Datasets:

webxos
/

synthetic_error_correction

license: mit
task_categories:
  - text-classification
  - text-generation
tags:
  - nlp
  - text
  - text classification
  - text generation
  - error
  - error correction
  - synthetic
  - generated
  - ocr
  - text-correction
  - code-correction

      ___           ___                         ___            ___           ___     
     /\  \         /\__\         _____         /|  |          /\  \         /\__\    
    _\:\  \       /:/ _/_       /::\  \       |:|  |         /::\  \       /:/ _/_   
   /\ \:\  \     /:/ /\__\     /:/\:\  \      |:|  |        /:/\:\  \     /:/ /\  \  
  _\:\ \:\  \   /:/ /:/ _/_   /:/ /::\__\   __|:|__|       /:/  \:\  \   /:/ /::\  \ 
 /\ \:\ \:\__\ /:/_/:/ /\__\ /:/_/:/\:|__| /::::\__\_____ /:/__/ \:\__\ /:/_/:/\:\__\
 \:\ \:\/:/  / \:\/:/ /:/  / \:\/:/ /:/  / ~~~~\::::/___/ \:\  \ /:/  / \:\/:/ /:/  /
  \:\ \::/  /   \::/_/:/  /   \::/_/:/  /      |:|~~|      \:\  /:/  /   \::/ /:/  / 
   \:\/:/  /     \:\/:/  /     \:\/:/  /       |:|  |       \:\/:/  /     \/_/:/  /  
    \::/  /       \::/  /       \::/  /        |:|__|        \::/  /        /:/  /   
     \/__/         \/__/         \/__/         |/__/          \/__/         \/__/

Synthetic Error Correction

UNDER DEVELOPMENT

Small synthetic dataset (28 examples: 22 train, 6 test) of error-correction pairs for training text correction models (e.g., OCR post-processing, typos).

Dataset Summary

This dataset was generated locally using the TYPE/RIGHTER offline-first PWA. It contains synthetic error-correction pairs for training text correction models.

Generated by TYPE/RIGHTER • Offline-First PWA, downloadable in the /typerighter/ folder of the repo

Supported Tasks and Leaderboards

Text error correction
OCR post-processing
Typographical error correction

Languages

English

Dataset Structure

Data Instances

Each instance contains:

original: The text as it appeared in the uploaded file
corrupted: The procedurally "damaged" version for error detection training
target: The cleaned, normalized version for error correction training
tags: Content-based classification tags
source: Original file name

Data Fields

original: string - Original text
corrupted: string - Text with injected errors
target: string - Cleaned target text
tags: Sequence[string] - Content classification tags
source: string - Source file name

Data Splits

Train: 22 examples
Test: 6 examples

Dataset Creation

Curation Rationale

This dataset was created to provide synthetic data for training error correction models from user-provided documents.

Source Data

The data was generated from user-uploaded documents in .txt, .csv, .md, or .jsonl formats.

Annotations

Annotations were automatically generated through:

Text normalization (simulated cleaning)
Controlled error injection based on configurable parameters
Content-based classification

Personal and Sensitive Information

All processing occurs locally in the browser. No data was uploaded to any server.

Considerations for Using the Data

Social Impact of Dataset

This dataset enables the creation of error correction models for various text domains.

Discussion of Biases

The error patterns are synthetic and may not perfectly represent real-world error distributions.

Other Known Limitations

Synthetic errors may not capture all real-world error patterns
Limited to text-based errors (no formatting or layout errors)

Additional Information

Dataset Curators

Generated by TYPE/RIGHTER PWA by webXOS

Licensing Information

MIT

Citation Information

@misc{typerighter2026,
  title={TYPE/RIGHTER Synthetic Error Correction Dataset},
  author={TYPE/RIGHTER PWA},
  year={2026},
  url={https://github.com/typerighter}
}

Contributions

Thanks to the user who generated this dataset using TYPE/RIGHTER.

Error Distribution

Keyboard Adjacent Errors: 20%
Case Corruption: 15%
Symbol Insertion: 10%
OCR Line Noise: 10%

Generated on

2026-01-07T01:04:58.601Z

Generated by TYPE/RIGHTER • Offline-First PWA • Privacy by Design • webXOS 2026