NeKo: Cross-Modality Post-Recognition Error Correction with Tasks-Guided Mixture-of-Experts Language Model

TL;DR

NeKo introduces a multi-task Mixture-of-Experts model for cross-modality post-recognition error correction, achieving state-of-the-art results across speech, language, and vision datasets with fewer parameters.

Contribution

The paper presents a novel Multi-Task Correction MoE that effectively learns dataset-specific features and routes tokens to specialized experts, improving error correction performance.

Findings

Achieves 5.0% relative WER reduction on Open ASR Leaderboard

Outperforms GPT-3.5 and Claude-Opus in zero-shot WER reduction

Performs well on grammar and post-OCR correction tasks

Abstract

Construction of a general-purpose post-recognition error corrector poses a crucial question: how can we most effectively train a model on a large mixture of domain datasets? The answer would lie in learning dataset-specific features and digesting their knowledge in a single model. Previous methods achieve this by having separate correction language models, resulting in a significant increase in parameters. In this work, we present Mixture-of-Experts as a solution, highlighting that MoEs are much more than a scalability tool. We propose a Multi-Task Correction MoE, where we train the experts to become an ``expert'' of speech-to-text, language-to-text and vision-to-text datasets by learning to route each dataset's tokens to its mapped expert. Experiments on the Open ASR Leaderboard show that we explore a new state-of-the-art performance by achieving an average relative 5.0% WER reduction and substantial improvements in BLEU scores for speech and translation tasks. On zero-shot evaluation, NeKo outperforms GPT-3.5 and Claude-Opus with 15.5% to 27.6% relative WER reduction in the Hyporadise benchmark. NeKo performs competitively on grammar and post-OCR correction as a multi-task model.