Researchers at Alibaba, Zhejiang University and the University of Zurich have developed a knowledge graph embedding (KGE) model called CrossE. The model overcomes the earlier challenges associated with knowledge graphs— the inability to learn from existing triples without additional help. The CrossE model overcomes the challenge by learning multiple triple-specific embeddings for each graph entity. In addition, the model proved to be able to provide reliable explanations for its choices in tests with complex datasets.
In knowledge graph triples, crossover interactions are effects passing either from entities to relations or from relations to entities. Importantly, these effects are bi-directional and hence impact both entities and relations simultaneously. Unfortunately, the earlier models overlooked this aspect and attempted to capture the information described by these effects using general embeddings or multiple separately learned embeddings.
CrossE’s researchers sought to improve this aspect of the earlier models. The researchers developed a novel KGE model that learns one general embedding for each entity and relation and generates multiple triple-specific interaction embeddings through a relation-specific interaction matrix. Furthermore, CrossE uses an embedding-based path-searching algorithm to provide explanations for the predictions offered by KGE.
To evaluate CrossE, researchers applied WN18, FB15k, and FB15k-237 datasets in a link prediction task, followed by a test of its ability to generate explanations for predicted triples. Compared with seven baseline models, CrossE achieved similar results with the WN18 data in some metrics while outperforming in others. With the more challenging FB15k and FB15k237 data, CrossE demonstrated state-of-the-art performance. Thereby indicating the consideration of crossover interactions improves its ability to encode diverse relations in knowledge graphs. Simultaneously, it outperformed baselines’ ability to provide supported explanations for their predictions.
Click here to read the article.
Please give your feedback on this article or share a similar story for publishing by clicking here.
