A study on automated invoice recognition and text correction
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Recently, with the growth of the domestic parcel delivery and e-commerce markets, there has been an increasing demand for automated parcel and package acceptance systems. The post office and convenience stores are using kiosks to replace them, but the usage rate of middle-aged and elderly people, who are the main users who are not familiar with digital devices, is low, making it difficult to call it a reasonable automation method. In addition, the existing paper-based receipt system is insufficient to meet modern requirements. Therefore, this paper proposes a system that automatically corrects and digitizes handwritten invoices and package information. The proposed system extracts invoice images that are resistant to optical and physical noise through an image segmentation model, and uses a visual document understanding model to correct address information that has been incorrectly recognized through an LLM-based correction algorithm to output the address. In particular, the system presents strategies for improving and commercializing the overall system by extracting areas of the invoice that were previously inferred using image processing algorithms alone, and by pre- and post-processing display models for handwriting recognition. This improves the performance of systems that process handwritten invoices and enables the detection and correction of errors and typos.
Abstract
Recently, with the growth of the domestic parcel delivery and e-commerce markets, there has been an increasing demand for automated parcel and package acceptance systems. The post office and convenience stores are using kiosks to replace them, but the usage rate of middle-aged and elderly people, who are the main users who are not familiar with digital devices, is low, making it difficult to call it a reasonable automation method. In addition, the existing paper-based receipt system is insufficient to meet modern requirements. Therefore, this paper proposes a system that automatically corrects and digitizes handwritten invoices and package information. The proposed system extracts invoice images that are resistant to optical and physical noise through an image segmentation model, and uses a visual document understanding model to correct address information that has been incorrectly recognized through an LLM-based correction algorithm to output the address. In particular, the system presents strategies for improving and commercializing the overall system by extracting areas of the invoice that were previously inferred using image processing algorithms alone, and by pre- and post-processing display models for handwriting recognition. This improves the performance of systems that process handwritten invoices and enables the detection and correction of errors and typos.
참고문헌 (28)
- K. He, "Mask rcnn", Proceedings of the IEEE international conference on computer vision, pp.2961~2969 (2017)
- R. N. Khobragade, "Challenges in recognition of online and off-line compound handwritten characters: a review", Smart Trends in Computing and Communications: Proceedings of SmartCom 2019, pp.375~383 (2020)
- S. Appalaraju, "Docformer: End-to-end transformer for document understanding", Proceedings of the IEEE, 2021, 993~1003
- Y. Xu, "Layoutlm: Pre-training of text and layout for document image understanding", Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp.1192~1200 (2020)
- K. Lee, "Pix2struct: Screenshot parsing as pretraining for visual language understanding", International Conference on Machine Learning, pp.18893~18912 (2023)
- D. Bolya, "Yolact: Realtime instance segmentation", Proceedings of the IEEE, 2019, 9157~9166
- V. Vuori, "Adaptive Methods for On-Line Recognition of Isolated Handwritten Characters", pp.2002
- M. Ma, "Online recognition of handwritten Korean and English characters", Journal of Information Processing Systems, 8(4), pp.653~668 (2012)
- T. Q. Phan, "Recognizing text with perspective distortion in natural scenes", Proceedings of the IEEE International Conference on Computer Vision, pp.569~576 (2013)
- Z. Zhang, "Multi-oriented text detection with fully convolutional networks", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.4159~4167 (2016)
- A. Gupta, "Synthetic data for text localisation in natural images", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.2315~2324 (2016)
- Y. Huang, "Layoutlmv3:Pre-training for document ai with unified text and image masking", Proceedings of the 30th ACM International Conference on Multimedia, pp.4083~4091 (2022)
- V. P. d’Andecy, "Field extraction by hybrid incremental and a-priori structural templates", 2018 13th IAPR International Workshop on Document Analysis Systems (DAS), pp.251~256 (2018)
- C. Biagioli, "Automatic semantics extraction in law documents", Proceedings of the 10th international conference on Artificial intelligence and law, pp.133~140 (2005)
- E. Francesconi, "Semantic Processing of Legal Texts: Where the Language of Law Meets the Law of Language" (2010), Springer
- J. Y. Lee, "A geocoding method implemented for hierarchical areal addressing system in Korea", Spatial Information Research, 14(4), pp.403~419 (2006)
- J. Voerman, "Evaluation of neural network classification systems on document stream", Document Analysis Systems: 14th IAPR International Workshop, DAS 2020, Proceedings 14, pp.262~276 (2020), Springer International Publishing
- A. Shapenko, "ABBYY: The digitization of language and text", Emerald Emerging Markets Case Studies, 8(2), pp.1~26 (2018)
- F. Cesarini, "Conceptual modelling for invoice document processing", Proceedings of Database and Expert Systems Applications. 8th International Conference, pp.596~603 (1997)
- M. Köppen, "Document Analysis Systems", pp.223~241 (1996), World Scientific
- E. Medvet, "A probabilistic approach to printed document understanding", International Journal on Document Analysis and Recognition, 14(4), pp.335~347 (2011)
- I. Taleb, "Leveraging large language models for enhancing literature-based discovery", Big Data and Cognitive Computing, 8(11), pp.146~ (2024)