Negative sampling (NS) is widely used in knowledge graph embedding (KGE), which aims to generate negative triples to make a positive-negative contrast during training. However, existing NS methods are unsuitable when multi-modal informa- tion is considered in KGE models. They are also inefficient due to their complex design. In this paper, we propose Modality- Aware Negative Sampling (MANS) for multi-modal knowledge graph embedding (MMKGE) to address the mentioned problems. MANS could align structural and visual embeddings for entities in KGs and learn meaningful embeddings to perform better in multi-modal KGE while keeping lightweight and efficient. Empirical results on two benchmarks demonstrate that MANS outperforms existing NS methods. Meanwhile, we make further explorations about MANS to confirm its effectiveness.
2024-02We preprint our Survey Knowledge Graphs Meet Multi-Modal Learning: A Comprehensive Survey [Repo].
The KGE model:
Modality-Aware Negative Sampling:
Our implemention is based on OpenKE. You should build Python dependencies based on the initialization steps in OpenKE.
The visual embeddings of FB15K and DB15K datasets can be download from the GoogleDrive. To run the code, you should download the embeddings and place them in visual/
You can run the shell scripts placed in scripts/ to run the negative sampling methods proposed in MANS. For example:
- MANS-V: run_visual_fb15k.sh
- MANS-T: run_twostage_fb15k.sh
- MANS-H: run_hybrid_fb15k.sh
- MANS-A: run_adaptive_fb15k.sh
Note that you can open the .sh file for parameter modification.
