https://github.com/willwade/dailydialogcorrections

Train a t5-small model to understand sentences without spaces and typos
https://github.com/willwade/dailydialogcorrections

Last synced: 3 months ago
JSON representation

Train a t5-small model to understand sentences without spaces and typos

• Host: GitHub
• URL: https://github.com/willwade/dailydialogcorrections
• Owner: willwade
• Created: 2024-02-05T11:03:58.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-02-06T09:37:31.000Z (over 2 years ago)
• Last Synced: 2025-12-30T22:33:13.159Z (7 months ago)
• Language: Python
• Size: 4.64 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Spoken "AAC-Like" corpora 

- We have dailyDialog and BNC2014 corpora extracted and tweaked

- Main tweaks - removal of double words like yeah and removal of speech components like 'mm' 

- Removal of quotes and other common written only aspects

- For the purposes of our tool we then add in typos and remove the spaces

- We then train a t5-small model on both these. Note we did it in two stages. First the dailyDialog (it has a lot of quotes in it). and next the BNC corpora