Learning to Compose Words into Sentences with Reinforcement Learning

Abstract

We use reinforcement learning to learn tree-structured neural networks for computing representations of natural language sentences. In contrast with prior work on tree-structured models in which the trees are either provided as input or predicted using supervision from explicit treebank annotations, the tree structures in this work are optimized to improve performance on a downstream task. Experiments demonstrate the benefit of learning task-specific composition orders, outperforming both sequential encoders and recursive encoders based on treebank annotations. We analyze the induced trees and show that while they discover some linguistically intuitive structures (e.g., noun phrases, simple verb phrases), they are different than conventional English syntactic structures.

References in corpus (2)

• Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars
• A Convolutional Neural Network for Modelling Sentences

Cited by in corpus (6)

• A Simple, Fast Diverse Decoding Algorithm for Neural Generation
• On-the-fly Operation Batching in Dynamic Computation Graphs
• Structured Reordering for Modeling Latent Alignments in Sequence Transduction
• BERE: An accurate distantly supervised biomedical entity relation extraction network
• Tensor Product Generation Networks for Deep NLP Modeling
• Leveraging Recursive Gumbel-Max Trick for Approximate Inference in Combinatorial Spaces