|
|||||
|
|
| 基于深度学习的生物医学文本分类研究 | |
| 引用本文： | 周永称,崔忠芳,范少萍,安新颖. 基于深度学习的生物医学文本分类研究[J]. 中华医学图书情报杂志, 2019, 28(11): 1-10 |
| 作者姓名： | 周永称 崔忠芳 范少萍 安新颖 |
| 作者单位： | 北京协和医学院/中国医学科学院医学信息研究所,泰康健康产业投资控股有限公司，北京 100020,中国医学科学院医学信息研究所，北京 100020,中国医学科学院医学信息研究所，北京 100020 |
| 基金项目： | 中央级高校基本科研业务费项目“临床医学高层次人才科研能力评价研究”（2018TX63002）；国家重点研发计划“精准医学文本知识网络构建”子课题“精准医学文本语料库构建”（2016YFC0901902-2）；国家自然科学基金项目“面向精准医学的基因-疾病-药物语义关系抽取研究”（71704188） |
| 摘 要： | 目的：探索基于深度学习的文本分类方法在生物医学文本的学科分类中是否具有更好的分类性能。方法：以中国医院科技量值研究中累积的神经病学科、消化病学科、肿瘤学科的SCI论文为数据来源，分别训练并测试CNN、LSTM、LSTM-CNN、LSTM-attention及SVM模型并评估其性能。结果：5类模型中，双层CNN模型的分类性能最好，CNN、LSTM、LSTM-CNN和LSTM-attention模型的分类性能均优于SVM模型。结论：基于深度学习的文本分类方法可提高生物医学文本的学科分类精度，推动医院评价和学科评估的发展。 |
| 关 键 词： | 深度学习；文本分类；医院评价；学科评估 |
| 收稿时间： | 2019-10-12 |
Deep-learning-based biomedical text classification |
|
| ZHOU Yong-cheng,CUI Zhong-fang,FAN Shao-ping and AN Xin-ying. Deep-learning-based biomedical text classification[J]. Chinese Journal of Medical Library and Information Science, 2019, 28(11): 1-10 | |
| Authors: | ZHOU Yong-cheng CUI Zhong-fang FAN Shao-ping AN Xin-ying |
| Affiliation: | Institute of Medical Information, Beijing Union Medical College/Chinese Academy of Medical Sciences, Beijing 100020, China,Taikang Health Industry Investment Proprietary Limited Company, Beijing 100020, China,Institute of Medical Information, Chinese Academy of Medical Sciences, Beijing 100020, China and Institute of Medical Information, Chinese Academy of Medical Sciences, Beijing 100020, China |
| Abstract: | Objective To study whether the classification performance of deep learning-based text classification is better in biomedical text classification. Methods The CNN, LSTM, LSTM-CNN, LSTM-attention and SVM models were trained and tested, and their classification performance was assessed with the cumulated SCI-covered papers on neurology, gastroenterology and oncology in scientific magnitude studies of Chinese hospitals as the data source. Results The classification performance of CNN model ranked first in the 5 models. The classification performance of CNN, LSTM, LSTM-CNN and LSTM-attention models was better than that of SVM model. Conclusion Deep-learning-based text classification can improve the classification accuracy of biomedical text and promote the development of hospital and discipline assessment. |
| Keywords: | Deep learning Text classification Hospital assessment Discipline assessment |
| 点击此处可从《中华医学图书情报杂志》浏览原始摘要信息 | |
| 点击此处可从《中华医学图书情报杂志》下载免费的PDF全文 | |