ghananlpcommunity/twi-sentiments-corpus-400k

This dataset is made available because of Ghana NLP's volunteer driven research work. Please consider contributing to any of our projects on Github

Twi Sentiment Corpus

Dataset Description

This dataset contains sentiment-labeled text data in Twi for binary sentiment classification (Positive/Negative). Sentiments are extracted and processed from the English meanings of the sentences using DistilBERT for sentiment classification. The dataset is part of a larger collection of African language sentiment analysis resources.

Dataset Statistics

• Total samples: 432,647
• Positive sentiment: 249237 (57.6%)
• Negative sentiment: 183410 (42.4%)

Dataset Structure

Data Fields

• Text Column: Contains the original text in Twi
• sentiment: Sentiment label (Positive or Negative only)

Data Splits

This dataset contains a single split with all the processed data.

Data Processing

The sentiment labels were generated using:

• Model: distilbert-base-uncased-finetuned-sst-2-english
• Processing: Batch processing with optimization for efficiency
• Deduplication: Duplicate entries were removed based on text content
• Filtering: Only Positive and Negative sentiments retained for binary classification

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("michsethowusu/twi-sentiments-corpus")

# Access the data
print(dataset['train'][0])

# Check sentiment distribution
from collections import Counter
sentiments = [item['sentiment'] for item in dataset['train']]
print(Counter(sentiments))

Use Cases

This dataset is ideal for:

• Binary sentiment classification tasks
• Training sentiment analysis models for Twi
• Cross-lingual sentiment analysis research
• African language NLP model development

Citation

If you use this dataset in your research, please cite:

@dataset{twi_sentiments_corpus,
  title={Twi Sentiment Corpus},
  author={Mich-Seth Owusu},
  year={2025},
  url={https://huggingface.co/datasets/michsethowusu/twi-sentiments-corpus}
}

License

This dataset is released under the MIT License.

Contact

For questions or issues regarding this dataset, please open an issue on the dataset repository.

Dataset Creation

Date: 2025-07-02 Processing Pipeline: Automated sentiment analysis using HuggingFace Transformers Quality Control: Deduplication, batch processing optimizations, and binary sentiment filtering applied