aarontengdechuan/mlwa-chinese-slu

Open source code for ICASSP 2021 Paper “Injecting Word Information with Multi-Level Word Adapter for Chinese Spoken Language Understanding”

mlwa-chinese-slu's Issues

请问ecdt的测试集和验证集可以公开吗

请问应该使用什么参数才可以得到你的论文复现结果呢

-dd data/cais_bies-True_token-hanlp -sd test -bs 16 -dr 0.3 -ced 128 -wed 128 -ehd 512 -aod 128 -sed 32 -sdhd 64
使用这种参数得到的效果是Accepted performance: (0.8764241893076249, 0.9466403162055336, 0.8527667984189723) at test dataset;

你好，论文中的结果是使用了词向量的结果，还是未使用词向量的结果呢

实验结果是使用了词向量的结果还是随机的呢

Dot product or element wise product of Vc and Vw?

Hi, it is a good job! Though I have a small question.

Seems that you use element wise product of Vc and Vw in the class BilinearLayer(nn.Module):.

While in Figure1 (c) in the paper, it is described as dot product?

Is my understanding of the details wrong? Wait for your reply, thanks~

请问ECDT-NLU数据集中的1033测试数据在哪里呀，从会议官网只能找到训练集

如题

数据集处理的问题

作者您好，我看到您好像用到了SMP这个数据集，请问您有处理数据的代码可以分享吗？因为我想尝试一下其他的json格式的意图识别数据集，但是不知道怎么能处理成和您一样的数据。

the result of the run with BERT

Hi, what is the result of the run with BERT?

igaword_chn.all.a2b.uni.ite50.vec和ctb.50d.vec是需要自己下载吗

如题，预训练词向量缺失，是需要自己下载吗？

When will you make the codes available?

I have read your paper. I found it interesting and cannot wait to try it out. I noticed that you mentioned in the paper you would make the datasets and codes available here, but it has been two months since the first commit, so do you still plan to upload them here? I am looking forward to them :-)