Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking.
Michael Sejr SchlichtkrullNicola De CaoIvan TitovPublished in: CoRR (2020)
Keyphrases
- neural network
- weighted graph
- undirected graph
- natural language processing
- disjoint paths
- edge weights
- vertex set
- directed graph
- genetic algorithm
- artificial neural networks
- graph representation
- wordnet
- graph matching
- graph structure
- loss function
- random walk
- back propagation
- graph model
- directed acyclic graph
- edge detection
- strongly connected
- bipartite graph
- empirically derived
- multi layer
- graph theory
- feed forward
- human visual system
- neural network model
- structured data
- text mining
- natural language
- free text
- multilayer perceptron
- connected components
- edge information
- question answering
- natural images
- information extraction
- pattern recognition
- learning algorithm