A Gutenberg Poetry Corpus Overview

Dataset Description

• Name: Gutenberg Poetry Corpus
• Creator: Allison Parrish
• Content: Approximately three million lines of poetry extracted from hundreds of books in Project Gutenberg.
• Format: Provided as gzip-compressed newline-delimited JSON.
• Structure: Each line of poetry is represented by a JSON object containing an s key (poetry text) and a gid key (source book ID).
• Usage: Particularly suitable for creative computational poetry text generation.

Usage

• Data Access: Obtain the dataset via the download link.
• Data Processing: Process the JSON format directly using programming languages such as Python.
• Example: Provides a Quick Experiments notebook demonstrating how to quickly use the dataset in Python.

Dataset Construction

• Generation Process: Uses the build.py script to access Project Gutenberg books via Gutenberg, dammit, selecting books with the poetry subject and extracting poetry lines based on textual features.
• Filtering Mechanism: Employs wordfilter to exclude lines that may contain offensive content.

Notes

• Content Review: The dataset has not undergone individual review of each poetry line; users must ensure suitability themselves.
• Copyright Status: The dataset includes only poetry lines from English books that are in the public domain in the United States.

License

• Data: Released under the CC0 Public Domain Dedication.
• Code: Released under the MIT License.