alconost/alconost-multilingual-speech-en-ja-ar-pl-v1

Multilingual Speech & Translation Dataset — EN↔JA/AR/PL (10 phrases, dual-take)

Description 10 English source phrases with expert human translations into Japanese, Egyptian Arabic, and Polish. Each target phrase is recorded by 5 native speakers per language (two takes each). Audio is WAV 48 kHz mono, 16‑bit PCM. Translations are produced and QA’d by professional linguists; recordings follow consistent orthography/style (AR: Egyptian; JA/PL: standard). Each row carries an explicit consent flag.

Contents

• 10 source phrases (English)
• 3 translation CSVs (JA, AR-EG, PL), 50 rows each (10 phrases × 5 speakers)
• 2 audio takes per row (per speaker)
• Speaker demographics (gender, age) and consent info

Use Cases Multilingual TTS/ASR fine‑tuning, cross‑lingual speech translation, multimodal alignment.

License & Consent Each translation row includes consent (TRUE in this release). Optional forms: /consent/{speaker_id}/.

Audio Specs WAV 48 kHz mono, 16-bit PCM, −23 LUFS ±2 LU.

Quality & Provenance Professional translation with bilingual QA; native‑speaker recordings (studio conditions); two takes per item; terminology/orthography consistency; verified consent.

Data layout

/alconost-multilingual-speech-en-ja-ar-pl-v1/
├── README.md
├── LICENSE.txt
├── en_source.csv
├── ja_translations.csv
├── ar_translations.csv
├── pl_translations.csv
└── audio/
    ├── ja/
    ├── ar/
    └── pl/

Audio file naming per speaker folder: audio/{lang}/{speaker_id}/seg_####_take{1|2}.wav

CSV schemas

• en_source.csv: id, domain, source_lang, source_text
• {ja,ar,pl}_translations.csv: id, source_id, domain, target_text, take1, take2, speaker_id, gender, age, consent

Citation Alconost, 2025. “Multilingual Speech & Translation Dataset (EN↔JA/AR/PL).”