微型电脑应用

现有电子病历命名实体识别方法由于引入外部专业知识,导致计算资源需求大幅度上升,对此,提出一种基于知识增强的中文电子病历命名实体识别模型。使用基于实体关联的知识增强模型(EAGBERT)对输入的电子病历文本进行编码,识别实体名词,通过实体名词与外部知识三元组的关联,进一步拼接生成包含外部知识的语料,并加入全局词语边界特征进行训练生成词向量。将词向量输送到双向长短期记忆(BiLSTM),向其添加注意力机制以更准确地提取上下文信息,使用条件随机场(CRF)输出最佳序列。实验表明,文章对中文电子病历命名实体识别的改进有效,为电子病历信息自动化处理的应用提供了有益的贡献。

The existing named entity recognition methods for electronic medical records have led to a significant increase in the demand for computing resources due to the introduction of external professional knowledge.In response to this,a named entity recognition model for Chinese electronic medical records based on knowledge enhancement is proposed.The input electronic medical record text is encoded using entity association generation-based knowledge BERT(EAGBERT) to identify entity nouns.The entity nouns are associated with external knowledge triples,further concatenated to generate corpora containing external knowledge,and global word boundary features are added for training to generate word vectors.Feed the word vectors to bi-directional long short-term memory(BiLSTM),add an attention mechanism to it to extract context information more accurately,and use conditional random field(CRF) to output the best sequence.Experiments show that the improvement of named entity recognition in Chinese electronic medical records in this article is effective,providing a beneficial contribution to the application of automatic processing of electronic medical record information.