English

Connecting Voices: LoReSpeech as a Low-Resource Speech Parallel Corpus

Computation and Language 2026-03-11 v2

Abstract

Aligned audio corpora are fundamental to NLP technologies such as ASR and speech translation, yet they remain scarce for underrepresented languages, hindering their technological integration. This paper introduces a methodology for constructing LoReSpeech, a low-resource speech-to-speech translation corpus. Our approach begins with LoReASR, a sub-corpus of short audios aligned with their transcriptions, created through a collaborative platform. Building on LoReASR, long-form audio recordings, such as biblical texts, are aligned using tools like the MFA. LoReSpeech delivers both intra- and inter-language alignments, enabling advancements in multilingual ASR systems, direct speech-to-speech translation models, and linguistic preservation efforts, while fostering digital inclusivity. This work is conducted within Tutlayt AI project (https://tutlayt.fr).

Keywords

Cite

@article{arxiv.2502.18215,
  title  = {Connecting Voices: LoReSpeech as a Low-Resource Speech Parallel Corpus},
  author = {Samy Ouzerrout},
  journal= {arXiv preprint arXiv:2502.18215},
  year   = {2026}
}

Comments

This paper is withdrawn because the LoReSpeech dataset described in Section 2 is not currently available, which affects the reproducibility of the work and the validity of the experimental results

R2 v1 2026-06-28T21:57:20.518Z