申晨,林鸿飞.基于图嵌入的社交媒体药物不良反应事件检测方法[J].,2020,60(5):547-554 |
基于图嵌入的社交媒体药物不良反应事件检测方法 |
Detection method of adverse drug events from social media based on graph embeddings |
|
DOI:10.7511/dllgxb202005011 |
中文关键词:社交媒体图嵌入对抗训练药物不良反应事件检测深度学习 |
英文关键词:social mediagraph embeddingsadversarial trainingadverse drug event detectiondeep learning\@ |
基金项目:国家自然科学基金资助项目(6177210361572102). |
|
摘要点击次数:261 |
全文下载次数:168 |
中文摘要: |
药物不良反应事件是造成患者发病、死亡的主要原因之一.传统的基于患者自发报告系统存在较为严重的漏报情况,近年来将推特等社交媒体作为数据来源进行药物不良反应事件检测的研究愈发受到重视.各种深度学习模型通常依赖于大量的训练样本,然而受限于数据来源的特质和耗时的数据标注工作,该领域的相关研究面临标注数据规模小、数据噪声大等问题,制约了这些模型发挥良好的效果.据此,在文本表示层面引入基于图嵌入数据增强和对抗训练两种正则化方法,提升模型在低资源高噪声下的药物不良反应事件检测效果.通过实验,具体分析和讨论两种方法的适用范围,结合卷积神经网络,提出一种同时发挥其优势的药物不良反应事件检测模型,实验结果显示其具有良好的适用性. |
英文摘要: |
Adverse drug event is a major cause of morbidity and mortality of hospitalized patients. Since the traditional spontaneous self reporting system is experiencing serious underreporting issues, and the research on the use of social media, such as Twitter, as a data source for the detection of adverse drug events has received increasing attention in recent years. Deep learning models usually require many training samples. However, due to the characteristics of user generated contents and time consuming data annotation process, related research is faced with the problems caused by small scale annotation but highly noisy datasets, which restricts deep learning models to achieve good results. Accordingly, two regularization methods are introduced at the representation level, which are graph embedding based data augment and adversarial training, to improve the performance of detecting adverse drug events under such condition. The applicable scopes of these two methods are analyzed and discussed through experiments. Combining with the convolutional neural network, an adverse drug event detection scheme that can make full use of the two methods is proposed, and the experimental results testify the feasibility. |
查看全文查看/发表评论下载PDF阅读器 |
| --> 关闭 |