DexterityLearn/dexterity-pidgin-voice-dataset-sample-v2

Dexterity Pidgin Voice Dataset — Sample v2 Creator: Dexterity Learn.

Dataset Description 49 human-created Nigerian Pidgin English (Naija) utterances with aligned home studio audio, covering 8 domains: faith, wisdom, money, business, relationships, community, social commentary, and everyday conversation. Three utterances (U20–U22) address consent and anti-harassment themes, included for social good NLP use.

What sets this dataset apart is its dual-translation structure — every utterance includes both a literal English translation and a Level 3 cultural translation capturing the full emotional and contextual meaning as a native speaker would understand it.

All text, translations and audio recordings are original works created by Dexterity Learn. No third-party scripts, scraped content, or copyrighted material was used in the production of this dataset.

Uses

Direct Use This dataset is suitable for:

-Automatic speech recognition (ASR) for Nigerian Pidgin English -Sentiment and emotion classification in low-resource African languages -English–Pidgin parallel translation and machine translation research -Cultural and contextual NLP model training -Linguistic research on Nigerian Pidgin English

Out-of-Scope Use

This dataset must not be used to: -Clone, synthesise, or replicate the speaker's voice -Train any text-to-speech (TTS) or voice conversion model that reproduces the speaker's voice -Impersonate or imitate the speaker in any form

• Generate audio content that could be mistaken for the speaker's actual voice

These restrictions apply regardless of whether the use is commercial, research-based, non-commercial, or open-source. Violation may constitute a breach of applicable law.

Dataset Structure Field Description utterance_id - Unique identifier (e.g. DL_S02_U01) pidgin_text - Original Nigerian Pidgin English utterance english_literal - Word-for-word English translation english_cultural - Level 3 — full cultural and emotional meaning domain - Broad topic area category - Sub-topic within domain speech_act - statement / advice / question utterance_type - proverb / declarative / conditional / imperative sentiment - positive / negative / neutral / cautionary audio_filename - Matching WAV file sample_rate_hz - 44,100 Hz duration_seconds - Recording length in seconds quality_flag - blank = clear · REVIEW · SENSITIVE

Domain Coverage

Domain/Utterances

wisdom/12 community/9 money/9 faith/5 relationships/5 business/4 conversational/4 social_commentary/1

Dataset Creation

Curation Rationale Nigerian Pidgin English (Naija) is spoken by an estimated 75–110 million people across Nigeria and West Africa, yet remains severely underrepresented in AI training data. Most existing low-resource African language datasets provide only basic literal translations, missing the cultural and emotional depth that makes language meaningful. This dataset was created to address both gaps — providing not only parallel text and audio data for Naija, but a Level 3 cultural translation layer that captures how utterances are actually understood by native speakers.

Source Data

Data Collection and Processing All utterances are original scripts written by the dataset creator in Nigerian Pidgin English, covering everyday topics relevant to Nigerian and West African life. Scripts were recorded in a home studio environment in a single speaker session. Audio files are in WAV format, mono, 44,100 Hz. Each transcript was human-verified against the recording to confirm accuracy before inclusion.

Who Are the Source Data Producers? The source data was produced entirely by the dataset creator — a Nigerian female native speaker of Nigerian Pidgin English with professional experience in content creation and educational media in Naija. All written scripts, translations, and audio recordings are original works.

Bias, Risks, and Limitations

Single speaker: This sample features one female Nigerian speaker. Accent, tone, and pronunciation will reflect one individual's variety of Nigerian Pidgin English and may not represent the full range of Naija dialects across Nigeria and West Africa.

Small sample size: At 49 utterances, this is a sample dataset. It is suitable for evaluation, prototyping, and research but not sufficient alone for training production-scale models.

ialect variation: Nigerian Pidgin English varies across regions. This dataset reflects one speaker's dialect and should be supplemented with broader speaker diversity for dialect-robust models.

Domain scope: While 8 domains are covered, representation is uneven. Wisdom (12) and community (9) are most represented; social commentary (1) is minimal.

Recommendations Users should be aware that models trained on this sample alone may not generalise well across all Nigerian Pidgin English speakers or regional varieties. This dataset is best used as part of a broader low-resource language training pipeline, supplemented with additional Naija data sources where available. For production use cases, contact Dexterity Learn regarding expanded dataset licensing.

⚠️ Copyright & Voice Cloning Restriction © 2025 Dexterity Learn. All rights reserved.

All text, translations, and audio recordings are protected under copyright law. This dataset is made available under CC BY 4.0 subject to the voice cloning and impersonation restrictions stated above. Unauthorised commercial exploitation outside the terms of the licence is prohibited.

Licence & Attribution CC BY 4.0 — commercial use, modifications, and redistribution permitted with attribution. Voice cloning prohibited.

Dexterity Learn. Dexterity Pidgin Voice Dataset — Sample v2 (v1.0). 2025. CC BY 4.0.

Commercial Licensing For commercial licensing, domain-specific packages, or custom dataset builds: 📧 dexteritylearnhq@gmail.com

Enterprise and bulk licensing is available. Custom datasets can be built to specification for specific domains, speaker profiles, and use cases.

About Dexterity Learn Dexterity Learn is a registered social enterprise building and licensing Nigerian Pidgin English AI training datasets, ensuring West African languages are represented in AI with accuracy and cultural depth.

Related: Dexterity Pidgin Voice Dataset — Sample v1 Dataset Card Contact 📧 dexteritylearnhq@gmail.com