README.md

Dataset Card for sarcasm_headlines_multilingual

size_categories:

• 100k<n source_datasets:
• original
• https://www.kaggle.com/datasets/rmisra/news-headlines-dataset-for-sarcasm-detection
• https://www.kaggle.com/datasets/maxscheijen/dutch-news-articles languages:
• en
• it
• nl multilinguality:
• multilingual task_categories:
• text_classification

Table of Contents

• Dataset Description
  • Dataset Summary
  • Supported Tasks
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information

Dataset Description

• Repository: https://github.com/helinivan/multilingual-sarcasm-detector

Dataset Summary

Dataset consists of news article headlines in Dutch, English and Italian. The news article headlines are both from actual news sources and sarcastic/satirical newspapers. The news article is determined sarcastic/non-sarcastic based on the news article source.

The sources of news articles are:

• The Huffington Post (en, non-sarcastic)
• The Onion (en, sarcastic)
• NOS (nl, non-sarcastic)
• De Speld (nl, sarcastic)
• Il Giornale (it, non-sarcastic)
• Lercio (it, sarcastic)

Languages

en, nl, it

Dataset Structure

Data Instances

total_length: 67480 sarcastic: 25,609 non_sarcastic: 41,817 english: 22,837 dutch: 20,771 italian: 23,871

Data Fields

• article_url: str
• article_title: str
• is_sarcastic: int
• lang: str
• title_length: int

Dataset Creation

Source Data

English news article titles are gathered from this Kaggle dataset: https://www.kaggle.com/datasets/rmisra/news-headlines-dataset-for-sarcasm-detection Dutch non-sarcastic news article titles are gathered from this Kaggle dataset: https://www.kaggle.com/datasets/maxscheijen/dutch-news-articles

Rest of the data is scraped directly from the newspapers.