基于多标签CRF的疾病名称抽取

发布时间：2018-08-30 12:48

【摘要】：生物医疗文本中的命名实体识别对于构建和挖掘大型临床数据库以服务于临床决策具有重要意义,而其中一个基础工作是疾病名称的识别。医疗文本中存在大量的复合疾病名称,难以分离抽取出其中的实体。针对这一问题,提出一种基于多标签的条件随机场算法,首先对数据标注多层标签,每层标签针对复合疾病名称中的不同疾病,然后用整合后的最终标签去训练模型,最后再对模型预测的标签进行分离。此方法能够识别传统条件随机场算法无法识别的复合疾病名称,实验结果验证了所提算法的有效性。
[Abstract]:The identification of named entities in biomedical texts is of great significance for constructing and mining large clinical databases to serve clinical decisions, and one of the basic tasks is the recognition of disease names. There are a large number of complex disease names in medical texts, so it is difficult to separate and extract the entities. In order to solve this problem, a conditional random field algorithm based on multi-label is proposed. Firstly, the data is labeled with multi-layer label, each layer label is aimed at different diseases in the name of complex disease, and then the model is trained with the integrated final label. Finally, the label of model prediction is separated. This method can recognize the complex disease names which can not be recognized by the traditional conditional random field algorithm. The experimental results show that the proposed algorithm is effective.
【作者单位】：武汉大学计算机学院;
【分类号】：TP391

【相似文献】