OxfordVGG Submission to the EGO4D AV Transcription Challenge

TL;DR

This paper details OxfordVGG's submission to the EGO4D AV Transcription Challenge, introducing WhisperX for efficient long-form speech transcription with word-level alignment, achieving top performance.

Contribution

The paper introduces WhisperX, a novel system for long-form audio transcription with word-level timing, and provides publicly available text normalizers, advancing speech recognition technology.

Findings

Achieved 56.0% WER on the challenge test set

Ranked 1st on the leaderboard

Provided publicly available code and models

Abstract

This report presents the technical details of our submission on the EGO4D Audio-Visual (AV) Automatic Speech Recognition Challenge 2023 from the OxfordVGG team. We present WhisperX, a system for efficient speech transcription of long-form audio with word-level time alignment, along with two text normalisers which are publicly available. Our final submission obtained 56.0% of the Word Error Rate (WER) on the challenge test set, ranked 1st on the leaderboard. All baseline codes and models are available on https://github.com/m-bain/whisperX.