Learning from others' mistakes: Finetuning machine translation models with span-level error annotations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Lily H., Dadkhahi, Hamid, Finkelstein, Mara, Trabelsi, Firas, Luo, Jiaming, Freitag, Markus
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910660141514752
author Zhang, Lily H.
Dadkhahi, Hamid
Finkelstein, Mara
Trabelsi, Firas
Luo, Jiaming
Freitag, Markus
author_facet Zhang, Lily H.
Dadkhahi, Hamid
Finkelstein, Mara
Trabelsi, Firas
Luo, Jiaming
Freitag, Markus
contents Despite growing interest in incorporating feedback to improve language models, most efforts focus only on sequence-level annotations. In this work, we explore the potential of utilizing fine-grained span-level annotations from offline datasets to improve model quality. We develop a simple finetuning algorithm, called Training with Annotations (TWA), to directly train machine translation models on such annotated data. TWA utilizes targeted span-level error information while also flexibly learning what to penalize within a span. Moreover, TWA considers the overall trajectory of a sequence when deciding which non-error spans to utilize as positive signals. Experiments on English-German and Chinese-English machine translation show that TWA outperforms baselines such as Supervised FineTuning on sequences filtered for quality and Direct Preference Optimization on pairs constructed from the same data.
format Preprint
id arxiv_https___arxiv_org_abs_2410_16509
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning from others' mistakes: Finetuning machine translation models with span-level error annotations
Zhang, Lily H.
Dadkhahi, Hamid
Finkelstein, Mara
Trabelsi, Firas
Luo, Jiaming
Freitag, Markus
Computation and Language
Machine Learning
Despite growing interest in incorporating feedback to improve language models, most efforts focus only on sequence-level annotations. In this work, we explore the potential of utilizing fine-grained span-level annotations from offline datasets to improve model quality. We develop a simple finetuning algorithm, called Training with Annotations (TWA), to directly train machine translation models on such annotated data. TWA utilizes targeted span-level error information while also flexibly learning what to penalize within a span. Moreover, TWA considers the overall trajectory of a sequence when deciding which non-error spans to utilize as positive signals. Experiments on English-German and Chinese-English machine translation show that TWA outperforms baselines such as Supervised FineTuning on sequences filtered for quality and Direct Preference Optimization on pairs constructed from the same data.
title Learning from others' mistakes: Finetuning machine translation models with span-level error annotations
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.16509