[ASAP] Some questions on datasets and raw data.

[revoluzionario]

In the statistics you provided, I see that MKG-W has 15000 entities, MKG-Y has 15000 entities but they respectively have 9203 folders and 14481 folders. Is this result okay? Does it have any problems?

Second, I don't understand what column "image" means, when I checked MKG-W raw dataset and I saw 27589 images, and MKG-Y raw dataset has 44386 images

And the number of texts is not the same with the number I checked through using "Find and Replace" with "dbpedia.net", though the error is negligible.

Hope to receive your answer soon!

[quqxui]

Thank you for your interest in our work.

Firstly, the MKG-W & -Y dataset consists of 15,000 entities, indicating the number of nodes in the knowledge graph. However, due to unavailability, a portion of the entities lacks both image and text data, resulting in an unequal number of image and text in the datasets.

In our published datasets, we provide three images for each entity. However, it's important to note that our method does not consider the learning of multiple images. As mentioned in section 4.3.1 of our paper, we randomly select one image as the representation for each entity.

Therefore, the raw datasets for MKG-W and MKG-Y contain more images than what we used in our paper.

If you have any further questions or need clarification on any aspect, please let me know.