Contextual Parameter Generation for Knowledge Graph Link Prediction
AAAI-2020
Reference Format
:
GB/T7714
Stoica G , Stretcu O , Platanios E A , et al. Contextual Parameter Generation for Knowledge Graph Link Prediction[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(3):3000-3008.
Datasets
:
UMLS、Kinship、WN18RR、FB15k-237、NELL-995
Problems
:
- Recent approaches significantly limits representational power. We treat relations as the context in which source entities are processed to produce predictions, by using relation embeddings to generate the parameters of a model operating over source entity embeddings.
- Entity and relation representations are combined in a way that only allows for additive interactions between them (e.g., they may be concatenated and then projected using a linear transformation).
Results
:
可参考点
:
- 现阶段连接预测的任务大致可以分为两类:单跳与多跳。其中单跳又可以分为:基于Trans的系列模型、基于卷积神经网络的系列模型。多跳则主要基于路径排序算法。
- 大部分连接预测模型可以分为4步:
其中Z是实体与关系融合后的表示,f()是预测函数,最终输出是在所有目标实体上的一个概率分布或者是针对当前三元组的分类(真/伪)结果。 - 可以通过Toy Case的方式阐述自己模型的优劣性。
原文与Code已上传,主页可Down