donalee/BUIR

CiteUlike and Ciao datasets

Closed this issue · 5 comments

chenchongthu commented

Would you please share the CiteUlike and Ciao datasets used in your paper? Thanks!

donalee commented

You can download the datasets from the following links:
CiteULike : https://github.com/changun/CollMetric/tree/master/citeulike-t
Ciao : https://github.com/CQU-CSE/DatasetCollection
Thanks!

chenchongthu commented

Thanks for your reply! Actually I want to reproduce the results reported in your paper. So would you please upload the datasets that exactly the same as those used in your paper? Or would you upload the data preprocessing code?

donalee commented

The data preprocessing code (e.g., filter_interactions) is already included in the Utils/data_utils.py. In my experiments, I used the raw data (in the same format with user.dat of the toy-dataset), and filtered long-tail users and items on-the-fly while loading each dataset from the raw file. Thanks!

chenchongthu commented

Thanks!

Coder-Yu commented

Thanks for your reply! Actually I want to reproduce the results reported in your paper. So would you please upload the datasets that exactly the same as those used in your paper? Or would you upload the data preprocessing code?

Hi Chong,
Have you successfully reproduced the results? I have read the official codes and re-implemented the method with Tensorflow. However, the performance is much lower than the reported. I conducted experiments on some datasets with long-tail users and items, and the results were a bit disappointing.