Wu Kun
Wu Kun
text2sql任务的数据框架已开源～
ERNIE的最大文本输入的长度只能在1到512之间，我认为你需要去掉一些更低置信度的value候选，保证ernie输入长度小于512才能继续预测
第一个问题：我看了一下，你发现的问题是存在的，你说的也是对的。 第二个问题：按照我的理解，abc的2-gram-list是['a', 'b'],拼接也是将['a', 'b']拼接成ab。 第三个问题我排查以后再和您讨论一下
> 您好，想问一下您训练NER模型指标能到多少，我这里训练出来在测试集上的指标f值只有0.22，不知道是不是哪里搞错了 我觉得最大可能是数据预处理结果出错了
> In Question Classification process: > > 1. where is the train data for the model "Question_classification/BERT_LSTM_word"? > 2. how to use the code in "PreScreen" directory? I update the...
可以提示一下哪个文件吗
您好，这是我们相关的中文论文 http://jcip.cipsc.org.cn/CN/abstract/abstract3196.shtml
I have not ran the codes,but I wanna know how many available data(subject mention could be found in the question) in the train(75910)&test(21678) can I get after preprocessing.Would you mind...
Well, it is a dependency parsing api. I tried to use dependency parsing to make a difference, but failed. Actrually you can just ignore bert-bilstm-word model and DO NOT SET...
> maybe you can simplify main() function in "Question_classification/BERT_LSTM_word/main.py", it contains too much unused code for a simple classification model. yes, you are right. I update the detail of readme....