Learning to Compose Words into Sentences with Reinforcement Learning
international conference on learning representations, 2017.
We demonstrated the benefit of learning task-specific composition order on four tasks: sentiment analysis, semantic relatedness, natural language inference, and sentence generation
We use reinforcement learning to learntree-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 orpredicted using supervision from explicit treebank annotations,the tree structures in this work are optim...More
PPT (Upload PPT)