search

liuhuanyong / MedicalNamedEntityRecognition

Medical Named Entity Recognition implement using bi-directional lstm and crf model with char embedding.CCKS2017中文电子病例命名实体识别项目,主要实现使用了基于字向量的四层双向LSTM与CRF模型的网络.该项目提供了原始训练数据样本(一般醒目,出院情况,病史情况,病史特点,诊疗经过)与转换版本,训练脚本,预训练模型,可用于序列标注研究.把玩和PK使用.
423 stars 195 forks source link

loss和accuracy #10

Open wangjiaxu opened 4 years ago

wangjiaxu commented 4 years ago

有知道为什么结果的loss很大，都是十几呢？正确率却挺高的。

loss: 17.5805 - crf_viterbi_accuracy: 0.9911 - val_loss: 15.6539 - val_crf_viterbi_accuracy: 0.8773 loss: 17.5799 - crf_viterbi_accuracy: 0.9915 - val_loss: 15.6227 - val_crf_viterbi_accuracy: 0.8833 loss: 17.5801 - crf_viterbi_accuracy: 0.9915 - val_loss: 15.6512 - val_crf_viterbi_accuracy: 0.8796 loss: 17.5808 - crf_viterbi_accuracy: 0.9909 - val_loss: 15.6785 - val_crf_viterbi_accuracy: 0.8777

houruihui commented 4 years ago

训练完后在测试的时候结果比较不理想

houyuchao commented 2 months ago

训练完后在测试的时候结果比较不理想

您好我运行预测文件时候遇到了这个问题怎么解决呢，改了半天改不出来 image