xcfcode / dhgn Goto Github PK

Dear authors,

I have a question about the data preprocessing step(Heterogeneous Dialogue Graph Construction Process).
So far, I have understood that the result of this step is in the '/data' folder. (e.g., {train/valid/test}.{node_type/edge_type/edge_index}.txt)
However, I cannot find the code to map the input data into this format.

Could you provide the code you used for making this txt files?
I am trying to apply the DHGN model to another dialogue summarization dataset and use it as a baseline in our paper.
It would really be appreciated if you could do so!

Most respectively,
Seungone Kim

torch的cuda版本是10.1 ，然而 pytorch_geometric 版本是10.0

Thanks for your remarkable work!

In your paper, Sec. 2.4, there are four types of edges that exist in the Heterogeneous Dialogue Graph: speak-by, know-by, rev-speak-by and rev-know-by.
And I also find there are three types of edges in the code (data/.edge_types.txt). I also read the data/.edge_index.txt file, and I guess the "speak-by" edge is not present.

I don't know if I understand it wrong. I sincerely hope you can answer it.

生成的摘要全在一个文件，如何对应上原对话