ETD

• deep learning
• lstm
• neural network
• transduction
• tree
• treelstm

The thesis describes a deep neural network for learning tree-to-tree transductions. The proposed approach uses TreeLSTM models to encode the input as a fixed size vector, followed by a decoding phase that generates the output. Experiments are provided to show how the model can learn structure-to-structure and structure-to-element transductions.