Summary of Pseudo-label Calibration Semi-supervised Multi-modal Entity Alignment, by Luyao Wang and Pengnian Qi and Xigang Bao and Chunlai Zhou and Biao Qin
Pseudo-Label Calibration Semi-supervised Multi-Modal Entity Alignment
by Luyao Wang, Pengnian Qi, Xigang Bao, Chunlai Zhou, Biao Qin
First submitted to arxiv on: 2 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computation and Language (cs.CL); Databases (cs.DB)
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 A novel approach to multi-modal entity alignment (MMEA) is introduced, tackling the challenge of integrating equivalent entities between two multi-modal knowledge graphs. The proposed Pseudo-label Calibration Multi-modal Entity Alignment (PCMEA) method leverages semi-supervised learning, exploiting both labeled and unlabeled data. PCMEA combines various embedding modules and attention mechanisms to extract features from visual, structural, relational, and attribute modalities. Mutual information maximization is used to filter noise and augment commonality between modalities. Pseudo-label calibration with momentum-based contrastive learning improves the quality of pseudo-labels and pulls aligned entities closer. Experimental results on two MMEA datasets demonstrate the effectiveness of PCMEA, achieving state-of-the-art performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Multi-modal entity alignment is a way to connect similar information from different sources. This paper introduces a new method called PCMEA that helps match entities (like people or things) across two types of data: visual and non-visual (like text). The approach combines different techniques to extract useful features from both types of data, while also removing noise and improving the accuracy of matching. The results show that this method performs better than previous methods in this area. |
Keywords
* Artificial intelligence * Alignment * Attention * Embedding * Multi modal * Semi supervised
