Qeustions about reproduced results on flat NER

ShannonAI / mrc-for-flat-nested-ner

Code for ACL 2020 paper `A Unified MRC Framework for Named Entity Recognition`

663 stars 118 forks source link

Qeustions about reproduced results on flat NER #38

Open berryxue opened 4 years ago

berryxue commented 4 years ago

您好！非常感谢您提供了MRC for NER思路，请问您所用的suquentialsampler对最后的结果影响有多大？在ontonotes数据集上您设置的最大句子长度为100，但在这个数据集的train, test, dev中句子长度超过90的非常多，请问您是怎么处理这种情况的呢？对比模型设置的最大句子长度也是100吗？期待您的回复！

leila-iust commented 3 years ago

in this repository i could not find any code and script about flat ner. how did u find these parameters?