taoqi98/pp-rec

Codes of the PP-Rec method published in ACL 2021.

pp-rec's Issues

predict_generator function

Hi there,
I had the chance to review your paper, and I must say, it's truly captivating!

While going through the general logic presented in the paper's accompanying Jupyter notebook, I noticed the usage of the predict_generator function.

However, while examining the files, it appears that this function does not exist.

Could you kindly provide an explanation of how the predict_generator function works? More specifically for the following code line

news_scoring = news_encoder.predict_generator(news_generator,verbose=True)

Best regards.

upload source code

Just remind you to upload the source code of PP-Rec published on ACL 2021

the paper dataset

Dear the author, could you provide the paper dataset?? If you can, thank you so much!

How to build file

Great works!
I want to ask some questions about build some specific files:
In NewsContent.py,
we have mergeimps.tsv:

and V21UrlDocs22_title_josn.tsv:

I want to ask how to build that files, from MIND Dataset.
One more, can you show me one sample of each text file, so i can rebuild for myself.
Many thanks

Request for data specifics

Dear author(s),

I, and two other fellow MSc Artificial Intelligence students, would love to reproduce your experiments. Seeing the previous issues around the datasets, do you mind sharing the MSN and the Feeds datasets now? Or do you recommend using https://msnews.github.io/index.html still? Thanks for considering our request!

Best regards,
Ivo

format about dataset

Great work! I have some question about Dataset.
In UserContent.py,

In the function about load_session, As I understand it, sessions are built by reading the user's history of clicks, What do pos and neg stand for? What is the list after split?

=====================================
In addition, in preprocessing.py

multi clicks were split, but I cant understand what happend about pos and neg.
Can you give me an example of understanding?
Thanks!