.png){kind=icon}
.png){kind=icon}
Return
A two-stage framework for pig disease knowledge graph fusing
.png){kind=icon}
PRE
AI
DOI：10.1016/j.compag.2024.109796.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Pig disease knowledge graphs (KGs) are crucial for the prevention and treatment of pig diseases. Due to the difficulty of knowledge mining in the field of traditional animal husbandry, there is a lack of high-quality KGs of pig diseases. To tackle this issue, a novel two-stage framework for pig disease KG fusing is proposed in this manuscript. In the first stage, a multi-view augmentation method for pig disease KGs is designed. The domain characteristics in the field of pig disease are considered and four valid strategies are utilized for augmenting triples, which not only enriches the pig disease KGs and provides abundant training data for KG embedding. In the second stage, an unsupervised entity alignment method is introduced to match entities. Importantly, the similarities of entity name, relation, attribute, and structure information are learned alternatively to avoid annotating data manually. Extensive experiments on the pig disease datasets and the public dataset MED_BBK_9K demonstrate that the proposed method can achieve state-of-the-art performance, i.e., the multi-view augmentation method improves hits@1 by 0.387 compared with the suboptimal model on the Pig1 dataset, and the entity alignment model outperforms the second-best model by 0.168 in terms of hits@1 on the Pig1_Pig2 dataset.
Keywords:
Pig disease knowledge graph
Knowledge graph embedding
Data augmentation
Entity alignment
Journal
IF:
8.9
Papers:
10.0K
Citations:
4.8W
Organization
No organization information available