nlpdz/Medical-Named-Entity-Rec-Based-on-Dilated-CNN

关于识别的准确率?

Opened this issue · 0 comments

你好,我看到你提供的服务测试地址的效果非常的好,但是我利用你的方法拉起服务后, 效果较差,
{"entities": [{"start": 0, "end": 2, "type": "FW", "word": ""乙"}, {"start": 0, "end": 3, "type": "CL", "word": "肝"}, {"start": 3, "end": 4, "type": "SYM", "word": "和"}, {"start": 5, "end": 6, "type": "SYM", "word": "他"}], "string": ""乙肝和其他问题“"},
请问这是不是很大可能是我们的embedding, 分词 不一样的原因?
谢谢