摘要
为进一步提高识别、提取和应用航海通告信息的效率和准确性,基于对深度学习的自然语言处理技术及中文电子航海通告文本特点的研究,通过语料标注、搭建、调参、训练与评价了5种命名实体识别模型,得出CNN-BiLSTM-CRF模型的识别精确率、召回率等主要指标达到70%以上,能够适应复杂的通告内容及其语境干扰,是目前最优的航海通告命名实体识别模型的结论。应用该模型可为航海通告内容的自动化、智能化提取,高效编制及快速与准确应用提供一种新的技术方法。
In order to further improve the efficiency and accuracy of identifying,extracting and applying marine notice information,based on the research of natural language processing technology of deep learning and the characteristics of Chinese electronic marine notice text,this paper constructs,adjusts,trains and evaluates five named entity recognition models through corpus annotation.It is concluded that the main indicators such as recognition accuracy and recall rate of CNN-BiLSTM-CRF model reach more than 70%,which can adapt to the complex notice content and its context interference,and is the current optimal marine notice named entity recognition model.The application of this model can provide a new technical method for the automation,intelligent extraction,efficient compilation and rapid and accurate application of marine notice content.
作者
黄洋
刘国辉
郭立新
HUANG Yang;LIU Guohui;GUO Lixin(College of Marine Sciences,Shanghai Ocean University,Shanghai 201306,China;Chart Information Center,Tianjin 300450,China)
出处
《海洋测绘》
CSCD
北大核心
2024年第1期79-82,共4页
Hydrographic Surveying and Charting
关键词
航海通告
命名实体
自然语言处理
深度学习
神经网络
marine notices
named entity
natural language processing
deep learning
neural networks
