tgdatasets's picture
Update README.md
fbef4ea verified
|
Raw
History Blame Contribute Delete
3.92 kB

Writing.com Archive - TSV Data Format

This document provides a reference for the column structure and data formats used in the shared Writing.com Interactive Stories

File Specification

  • Format: Tab-Separated Values (TSV)
  • Encoding: UTF-8 (with BOM)
  • Headers: Included as the first line

Column Definitions

Index Column Name Description
1 STORY_TITLE The overall title of the story/book.
2 CHAPTER_TITLE The specific title of the chapter or node.
3 AUTHOR_DISPLAY The display name of the author of this entry.
4 AUTHOR_HANDLE The raw username/handle of the author.
5 CONTENT The body text of the chapter. (See Content Formatting).
6 CHOICES Delimited list of branching options. (See Choices Formatting).
7 DATE_POSTED The date the entry was originally posted (usually MM-DD-YYYY).
8 MAP_ID Crucial unique identifier. (See Understanding MAP_ID).
9 STORY_URL The original Writing.com URL for the story.
10 COVER_IMAGE (See Warning below) URL to the story's cover image.
11 SYNOPSIS (See Warning below) The summary/description of the story.
12 RELATED_URL Secondary or related external links.

Metadata Consistency: To maintain data integrity and reduce redundancy, the SYNOPSIS and COVER_IMAGE columns are generally only populated in rows where the MAP_ID indicates an -Intro chapter. Since these values are the same for every chapter in a story, you should look to the Intro row as the primary source for this information.


Understanding MAP_ID

The MAP_ID is the most important field for organizing, linking, and deduplicating the data. It follows a specific hierarchical format:

Format: [STORY_ID]-[CHAPTER_ID]-[UNIX_TIMESTAMP]

  • STORY_ID: A unique numeric ID for the entire story (e.g., 1234567).
  • CHAPTER_ID: The identifier for the specific chapter node.
    • The Intro: The starting node of any story is marked with -Intro (e.g., 1234567-Intro).
      • Note: You may find multiple Intro entries with different timestamps. This is intentional and allows the archive to preserve different variations of a story's introduction as they evolved over time.
    • Branches: Subsequent chapters have numeric or alphanumeric IDs (e.g., 1234567-1411).
  • UNIX_TIMESTAMP: A versioning suffix. Because the archive may contain multiple "snapshots" of the same chapter over time, this timestamp tells you which version is the most recent.

Deduplication Strategy: When processing this file, if you encounter multiple rows with the same STORY_ID and CHAPTER_ID, you must prioritize the one with the highest timestamp.


Technical Details

Content Formatting

The CONTENT column may contain:

  • Raw plain text.
  • HTML snippets (e.g., <br>, <div>).
  • JSON-encoded HTML objects, e.g., {"html": "Body text here..."}. This format is used to preserve complex formatting or handle special characters.

Choices Formatting

The CHOICES column contains the options available to a reader at the end of a chapter. Multiple choices are separated by a double-pipe delimiter: ||||. Example: Go through the door||||Wait in the hall||||Run away


Processing Recommendations

  1. Deduplication First: Filter the dataset by STORY_ID-CHAPTER_ID (the first two parts of the MAP_ID) and keep only the entry with the largest timestamp.
  2. Global Metadata: Link story-level information (SYNOPSIS, COVER_IMAGE) from the -Intro chapter to all other chapters belonging to that same STORY_ID.
  3. JSON Handling: Be prepared to parse JSON in the CONTENT column to extract the underlying HTML/text.

license: mit