README.md

metadata

configs:
  - config_name: default
    data_files:
      - path: train-00000-of-00001.parquet
        split: train
      - path: val-00000-of-00001.parquet
        split: val
      - path: test-00000-of-00001.parquet
        split: test
dataset_info:
  config_name: default
  dataset_size: 10208409
  download_size: 10208409
  features:
    - dtype: string
      name: id
    - dtype: string
      name: author
    - dtype: string
      name: title
    - dtype: string
      name: topic
    - dtype: string
      name: text
    - dtype: string
      name: topic_code
  splits:
    - name: train
      num_bytes: 6060818
      num_examples: 8574
    - name: val
      num_bytes: 2150537
      num_examples: 2825
    - name: test
      num_bytes: 1997054
      num_examples: 2935

English Poetry Topic Clustering Benchmark

Dataset Description

English poetry topic clustering benchmark for evaluating embedding models' performance on unsupervised poetry topic clustering tasks. Each record contains a poem and its topic clustering label.

This dataset is derived from AJMC2002/poems and split into train/validation/test sets with a 6:2:2 ratio while maintaining topic distribution.

Dataset Statistics

Metric	Value
Total Poems	14,334 poems
Train Set	8,574 poems
Validation Set	2,825 poems
Test Set	2,935 poems
Total File Size	9.7 MB

Data Fields

• id: Unique poem identifier
• author: Poem author
• title: Poem title
• text: Poem body/content
• topic: Original topic label
• topic_code: Topic clustering label (used for clustering evaluation)

Dataset Splits

The dataset is split into three sets:

• train: Training set (60% of data)
• val: Validation set (20% of data)
• test: Test set (20% of data)

The split maintains topic distribution across all sets.

Usage

from datasets import load_dataset

dataset = load_dataset("PoetryMTEB/EnglishPoetryTopicClustering")

# Access different splits
train_data = dataset['train']
val_data = dataset['val']
test_data = dataset['test']

Citation

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

• Original dataset: AJMC2002/poems