Advancing Speech Translation: A Corpus of Mandarin-English Conversational Telephone Speech

2024/03/25 by Shannon Wotherspoon, William M. Hartmann, Wotherspoon, Shannon +3 · 1 citation
Arts and Humanities · Computer Science · #Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Natural Language Processing Techniques #Sound (cs.SD) #Subtitles and Audiovisual Media #Translation Studies and Practices #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2404.11619

openalex publication_date 2024/03/25 · openalex created_date 2024/04/20 · openalex updated_date 2026/07/28

Abstract

This paper introduces a set of English translations for a 123-hour subset of the CallHome Mandarin Chinese data and the HKUST Mandarin Telephone Speech data for the task of speech translation. Paired source-language speech and target-language text is essential for training end-to-end speech translation systems and can provide substantial performance improvements for cascaded systems as well, relative to training on more widely available text data sets. We demonstrate that fine-tuning a general-purpose translation model to our Mandarin-English conversational telephone speech training set improves target-domain BLEU by more than 8 points, highlighting the importance of matched training data.

Cited by

• Whisper-UT: A Unified Translation Framework for Speech and Text

Related

• BSTC: A Large-Scale Chinese-English Speech Translation Dataset
• Speech-to-Speech Translation For A Real-world Unwritten Language
• AlloST: Low-resource Speech Translation without Source Transcription
• Direct Speech-to-speech Translation without Textual Annotation using Bottleneck Features
• Speech Translation with Large Language Models: An Industrial Practice