Summary of Start From Zero: Triple Set Prediction For Automatic Knowledge Graph Completion, by Wen Zhang et al.
Start from Zero: Triple Set Prediction for Automatic Knowledge Graph Completion
by Wen Zhang, Yajing Xu, Peng Ye, Zhiwei Huang, Zezhong Xu, Jiaoyan Chen, Jeff Z. Pan, Huajun Chen
First submitted to arxiv on: 26 Jun 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 novel graph-level automatic knowledge graph (KG) completion task called Triple Set Prediction (TSP), which aims to predict a set of missing triples given a set of known triples, without assuming any elements in the missing triple are provided. To evaluate this new task, four evaluation metrics are proposed, including three classification metrics and one ranking metric, considering both partial-open-world and closed-world assumptions. The paper also presents a novel subgraph-based method GPHT to efficiently predict the triple set. RuleTensor-TSP and KGE-TSP are proposed as baselines applying existing rule- and embedding-based methods for TSP. The proposed methods are evaluated on two datasets extracted from Wikidata, with results showing that they can successfully generate missing triples and achieve reasonable scores. GPHT performs better than baselines with significantly shorter prediction time. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper proposes a new way to fill in missing information in a knowledge graph. Instead of starting with some pieces of the puzzle, it tries to predict an entire set of missing relationships. The authors develop four ways to measure how well this task is done and test their methods on two datasets from Wikidata. They also create a special dataset for testing purposes. The results show that their methods can do this task successfully, and one method works better than others with much less time spent. |
Keywords
» Artificial intelligence » Classification » Embedding » Knowledge graph
