README.md

metadata

license: cc-by-sa-4.0
task_categories:
  - text-classification
task_ids:
  - multi-class-classification
language:
  - en
  - de
  - es
  - cs
multilinguality:
  - multilingual
size_categories:
  - 10K<n<100K
pretty_name: Multilingual Poetry Meter Classification
tags:
  - poetry
  - meter-classification
  - prosody
  - multilingual
  - text-classification
  - mteb
  - poetrymteb
  - embedding-evaluation
annotations_creators:
  - found
source_datasets:
  - riazhsks/multilingual-annotated-poems
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: title
        dtype: string
      - name: author
        dtype: string
      - name: language
        dtype: string
      - name: label
        dtype: int64
      - name: label_name
        dtype: string
    splits:
      - name: train
        num_examples: 24464
      - name: test
        num_examples: 6114

Multilingual Poetry Meter Classification

Multilingual metrical foot classification for PoetryMTEB embedding evaluation, built from EN/DE/ES/CS corpora with full text in multilingual-annotated-poems.

Hungarian and modern Czech (C3P) standalone annotation files are excluded (no poem text in-repo).

Dataset Card

Item	Description
Source	https://github.com/riazhsks/multilingual-annotated-poems English / German / Spanish / CCV JSONL
Languages	en, de, es, cs
Size	train=24464; test=6114 (no validation)
Classes	7 meters with frequency ≥ 10 (after per-language cap)
Filtering	Valid metrical_foot; drop noisy tags (s/c/single/…); length ∈ [40, 8000]; ≤10000 poems / language (stratified by meter)
Splits	Stratified by language×meter ≈ 80% / 19%, seed=42
License	CC BY-SA 4.0
Dataset version	1.0.0
Evaluation metrics	accuracy, macro/weighted F1

Features

Field	Type	Description
id	string	Example id
poem	string	Poem text
title	string	Title
author	string	Author
language	string	en / de / es / cs
label	int64	Meter class index
label_name	string	Metrical foot (e.g. iambic)

Language distribution

language	train	test	total
en	1067	267	1334
de	7995	1998	9993
es	7402	1849	9251
cs	8000	2000	10000

Label inventory (kept)

id	label_name	gloss	train	test	total
0	iambic	Iambic (抑扬格)	14785	3696	18481
1	trochaic	Trochaic (扬抑格)	4375	1094	5469
2	polymetric	Polymetric / mixed (多格律)	4326	1081	5407
3	dactylic	Dactylic (扬抑抑格)	514	127	641
4	anapestic	Anapestic (抑抑扬格)	317	79	396
5	amphibrachic	Amphibrachic (抑扬抑格)	136	34	170
6	hexameter	Hexameter (六音步)	11	3	14

Dropped / unused meters (examples)

meter	count
sapphicusminor	2
prosodiakos	2
pherekrateus	1
elegiambus	1
glykoneus	1

Construction method

1. Load EN/DE/ES/CCV JSONL with text + metrical_foot.
2. Normalize meters; drop invalid/noisy tags; length-filter.
3. Cap each language at 10000 (stratified by meter) for balance.
4. Keep meters with ≥ 10 examples; assign frequency-descending ids.
5. Stratified train/test by language×meter.

How to load

from datasets import load_dataset

ds = load_dataset("PoetryMTEB/MultilingualPoetryMeterClassification")
print(ds["test"][0]["language"], ds["test"][0]["label_name"])

Citation

@misc{riazhsks_multilingual_annotated_poems,
  title={Multilingual Annotated Poems},
  author={riazhsks},
  year={2025},
  url={https://github.com/riazhsks/multilingual-annotated-poems}
}

Also cite the upstream corpora used in that repository (CCV, DLK, Spanish sonnet corpora, poetry-eval) as appropriate.

Hub packaging: PoetryMTEB/MultilingualPoetryMeterClassification (v1.0.0).