Saved in:
Bibliographic Details
Main Authors: Ma, Zhengrui, Feng, Yang, Zhang, Min
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.17170
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912399754264576
author Ma, Zhengrui
Feng, Yang
Zhang, Min
author_facet Ma, Zhengrui
Feng, Yang
Zhang, Min
contents Streaming generation models are utilized across fields, with the Transducer architecture being popular in industrial applications. However, its input-synchronous decoding mechanism presents challenges in tasks requiring non-monotonic alignments, such as simultaneous translation. In this research, we address this issue by integrating Transducer's decoding with the history of input stream via a learnable monotonic attention. Our approach leverages the forward-backward algorithm to infer the posterior probability of alignments between the predictor states and input timestamps, which is then used to estimate the monotonic context representations, thereby avoiding the need to enumerate the exponentially large alignment space during training. Extensive experiments show that our MonoAttn-Transducer effectively handles non-monotonic alignments in streaming scenarios, offering a robust solution for complex generation tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17170
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Overcoming Non-monotonicity in Transducer-based Streaming Generation
Ma, Zhengrui
Feng, Yang
Zhang, Min
Computation and Language
Artificial Intelligence
Streaming generation models are utilized across fields, with the Transducer architecture being popular in industrial applications. However, its input-synchronous decoding mechanism presents challenges in tasks requiring non-monotonic alignments, such as simultaneous translation. In this research, we address this issue by integrating Transducer's decoding with the history of input stream via a learnable monotonic attention. Our approach leverages the forward-backward algorithm to infer the posterior probability of alignments between the predictor states and input timestamps, which is then used to estimate the monotonic context representations, thereby avoiding the need to enumerate the exponentially large alignment space during training. Extensive experiments show that our MonoAttn-Transducer effectively handles non-monotonic alignments in streaming scenarios, offering a robust solution for complex generation tasks.
title Overcoming Non-monotonicity in Transducer-based Streaming Generation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2411.17170