Knowledge Graph Embedding for Data Mining vs. Knowledge Graph Embedding for Link Prediction - Two Sides of the same Coin?

Tracking #: 2892-4106

Authors: 
Jan Portisch
Nicolas Heist
Heiko Paulheim

Responsible editor: 
Guest Editors DeepL4KGs 2021

Submission type: 
Full Paper
Abstract: 
Knowledge Graph Embeddings, i.e., projections of entities and relations to lower dimensional spaces, have been proposed for two purposes: (1) providing an encoding for data mining tasks, and (2) predicting links in a knowledge graph. Both lines of research have been pursued rather in isolation from each other so far, each with their own benchmarks and evaluation methodologies. In this paper, we argue that both tasks are actually related, and we show that the first family of approaches can also be used for the second task and vice versa. In two series of experiments, we provide a comparison of both families of approaches on both tasks, which, to the best of our knowledge, has not been done so far. Furthermore, we discuss the differences in the similarity functions evoked by the different embedding approaches.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Accept

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Angelo Salatino submitted on 17/Oct/2021
Suggestion:
Accept
Review Comment:

The paper looks good to me now.

The only thing, but I am sure the publisher will say the same, please make Figure 2 as a table.

You can use tools like https://www.tablesgenerator.com.

Review #2
Anonymous submitted on 18/Oct/2021
Suggestion:
Accept
Review Comment:

The paper has largely improved and I appreciate the efforts made by the authors to address my concerns. The paper has a clear theme, the performed experiments sound. I have few suggestions which can further improve the paper, however, I do not see them as necessary for its publication.

“... ignoring the different types of relations and treating all edges the equally” does not seem correct.
In the last section of the related work, comparisons between KGE embeddings from previous works are reported. However, I would suggest including the reasons on why a model was better than another one. Now the discussion only reports A outperforms B, without giving insights or rationales to the reader.

Review #3
Anonymous submitted on 26/Oct/2021
Suggestion:
Accept
Review Comment:

The authors have addressed my comments. I think it's good to accept the paper.

One minor thing, the resolution of the figures is still a bit grainy so it would be nice to try and fix that for the final version.