In deep learning, a graph is represented as $G=(V,E)$, where $V=\{v_1,v_2,..,v_N\}$  is the set of nodes, and 	$E\subseteq V\times V$ is the set of edges between nodes. $N=|V|$ and $M=|E|$. $\mathbf{A}$ is the adjacency matrix of this graph.

Graphs Representation In Deep Learning

GNNs that build word/entity graphs for reasoning can be used


Columbia University

- Recursive neural networks
- Convolutional neural networks 
- Recurrent neural networks 
- Graph neural networks 
- Attention-based neural networks


Architectures explored for NRE


Recursive neural networks learn compositional representations for sentences recursively. Recursive neural networks based on augmented dependency paths can be used for relation extraction in order to integrate syntactic information. 


Recursive neural networks for NRE


CNNs effectively model local textual patterns and can be adopted over shortest dependency paths in order to integrate syntactic information.


Convolutional neural networks for NRE


RNNs can better handle long sequential data and can be adopted over shortest dependency paths in order to integrate syntactic information. Deep RNNs make further use of dependency paths as well. 


Recurrent neural networks for NRE


Graph neural networks for NRE


Attention-based neural networks for NRE use attention mechanism to aggregate global relational information 


Learn Before

Related

Learn After