摘要
针对标准文件中图片文本扫描不清晰、文本自动化识别准确率不高等问题设计了针对标准文件的跨模态的图片文本命名实体识别模型O2BSC(OCR-BERT-Bi-LSTM-CRF),重点介绍了模型结构及模型处理流程,并通过实验证明了该模型在命名实体识别上具有一定优势,相关识别效果更佳,从而为标准化实施工作及标准文件内容解读提供了一种有效的能力支撑。
Addressing issues such as unclear scanning in images and text in standard documents,and low accuracy of automated text recognition,a cross modal image text named entity recognition model O2BSC(OCR-BERT Bi LSTM CRF)was designed to address the issues of unclear Chinese font scanning and low accuracy in automated text recognition in standard files.The model structure and processing flow were highlighted,and experiments were conducted to demonstrate that the model has certain advantages in named entity recognition,with better recognition performance.This provides an effective capability support for standardization implementation and interpretation of standard document content.
出处
《信息技术与标准化》
2025年第7期49-55,共7页
Information Technology & Standardization
关键词
标准文件
图片文本
命名实体识别
跨模态
关键词提取
standard documents
image text
named entity recognition
cross modal
keyword extraction
