metadata
license: apache-2.0
Simple autocorrect dataset, dataset format:
<|start|><|user|>misspelled<|user|><|model|>correct<|model|><|end|>
Rough dataset word composition:
- Manual / basic : 1,386
- Wiki phrases : 9,975
- NLTK words : 88,639
- Total words : 100,000
Wiki phrases and Manual / basic can include more words, wiki 1 to 10 words per example, manual / basic 2.
Words / sentences ware misspelled using nearest key match, double tap, missing char, swap 2 chars, numbers, random insert Examples per word / sentence (misspells per word / sentence) : 164
Its recomended to use atleast 5k examples, the examples ware alredy randomly sorted.
Examples:
<|start|><|user|>cokmonty<|user|><|model|>commonty<|model|><|end|>
<|start|><|user|>hlm local sponsor pound pound which paid<|user|><|model|>him local sponsor pound pound which paid<|model|><|end|>
<|start|><|user|>HHardenbrgia<|user|><|model|>Hardenbergia<|model|><|end|>
<|start|><|user|>draftswomanshi<|user|><|model|>draftswomanship<|model|><|end|>
<|start|><|user|>xmieter<|user|><|model|>dimeter<|model|><|end|>
<|start|><|user|>cvaustery<|user|><|model|>clustery<|model|><|end|>
<|start|><|user|>carpospeerm<|user|><|model|>carposperm<|model|><|end|>
<|start|><|user|>cinsenter<|user|><|model|>consenter<|model|><|end|>
<|start|><|user|>into the game the wait `won<|user|><|model|>into the game the wait won<|model|><|end|>
<|start|><|user|>he=exx<|user|><|model|>hex<|model|><|end|>