README.md

---
license: mit
---

# Populist Argument Schemes 🗣️

A curated dataset of translated tweets from four political leaders, annotated with argument schemes according to argumentation theory. The dataset enables fine-tuning and evaluation of language models for argument mining, rhetorical analysis, and populist discourse detection.

## Dataset Summary

This dataset contains **English-translated tweets** from:
- **Matteo Salvini** 🇮🇹
- **Jair Bolsonaro** 🇧🇷
- **Donald Trump** 🇺🇸
- **Joe Biden** 🇺🇸

Each tweet is annotated with its **primary argumentation scheme** (based on *Argument 1*) following a scheme taxonomy inspired by Walton and Macagno. The dataset is derived from Fabrizio Macagno's original annotated corpus on the *Language of Populism* and translated for cross-lingual NLP applications.

## Features

- `Politician`: Name of the political figure (e.g., "Trump")
- `Argument`: The tweet text (translated into English)
- `Argument Scheme`: The full name of the primary argumentation scheme (e.g., "Argument from Consequences")

## Argument Scheme Labels

The dataset includes 13 argument schemes:

| Code  | Argument Scheme                        |
|-------|----------------------------------------|
| AA    | Argument from Analogy                  |
| AC    | Argument from Consequences             |
| AH    | Ad Hominem                             |
| AS    | Argument from Sign                     |
| AV    | Argument from Values                   |
| BEX   | Argument from Best Explanation         |
| CE    | Argument from Cause to Effect          |
| CLASS | Argument from Classification           |
| CO    | Argument from Commitment               |
| PK    | Argument from Position to Know         |
| PO    | Argument from Popular Opinion          |
| PR    | Argument from Practical Reasoning      |
| VV    | Victimization                          |

## Use Cases

This dataset can be used for:
- Fine-tuning LLMs for **argument scheme classification**
- Training models for **argument mining** and **fallacy detection**
- Studying **populist rhetoric** and comparative discourse analysis
- Building educational tools for **teaching argumentation theory**

## How to Use 🧠

You can easily load this dataset using the 🤗 `datasets` library, which allows seamless integration with Hugging Face Transformers, evaluation tools, and fine-tuning pipelines.

### 📥 Load the Dataset

```python
from datasets import load_dataset

# Load the dataset from the Hugging Face Hub
dataset = load_dataset("MidhunKanadan/populist-argument-schemes")

# View a sample entry
print(dataset["train"][0])
```

### 🖨️ Expected Output

```python
{
    'Politician': 'Biden',
    'Argument': 'It matters whether you continue to wear a mask. It matters whether you continue to socially distance. It matters whether you wash your hands. It all matters and can help save lives.',
    'Argument Scheme': 'ARGUMENT FROM CONSEQUENCES'
}
```

## Citation

If you use this dataset, please cite the original author:

> Macagno, F. (2022). Argumentation schemes, fallacies, and evidence in politicians' argumentative tweets – a coded dataset. *Data in Brief*, 44, 108501. https://doi.org/10.1016/j.dib.2022.108501