Improving Sequence-to-Sequence Learning via Optimal Transport

Abstract

Sequence-to-sequence models are commonly trained via maximum likelihood estimation (MLE). However, standard MLE training considers a word-level objective, predicting the next word given the previous ground-truth partial sentence. This procedure focuses on modeling local syntactic patterns, and may fail to capture long-range semantic structure. We present a novel solution to alleviate these issues. Our approach imposes global sequence-level guidance via new supervision based on optimal transport, enabling the overall characterization and preservation of semantic features. We further show that this method can be understood as a Wasserstein gradient flow trying to match our model to the ground truth sequence distribution. Extensive experiments are conducted to validate the utility of the proposed approach, showing consistent improvements over a wide variety of NLP tasks, including machine translation, abstractive text summarization, and image captioning.

References in corpus (20)

• Sequence to Sequence Learning with Neural Networks
• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
• Categorical Reparameterization with Gumbel-Softmax
• Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks
• Improved Image Captioning via Policy Gradient optimization of SPIDEr
• MaskGAN: Better Text Generation via Filling in the______
• Professor Forcing: A New Algorithm for Training Recurrent Networks
• Image Captioning with Semantic Attention
• How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
• Generating Informative and Diverse Conversational Responses via Adversarial Information Maximization
• Adversarial Ranking for Language Generation
• Computational Optimal Transport
• Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distance
• Distilled Wasserstein Learning for Word Embedding and Topic Modeling
• A Reinforced Topic-Aware Convolutional Sequence-to-Sequence Model for Abstractive Text Summarization
• Policy Optimization as Wasserstein Gradient Flows
• Hierarchically Structured Reinforcement Learning for Topically Coherent Visual Story Generation
• Sequence Generation with Guider Network
• Natural gradient via optimal transport

Cited by in corpus (7)

• A Survey on Optimal Transport for Machine Learning: Theory and Applications
• Neural Machine Translation: A Review of Methods, Resources, and Tools
• word2ket: Space-efficient Word Embeddings inspired by Quantum Entanglement
• Data Augmentation for Text Generation Without Any Augmented Data
• Adversarial Machine Learning in Text Analysis and Generation
• Improving Adversarial Text Generation by Modeling the Distant Future
• Differentiable Combinatorial Losses through Generalized Gradients of Linear Programs