Browse Datasets

AfriSenti

Sentiment Analysis for 14 African Languages

sentiment-analysis text-classification 25MB CC-BY-4.0
Access Dataset Read Paper

Description

AfriSenti is a multilingual sentiment analysis dataset for 14 African languages. It contains Twitter data annotated for positive, negative, and neutral sentiment. The dataset enables research on sentiment analysis in low-resource African languages.

Quick Start

Python 
from datasets import load_dataset

# Load Hausa sentiment data
dataset = load_dataset("masakhane/afrisenti", "hau")

# View a sample
print(dataset['train'][0])
# Output: {'text': '...', 'label': 1}

# Label mapping: 0=negative, 1=neutral, 2=positive

Project Ideas

Social Media Sentiment Monitor

Build a dashboard that tracks sentiment about key topics in African social media

Intermediate

Multilingual Sentiment API

Create an API that provides sentiment analysis for multiple African languages

Intermediate

Ethical Considerations

  • Twitter data may not represent general population sentiment
  • Urban bias in social media data
  • Potential for misuse in surveillance

Additional Information

AfriSenti was created as part of the SemEval-2023 shared task on sentiment analysis for African languages. It represents one of the largest multilingual sentiment datasets for African languages.

Label Distribution

  • Positive: Tweets expressing positive sentiment
  • Negative: Tweets expressing negative sentiment
  • Neutral: Tweets with neutral or no clear sentiment

Citation

@inproceedings{muhammad-etal-2023-afrisenti,
    title = "{A}fri{S}enti: A {T}witter Sentiment Analysis Benchmark for {A}frican Languages",
    author = "Muhammad, Shamsuddeen Hassan and others",
    booktitle = "EMNLP",
    year = "2023"
}