This paper proposes to transform the (head entity, relation) and tail entity into the same feature space through the Pseudo-Siamese network, and calculate the similarity between the two parts in this feature space, embedding vector of entity and relation have been optimization for constraint conditions.In this paper, the triple is regarded as the abstraction of the answer pair in the factual simple question-answering. According to the corresponding relation of the answer, the corresponding relation between the (head entity, relation) and the tail entity in the triple is obtained, and the constraint characteristics of the elements in the triple are modeled. And then by constructing inverse relations to build a new triple, thus the number of training samples is expanded to improve the learning results of the model.
[1]
Tim Weninger,et al.
ProjE: Embedding Projection for Knowledge Graph Completion
,
2016,
AAAI.
[2]
Luca Bertinetto,et al.
Fully-Convolutional Siamese Networks for Object Tracking
,
2016,
ECCV Workshops.
[3]
Yu Hu,et al.
Probabilistic Reasoning via Deep Learning: Neural Association Models
,
2016,
ArXiv.
[4]
Nikos Komodakis,et al.
Learning to compare image patches via convolutional neural networks
,
2015,
2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
[5]
Geoffrey E. Hinton,et al.
Rectified Linear Units Improve Restricted Boltzmann Machines
,
2010,
ICML.