README.md

---
license: other
task_categories:
- text-classification
task_ids:
- multi-class-classification
language:
- fr
multilinguality:
- monolingual
size_categories:
- 1K<n<10K
pretty_name: FrenchModernPoetryStyleClassification
tags:
- poetry
- french
- style-classification
- literary-movement
- symbolisme
- romantisme
- parnasse
- mteb
- poetrymteb
- embedding-evaluation
annotations_creators:
- expert-generated
source_datasets:
- wkwg429/FrenchLiteratureTextClassification
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  default: true
dataset_info:
- config_name: default
  features:
  - name: id
    dtype: string
  - name: poem
    dtype: string
  - name: label
    dtype: int64
  - name: label_name
    dtype: string
  splits:
  - name: train
    num_examples: 1143
  - name: test
    num_examples: 494
---

# FrenchModernPoetryStyleClassification

Single-label **French poetry literary-style / movement classification** for PoetryMTEB, derived from [FrenchLiteratureTextClassification](https://github.com/wkwg429/FrenchLiteratureTextClassification) (Yang, Wang & Wang, *Information* 2024).

Three classes: **Parnasse** (classical lyric), **Romantisme**, **Symbolisme** (modernist).

## Dataset Card

| Item | Description |
|------|-------------|
| **Source** | [https://github.com/wkwg429/FrenchLiteratureTextClassification](https://github.com/wkwg429/FrenchLiteratureTextClassification) |
| **Paper** | [Reimagining Literary Analysis: Utilizing Artificial Intelligence to Classify Modernist French Poetry](https://www.mdpi.com/2078-2489/15/2/70) ([DOI 10.3390/info15020070](https://doi.org/10.3390/info15020070)) |
| **Languages** | French (`fr`) |
| **Unit** | Poem text (`poem`; upstream field `content`) |
| **Classes** | 3 literary styles / movements |
| **Size** | train=1143; test=494 |
| **Splits** | Upstream `train_set.csv` / `test_set.csv`; train texts that exactly match any test poem were **removed** (9 dropped) to avoid leakage |
| **Text note** | Upstream preprocessing: typically lowercased French with diacritics removed |
| **Evaluation metrics** | Classification on embeddings: **accuracy**, **macro/weighted F1** |

## Label taxonomy (3)

| id | label_name | name_zh | gloss | train | test | total |
|---:|------------|---------|-------|------:|-----:|------:|
| 0 | `Parnasse` | 古典抒情 | Parnassian / classical lyric style | 387 | 171 | 558 |
| 1 | `Romantisme` | 浪漫主义 | Romanticism | 448 | 183 | 631 |
| 2 | `Symbolisme` | 现代主义 | Symbolism / modernist French poetry | 308 | 140 | 448 |

Codebook: `label_taxonomy.json`.

## Features

| Field | Type | Description |
|-------|------|-------------|
| `id` | string | Example id |
| `poem` | string | Poem body (preprocessed French) |
| `label` | int64 | Class index 0–2 |
| `label_name` | string | `Parnasse` / `Romantisme` / `Symbolisme` |

## Construction method

1. Load upstream `ml/data/french/train_set.csv` and `test_set.csv` (`genre`, `content`).
2. Map `genre` ids to readable movement names (same mapping as upstream TF-IDF/Doc2Vec scripts).
3. Drop train rows whose `poem` text exactly equals any test poem.
4. Export parquet splits for PoetryMTEB embedding classification.

## Citation

```bibtex
@Article{info15020070,
  AUTHOR = {Yang, Liu and Wang, Gang and Wang, Hongjun},
  TITLE = {Reimagining Literary Analysis: Utilizing Artificial Intelligence to Classify Modernist French Poetry},
  JOURNAL = {Information},
  VOLUME = {15},
  YEAR = {2024},
  NUMBER = {2},
  ARTICLE-NUMBER = {70},
  URL = {https://www.mdpi.com/2078-2489/15/2/70},
  ISSN = {2078-2489},
  DOI = {10.3390/info15020070}
}
```

Upstream code/data: https://github.com/wkwg429/FrenchLiteratureTextClassification

## License

Follow the upstream GitHub repository and MDPI article terms for reuse.