Summary of Inference Over Unseen Entities, Relations and Literals on Knowledge Graphs, by Caglar Demir et al.
Inference over Unseen Entities, Relations and Literals on Knowledge Graphs
by Caglar Demir, N’Dah Jean Kouagou, Arnab Sharma, Axel-Cyrille Ngonga Ngomo
First submitted to arxiv on: 9 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a new approach to address the limitations of existing knowledge graph embedding models, which can only reason over seen entities, relations, and literals in the transductive setting. The authors introduce the attentive byte-pair encoding layer (BytE) to construct triple embeddings from subword units of entities and relations. BytE enables massive feature reuse via weight tying, reducing the size of embedding matrices. Experimental results show improved link prediction performance on datasets where syntactic representations are semantically meaningful. However, benefits dissipate when entities and relations are represented with plain numbers or URIs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about finding a way to make knowledge graph embedding models work better in real-world situations. Right now, these models can only handle information they’ve seen before, which isn’t very helpful for dynamic knowledge graphs that change over time. The authors came up with an idea called BytE, which helps the model learn from smaller pieces of information instead of just big chunks. This makes it better at predicting links between things and answering questions about the graph. It works well when the information is represented in a specific way, but not as well when it’s just plain numbers or names. |
Keywords
» Artificial intelligence » Embedding » Knowledge graph
