While I understood the general structure of the dataset and how to produce most of the elements, I still can't figure out how triples_n files are generated. Can you help me?

Traceback (most recent call last):
File "run.py", line 5, in
trainer = Trainer(seed=2020)
File "/home/wangzhen/program/GraduationProject/SelfKG/model/layers.py", line 315, in init
myset1 = Mydataset(token1.id_features_dict) # dataset
TypeError: init() missing 1 required positional argument: 'adj_tensor_dict'

I have some questions about the experiments of self negative sampling.

1. In Table 4, the results show that removing self negative sampling can hurt the performance. Have you tried to sample negative entities from both the Gx and Gy?
2. Since the alignment accuracy are quite good on the examined datasets, have you tried to reduce the number of conflicts when sampling from another KG using the alignment result?

Hi, @HaoyunHong

I'm very excited at your work of self-supervised learning in EA. According to your paper, SelfKG performs well in DBP15k and DWY100K. These datasets are relatively simple EA datasets, and I wonder how SelfKG performs in more complicated datasets, such as DBP100K or DBP1M. If you ever tried SelfKG in other datasets? Can you offer more experimental results?

best,
Meihao

I can't find run_LaBSE_SSL_DWY.py and run_LaBSE_SSL.py.

When i run the run_LaBSE_SSL.py, ModuleNotFoundError: No module named 'script.preprocess.deal_fasttext'.

I just visited the paper link
but it is unavailable
Thank you for this great work

Sorry, i can‘t expression language clearly in English. Also, After download the data "LaBSE.zip", when i unzip it， The system tells me "pytorch_model.bin" is broken.

No module named 'model.layers_LaBSE_SSL',your model file don't have layers_LaBSE_SSL.py. Please sure the run_LaBSE_SSL.py can run. Thanks!

恭喜大佬喜中best paper。不读个10遍对不起best paper。

Thank you for this great work
I cant find the paper from the paper link
Could you help me
Thanks again

Thanks for your great work. When I run your model on different datasets you provided, I found there are maybe some missing files and corrupted file.

For DBP15K fr_en and DWY100K dbp_yg, some files seem corrupted and unreadable. There are errors like (both DBP15K fr_en and DWY100K dbp_yg):
Traceback (most recent call last):
File "run_LaBSE_neighbor.py", line 5, in
trainer = Trainer(seed=37)
File "SelfKG/model/layers_LaBSE_neighbor.py", line 193, in init
loader1 = DBP15KRawNeighbors(self.args.language, "1")
File "SelfKG/loader/DBP15KRawNeighbors.py", line 16, in init
self.load()
File "SelfKG/loader/DBP15KRawNeighbors.py", line 22, in load
self.id_entity = pickle.load(f)
_pickle.UnpicklingError: invalid load key, '\x00'.

For DWY100K dataset, some files may be missing, there is error like:
No such file or directory: 'SelfKG/data/DWY100K/dbp_wd/valid.ref'

你好，我注意到数据集文件中只提供实体对应的标注序号

(xxxx) [xx@xxx xxxx]$ head -5 /xxxx/data/DWY100K/dbp_yg/id_ent_1
0       Simeon Burt Wolbach
1       Peter Kyros
2       Yoshihito Fujita
3       Angus M. Woodbury
4       Avianca Perú

而关系类型的标注文件没有提供，无法了解两个实体之间的关系

(xxxx) [xx@xxx xxxx]$ head -5 /xxxx/selfkg/data/DWY100K/dbp_yg/triples_1
11495   75     23339
24003   288     157884
7342    268     39411
23111   216     67889
150614  230     27619

例如对三元组(11495,75,23339)，75对应的关系没有对应的文件指出。

请问能提供DWY100K、和DBP15K 关系对应的序号文件吗？
期待回复😁！

您好，我注意到你们代码里提供的初始嵌入向量文件 ./data/DBP15K/zh_en/raw_LaBSE_emb_1.pkl 与./data/DBP15K/zh_en/raw_LaBSE_emb_2.pkl 包含5244个重合的向量，感觉初始嵌入向量的重合率有点大，请问这是 LaBSE 在未使用翻译技术的前提下得到的嵌入向量，还是使用了翻译技术后得到的嵌入向量呢？如果是后者，请问可以提供未使用翻译技术的嵌入向量吗？
另外，请问可以提供不使用翻译技术、嵌入模型为 FastText 的初始嵌入向量吗？
期待您的回复！谢谢！