREG", CR-1278), 1983
- Xie, Q., "Self-training with Noisy Student improves ImageNet classification", Proceedings of the IEEE, 2020, 10687~10698
- Yan, Z., "A partition filter network for joint entity and relation extraction", Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pp.185~197 (2021)
- Yoo, S., "Korean contextual information extraction system using BERT and knowledge graph", Journal of Internet Computing and Services, 21(3), pp.123~131 (2020)
- Zhong, Z., "A frustratingly easy approach for entity and relation extraction", Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp.50~61 (2021)