ID-CNN-CWS

ID-CNN-CWS copied to clipboard

Question about the embedding vector.

1

Hi~ I have successfully run your wonderful code. But I want to know more details about embedding character (5103 vectors) What is the source of these characters and have you...

deardeerluluu

1、在进入空洞卷积之前为什么有一个普通的二维卷积？是必须的吗？意义何在？ 2、一般普通的在文本上的卷积输入一本是[batch,seq_len,char_dim,1],为什么在这里这个二维卷积的输入是[bacth,1,seq_len,char_dim]?

zc-Chao

KeyError in tf1.13.1

2

``` File "train.py", line 41, in main logger.info('\n'.join(sorted(["%s : %s" % (str(k), str(v)) for k, v in FLAGS.__dict__['__flags'].items()]))) KeyError: '__flags' ```

snakecy

ub16c9@ub16c9-gpu:/media/ub16c9/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/ID-CNN-CWS$ ll result/all/001/ total 142260 drwxrwxr-x 2 ub16c9 ub16c9 4096 Jun 9 00:09 ./ drwxrwxr-x 3 ub16c9 ub16c9 4096 Dec 7 2018 ../ -rw-rw-r-- 1 ub16c9 ub16c9 73 Jun 8...

SeekPoint

Results of the experiment cannot be reproduced on PKU, trainning stopped after 1 iteration

2

我们的实验环境 : Python 3.5 Tensorflow 1.4.0 4x Nvidia Geforce 1080Ti 对代码进行了小幅调整以便在我们的环境上运行 第一个实验测试的是Sighan2005-PKU 但是，无论BI-LSTM-CRF还是ID-CNN-CRF 每次训练完第1个迭代后，验证集、测试集都只有0.58-0.62 然后提示 'Score too low, break to save time' 退出。 同时，用同样脚本分割的语料训练 foolNLTK（https://github.com/rockyzhengwu/FoolNLTK）， 几乎一样的BI-LSTM-CRF第一轮训练后验证集有0.88的F1值。 初步怀疑评测脚本有问题，验证集、测试集分数一致

TylunasLi

tsv_to_tfrecords.py中naacl-data.tsv文件作用是什么，存放了什么数据，能给个例子吗？

wwn199219

function "preprocess" in file "convert_corpus.py"

1

after "text=preprocess(text)"; some Chinese character change to garbled. such as "充满" to "？满"， “？” for garbled. Is there something wrong? I think this function is to normalize all num to...

youngornever

paper

4

您好，感激您的分享，关于这个项目中提及的论文《Iterated Dilated Convolutions for Chinese Word Segmentation》，我在网上没有找到，请问您这边可以分享一份吗？谢谢！

ligaoyi101