Dialogs Re-enacted Across Languages

TL;DR

This paper introduces a protocol and dataset for collecting closely matched bilingual dialog pairs to enhance cross-language speech translation, providing resources and insights for researchers in speech and language processing.

Contribution

It presents a novel protocol for collecting bilingual dialog pairs, along with a publicly available dataset and observations to aid speech-to-speech translation research.

Findings

Dataset of matched bilingual dialogs released publicly

Protocol facilitates cross-language prosodic mapping research

Initial observations on dialog characteristics across languages

Abstract

To support machine learning of cross-language prosodic mappings and other ways to improve speech-to-speech translation, we present a protocol for collecting closely matched pairs of utterances across languages, a description of the resulting data collection and its public release, and some observations and musings. This report is intended for: people using this corpus, people extending this corpus, and people designing similar collections of bilingual dialog data.