dgai91 / pytorch-acnn-model Goto Github PK

Hi, I have carefully read your code and found two bugs that could potentially cause your inferior performance compared with the original paper. 1. in the new_convolution function, there is missing use of self.tanh() as the activation after the convolution layer; 2. in the original paper, the convolution kernel is 1 since the input is already trigram, so there is no need to use kernel size of 3 in the new_convolution function. If you have doubt on my comments, welcome to discuss with me, thanks!

After I clone the repo and also copy some code from the acnn repo, I still couldn't run the code actually.

原论文中在计算矩阵G的时候未写明三个矩阵的size，但是根据原论文中此处引用的2个文献里的计算方式：
1 ABCNN: attention-based convolutional neural network for modeling sentence pairs.
2 Attentive pooling networks
我觉得原文中G的size应该是 n * nr，对应的，R_star * AP的size是(dc,nr) 每行取max后缩小为向量(dc,1)
而你的代码里G的size是 n * dc ，在这个地方，我觉得与我的理解不太一样，欢迎交流，

第二个是，我觉得得到R_star的那个卷积，过滤器宽度设置为1就可以了，

embed_file = 'embeddings.txt'
vac_file = 'words.lst'
这两个文件怎么找啊，分别 是什么文件啊？

mldl@ub1604:/ub16_prj/relation-classification-via-attention-model$ python3.6 acnn_train.py
Traceback (most recent call last):
File "acnn_train.py", line 74, in
train_datasets = D.TensorDataset(data_tensor=train, target_tensor=y_tensor)
TypeError: init() got an unexpected keyword argument 'data_tensor'
mldl@ub1604:/ub16_prj/relation-classification-via-attention-model$

Something wrong with your program ?

DP = 25
DC = 500
N = 123
NP = 123
NR = 19
KP = 0.6
K = 3
请问一下这些参数分别是？

Hi,

In function pos in file data_pro.py, relative distance of words is mapped to [0,123). Why is 123 chosen? Is it related to the maximum sentence length of the data?

Thanks,
Nigel

attention模块的代码为什么没有提交呢？ 另求此模型当前的f1值。

Hi,
I was trying to run the code, but the words.lst file and embedding file is not there and I see words.lst was deleted. I am new to this, could you please tell me how to get these files. Do I have to generate it?

请问这个函数是什么意思？是不是漏写了

有没有同学解释一下数据集是具体怎么是这种格式，每条样例的前面五个参数是什么，还有，是在哪里标注的