ngl567 / cake Goto Github PK

Sorry but I'm a beginner. I'd like to know how to run with the mode “TransE+MVLP”“TransE+CANS”“HAKE+CAKE”？
Thank you very much

Hi, I'm very interested in your work, and I want to know when your code will be uploaded to GitHub?

Hi there, very interesting and solid work!
I'm very interested in your work and I'm a little confused since there are no concept-level data aligned with the instance-level data (i.e. the instance-of relations between entities and concepts) in the original KGC datasets such as FB15k. So which version of concept data did you use in the ACG part of your work? And did you use the same concept-level data across different datasets or just use the dataset-specific schemas for each one of them? And more specifically, how can I fetch those data online?

Thanks
ROIM

Can you provide the file 'domain2id.txt'?

作者您好，有一个问题想请问一下。我看了一下ent_dom这个数据集，发现对于每个实体对应的concept都有1或2，那么在构造高质量的负样本或者多视角链接预测时，会造成几乎所有的实体都能通过concept的过滤。从而造成引入常识和不引入常识区别不大。所以想请问一下，我的这个结论是否正确，非常感谢。

Hi, I am glad to have read your paper, I am very inspired by your work. However, I have some questions about the experimental results in this paper.
In this paper you say that you reproduced the results of baseline (e.g. TransE, RotatE, HAKE) using their source code and recommended parameters, but I notice that the reproduced results are very different from the results in the original paper, and even your own method is lower than the results in the original paper of baselines. Can you explain why?
The following table shows the results of the baseline from HAKE.