README.md

metadata

license: cc-by-4.0

Eng-PidginEdu Dataset

Overview

Eng-PidginEdu is an English–Nigerian Pidgin parallel corpus developed to support machine translation, multilingual NLP, and educational accessibility research for low-resource African languages.

The dataset contains 26,232 parallel sentence pairs collected across 8 Nigerian secondary school subjects, comprising approximately 1.09 million total tokens with high-quality sentence-level alignment and human-validated translations.

Nigerian Pidgin serves as a major lingua franca spoken across Nigeria and West Africa. This dataset was created to improve access to educational technologies and language resources for Pidgin-speaking communities.

---

Dataset Composition

Statistics

Metric	Value
Total Sentence Pairs	26,232
English Tokens	555,398
Pidgin Tokens	533,657
Avg English Sentence Length	21.17 tokens
Avg Pidgin Sentence Length	20.34 tokens
English TTR	0.1032
Pidgin TTR	0.0883

---

Dataset Splits

Split	Size
Train	20,985
Validation (Dev)	2,623
Test	2,624

The dataset was randomly shuffled using a fixed seed (42) to ensure reproducibility.

---

Languages

Role	Language	Code
Source	English	en
Target	Nigerian Pidgin	pcm

---

Data Collection Process

The dataset was compiled from publicly accessible Nigerian secondary school educational materials and curriculum-aligned learning resources.

The data preparation pipeline included:

1. Document extraction using:

   • PyPDF2
   • python-docx

2. Sentence segmentation using:

   • NLTK Punkt tokenizer

3. Text normalization and cleaning using:

   • spaCy
   • textacy
   • regular-expression filtering

4. Translation generation using:

   • Davlan/mt5-small-en-pcm

5. Human validation and correction by native Nigerian Pidgin speakers

6. Alignment verification and post-processing

---

Key Features

• Large-scale English–Pidgin educational parallel corpus
• Coverage across multiple academic subjects
• Human-validated translations
• Natural Nigerian Pidgin expressions and code-switching
• Suitable for low-resource machine translation research
• Reproducible train/dev/test splits

---

Intended Use

This dataset is intended for:

• Machine Translation (MT)
• Low-resource NLP research
• Educational AI systems
• Cross-lingual learning applications
• Multilingual language modeling
• Evaluation of African language translation systems

---

Limitations

• Educational domain only
• Uneven subject distribution
• Contains orthographic variation in Nigerian Pidgin
• Includes informal and code-switched language patterns
• Does not represent all regional varieties of Nigerian Pidgin

---

Bias and Ethical Considerations

The dataset reflects naturally occurring educational language and may contain:

• Curriculum-specific biases
• Subject imbalance
• Linguistic variation associated with Nigerian Pidgin usage

No personally identifiable or sensitive information is included.

Glossary and Terminology Resource

In addition to the parallel sentence corpus, Eng-PidginEdu includes a bilingual educational glossary designed to improve terminology consistency and domain adaptation for machine translation systems.

Glossary Statistics

Metric	Value
Total Glossary Entries	13,634
Source Language	English
Target Language	Nigerian Pidgin

---

Glossary Structure

Each glossary entry contains:

Field	Description
Technical_terms	Educational or domain-specific English terminology
Literal meaning	Standard English explanation or definition
pidgin meaning	Nigerian Pidgin translation or interpretation

---

Example Entries

Technical Term	Literal Meaning	Pidgin Meaning
abandoned	Having been deserted or left	Leave am / throw away
abandonment	The act of giving something up	Di act of giving something up
abbreviated	Shortened form of something	Abbreviated na short form

---

Purpose of the Glossary

The glossary was developed to:

• Improve translation consistency across subjects
• Support educational terminology alignment
• Reduce mistranslation of technical concepts
• Assist low-resource terminology grounding
• Improve domain adaptation for MT systems

The glossary was also used during post-processing and human validation to maintain consistency in subject-specific translations across Computer Science, Government, Biology, Business Studies, and related domains.

---

Intended Use of the Glossary

The glossary can be used for:

• Terminology-aware machine translation
• Lexicon-constrained decoding
• Educational chatbot systems
• Bilingual terminology extraction
• Low-resource lexicon research
• Prompt engineering for LLM translation systems

---

Notes

• Some entries include naturally occurring Nigerian Pidgin paraphrases rather than strict literal translations.
• Orthographic variation may occur due to non-standardized spelling conventions in Nigerian Pidgin.
• Certain educational concepts may have multiple valid Pidgin renderings depending on context.

---

Glossary-Augmented Dataset (Eng-PidginEdu v2)

Eng-PidginEdu_glossary_augmented.csv, also included in this repository, is the parallel corpus above with the glossary merged directly into the Pidgin translations: matched academic terms get their Pidgin meaning inserted inline, in brackets, immediately after the term.

en:            The ability for educators and students to adopt the
               materials for free.
pcm:           Teachers and students fit get the materials for free.
pcm_augmented: Teachers and students fit get the materials (tins wey
               dem use for teaching) for free.

This glossary-augmented file -- not the plain parallel corpus -- is the actual training target used to fine-tune all 14 models in the Eng-PidginEdu benchmark, and is the source of this project's novel glossary-accuracy evaluation metric.

Columns

In addition to en, pcm, and Subject above:

Column	Description
pcm_augmented	The Pidgin translation with inline glosses inserted
glossed_terms	Semicolon-separated list of which terms were glossed in this row
n_glosses	Number of glosses inserted in this row

Augmentation statistics

Metric	Value
Rows with at least one gloss	16,114 / 26,232 (61.4%)
Total glosses inserted	28,243
Distinct terms glossed	2,473

Reproducing it

The exact preprocessing script that produces this file from the raw corpus and the standalone glossary table above is glossary_augment.py, published in the benchmark's GitHub repository -- verified to reproduce this file exactly, byte-for-byte, across all 26,232 rows.

---

Licensing

This dataset is released under the CC-BY-4.0 License.

---

Citation

If you use this dataset, please cite:

@dataset{eng_pidginedu_2026,
  author = {Oladipupo, F., 
Flora Oladipupo, Felix Akintola, Aleem Aduragbemi, Emmanuel Macaulay, Olisaemeka Nwosu, Adeniran Treasure, Michael Kanu, Jennifer Ochonogor, Victor Jayeoba, Zion O. Oluwasegun, Gift Warieta, Joy Adebiyi, Consolation U. Obazee},
  title = {Eng-PidginEdu: An English–Nigerian Pidgin Educational Parallel Corpus},
  year = {2026},
  publisher = {Hugging Face}
}