README.md

metadata

license: apache-2.0
task_categories:
  - text-classification
task_ids:
  - multi-class-classification
language:
  - ar
multilinguality:
  - monolingual
size_categories:
  - 1K<n<10K
pretty_name: FannOrFlopClassification
tags:
  - poetry
  - arabic
  - text-classification
  - genre-classification
  - meter-classification
  - era-classification
  - mteb
  - poetrymteb
  - embedding-evaluation
annotations_creators:
  - expert-generated
source_datasets:
  - omkarthawakar/FannOrFlop
configs:
  - config_name: genre
    data_files:
      - split: train
        path: genre/train-*
      - split: validation
        path: genre/validation-*
      - split: test
        path: genre/test-*
    default: true
  - config_name: meter
    data_files:
      - split: train
        path: meter/train-*
      - split: validation
        path: meter/validation-*
      - split: test
        path: meter/test-*
  - config_name: era
    data_files:
      - split: train
        path: era/train-*
      - split: validation
        path: era/validation-*
      - split: test
        path: era/test-*
dataset_info:
  - config_name: genre
    features:
      - name: id
        dtype: string
      - name: title
        dtype: string
      - name: author
        dtype: string
      - name: poem
        dtype: string
      - name: label
        dtype: int64
      - name: label_name
        dtype: string
    splits:
      - name: train
        num_examples: 5176
      - name: validation
        num_examples: 657
      - name: test
        num_examples: 1106
  - config_name: meter
    features:
      - name: id
        dtype: string
      - name: title
        dtype: string
      - name: author
        dtype: string
      - name: poem
        dtype: string
      - name: label
        dtype: int64
      - name: label_name
        dtype: string
    splits:
      - name: train
        num_examples: 5176
      - name: validation
        num_examples: 657
      - name: test
        num_examples: 1106
  - config_name: era
    features:
      - name: id
        dtype: string
      - name: title
        dtype: string
      - name: author
        dtype: string
      - name: poem
        dtype: string
      - name: label
        dtype: int64
      - name: label_name
        dtype: string
    splits:
      - name: train
        num_examples: 5176
      - name: validation
        num_examples: 657
      - name: test
        num_examples: 1106

FannOrFlopClassification

Single-label Arabic poetry classification for PoetryMTEB embedding evaluation, derived from Fann or Flop (EMNLP 2025).

Three configs share the same poems and the same poet-aware train/validation/test split:

Config	Task	#classes
genre	poetic genre / theme	14
meter	poetic meter (ʿarūḍ)	19
era	historical era	12

Dataset Card

Item	Description
Source	omkarthawakar/FannOrFlop; paper arXiv:2505.18152
Languages	Arabic (ar)
Size	train=5176; validation=657; test=1106 (after filtering)
Filtering	Removed poems with unlabeled meter غير محدد (n=45). Raw=6984 → kept=6939.
Label type	Single-label multiclass (label int + label_name string)
Splits	Poet-aware ~8:1:1 (train / validation / test), shared across configs
License	Apache-2.0 (same as upstream)
Evaluation metrics	Classification on embeddings: accuracy, macro/weighted F1

---

Features

Field	Type	Description
id	string	Upstream poem id
title	string	Poem title
author	string	Poet name
poem	string	Poem body (verse index lines removed)
label	int64	Class index (0 … C−1)
label_name	string	Arabic class name

Label inventories: label_taxonomy.json.

---

Label inventories (all splits pooled)

genre

id	name	count
0	مدح	1354
1	غزل	894
2	هجاء	872
3	حكمه	705
4	عامه	609
5	عتاب	601
6	رثاء	466
7	فراق	458
8	سياسية	261
9	دينية	244
10	شوق	211
11	اعتذار	106
12	صبر	97
13	حزينة	61

meter

id	name	count
0	الطويل	1791
1	الكامل	1298
2	البسيط	892
3	الوافر	783
4	الخفيف	533
5	السريع	329
6	المتقارب	303
7	الرمل	297
8	الرجز	289
9	المنسرح	202
10	المجتث	87
11	الهزج	80
12	المديد	35
13	المتدارك	5
14	الدوبيت	4
15	موشح	4
16	المقتضب	3
17	المواليا	3
18	المضارع	1

era

id	name	count
0	العصر العباسي	2342
1	العصر الحديث	699
2	العصر الأموي	556
3	المغرب والأندلس	446
4	الدولة الايوبية	413
5	الدولة الفاطمية	409
6	الدولة المملوكية	408
7	قبل الإسلام	368
8	الشعراء المخضرمون	361
9	العصر العثماني	326
10	عصر بين الدولتين	321
11	العصر الإسلامي	290

---

Construction method

1. Load omkarthawakar/FannOrFlop.
2. Drop poems whose meter is غير محدد (unlabeled).
3. Normalize poem_verses into poem (strip numeric verse indices).
4. Build frequency-ordered label vocabularies for genre, meter, era.
5. Poet-aware split ≈ 80% / 10% / 10%; reuse the same split for all three configs.

---

How to load

from datasets import load_dataset

genre = load_dataset("PoetryMTEB/FannOrFlopClassification", "genre")
meter = load_dataset("PoetryMTEB/FannOrFlopClassification", "meter")
era = load_dataset("PoetryMTEB/FannOrFlopClassification", "era")

print(genre["test"][0]["title"], genre["test"][0]["label_name"])

For embedding evaluation, encode poem (optionally prepend title), then train a linear / MLP classifier on label.

---

License

Distributed under Apache License 2.0, consistent with the upstream FannOrFlop release.

---

Citation / provenance

@inproceedings{alghallabi-etal-2025-fann,
  title = {{Fann or Flop: A Multigenre, Multiera Benchmark for Arabic Poetry Understanding in LLMs}},
  author = {{Alghallabi, Wafa and Thawkar, Ritesh and Ghaboura, Sara and More, Ketan and Thawakar, Omkar and Cholakkal, Hisham and Khan, Salman and Anwer, Rao Muhammad}},
  booktitle = {{Proceedings of EMNLP 2025}},
  year = {{2025}},
  url = {{https://arxiv.org/abs/2505.18152}},
}

• Upstream dataset: https://huggingface.co/datasets/omkarthawakar/FannOrFlop
• This Hub packaging: PoetryMTEB/FannOrFlopClassification