Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Hao, Morimura, Tetsuro, Honda, Ukyo, Kawahara, Daisuke
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2405.01280
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913413139005440
author Wang, Hao
Morimura, Tetsuro
Honda, Ukyo
Kawahara, Daisuke
author_facet Wang, Hao
Morimura, Tetsuro
Honda, Ukyo
Kawahara, Daisuke
contents Non-autoregressive (NAR) language models are known for their low latency in neural machine translation (NMT). However, a performance gap exists between NAR and autoregressive models due to the large decoding space and difficulty in capturing dependency between target words accurately. Compounding this, preparing appropriate training data for NAR models is a non-trivial task, often exacerbating exposure bias. To address these challenges, we apply reinforcement learning (RL) to Levenshtein Transformer, a representative edit-based NAR model, demonstrating that RL with self-generated data can enhance the performance of edit-based NAR models. We explore two RL approaches: stepwise reward maximization and episodic reward maximization. We discuss the respective pros and cons of these two approaches and empirically verify them. Moreover, we experimentally investigate the impact of temperature setting on performance, confirming the importance of proper temperature setting for NAR models' training.
format Preprint
id arxiv_https___arxiv_org_abs_2405_01280
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Reinforcement Learning for Edit-Based Non-Autoregressive Neural Machine Translation
Wang, Hao
Morimura, Tetsuro
Honda, Ukyo
Kawahara, Daisuke
Computation and Language
Non-autoregressive (NAR) language models are known for their low latency in neural machine translation (NMT). However, a performance gap exists between NAR and autoregressive models due to the large decoding space and difficulty in capturing dependency between target words accurately. Compounding this, preparing appropriate training data for NAR models is a non-trivial task, often exacerbating exposure bias. To address these challenges, we apply reinforcement learning (RL) to Levenshtein Transformer, a representative edit-based NAR model, demonstrating that RL with self-generated data can enhance the performance of edit-based NAR models. We explore two RL approaches: stepwise reward maximization and episodic reward maximization. We discuss the respective pros and cons of these two approaches and empirically verify them. Moreover, we experimentally investigate the impact of temperature setting on performance, confirming the importance of proper temperature setting for NAR models' training.
title Reinforcement Learning for Edit-Based Non-Autoregressive Neural Machine Translation
topic Computation and Language
url https://arxiv.org/abs/2405.01280