How to produce dataset about selfkg HOT 5 CLOSED

Thanks for your attention. As we mentioned in our paper, the DWY100K dataset we use is built by

@inproceedings{sun2018bootstrapping,
  title={Bootstrapping Entity Alignment with Knowledge Graph Embedding},
  author={Sun, Zequn and Hu, Wei and Zhang, Qingheng and Qu, Yuzhong},
  booktitle={ IJCAI, vol.18},
  pages={4396--4402},
  year={2018}
}

and the DBP15k we used is built by

@inproceedings{JAPE,
  title={Cross-lingual entity alignment via joint attribute-preserving embedding},
  author={Sun, Zequn and Hu, Wei and Li, Chengkai},
  booktitle={ISWC},
  pages={628--644},
  year={2017},
  organization={Springer}
}

The dataset we provided is based on the original datasets mentioned above : )

from selfkg.

According to BootEA papers

We then extracted all the triples that only involve the entities in the alignment.

I understood it like this: they extract all entities (e1, e2) that connect directly to entity e.

from selfkg.

A triple is like (e1, r, e2), where r is the relation between entity e1 and entity e2.

I think this sentence:

We then extracted all the triples that only involve the entities in the alignment.

means that if a triple contains an entity e_a in Knowledge Graph KG_a that has an equivalent entity e_b in the other Knowledge Graph KG_b, that means e_a is in the alignment, then they extract this triple into their dataset.

from selfkg.

In their dataset, they extract concepts of different themes. Therefore it is possible for them to extract entities and relationships between them. My case is a bit different. I only need to align entities of one theme (e.g Airports), and it's impossible to find an Aiport that is 1 op (relationship) away to another airport. My training set thus contains non-Airport entities, will it affect my results?

from selfkg.

Sorry, your question is not very clear to me and is not related to this project. Maybe others can help you with your question : )

from selfkg.