Model Suggesting Synonyms as Spelling Errors: Seeking Guidance on Strict Spelling Error Detection
I’m using the vennify/t5-base-grammar-correction model for spell-checking, but it’s suggesting synonyms as spelling errors. For instance, the model suggested changing "notice." to "attention." in my text, which seems to be more of a context-based suggestion rather than a true spelling error.
Example:
Original: "Please take notice of the changes."
Suggested Correction: "Please take attention of the changes."
In this case, "notice" and "attention" are synonyms, and "attention" is not a correct spelling correction. I’m looking to refine the model or processing to focus strictly on genuine spelling errors and avoid such context-based suggestions.
How can I achieve this? Any tips on filtering, adjusting model settings, or using a custom dictionary to ensure only actual spelling mistakes are flagged would be greatly appreciated.