Summary of Chartkg: a Knowledge-graph-based Representation For Chart Images, by Zhiguang Zhou et al.
ChartKG: A Knowledge-Graph-Based Representation for Chart Images
by Zhiguang Zhou, Haoxuan Wang, Zhengqing Zhao, Fengling Zheng, Yongheng Wang, Wei Chen, Yong Wang
First submitted to arxiv on: 13 Oct 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Information Retrieval (cs.IR)
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 Chart images, such as bar charts, pie charts, and line charts, are produced at an unprecedented rate due to the widespread use of data visualizations. As a result, knowledge mining from chart images is becoming increasingly important, with potential benefits for downstream tasks like chart retrieval and knowledge graph completion. However, existing methods primarily focus on converting chart images into raw data, often disregarding their visual encodings and semantic meanings, which can lead to information loss. This paper proposes ChartKG, a novel KG-based representation that models the visual elements in a chart image and semantic relations among them, including visual encodings and insights. The framework integrates various image processing techniques to identify visual elements and relations, such as CNNs for classifying charts, yolov5 and optical character recognition for parsing charts, and rule-based methods for constructing graphs. The proposed representation can benefit applications like semantic-aware chart retrieval and chart question answering. Quantitative evaluations assess the two fundamental building blocks of the framework: object recognition and optical character recognition. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Chart images are produced in large numbers due to data visualizations. To get knowledge from these charts, we need new methods that work well. Current methods mainly turn chart images into raw data but forget important details like what things mean visually. This paper proposes a new way to represent chart images using knowledge graphs (KGs). The KG representation can model the main parts of a chart and how they relate to each other. We also developed a general framework that uses image processing techniques like computer vision, optical character recognition, and rule-based methods to turn chart images into this KG representation. This new way of representing charts can help with tasks like finding specific charts and answering questions about them. |
Keywords
» Artificial intelligence » Knowledge graph » Parsing » Question answering
