In this case, both fair and fare are words in English. Which shows that spell checking needs to know a lot about grammar and context to work in general. Basically you need an LLM. Or if not a 'large language model', perhaps at least a small language model.
I wonder how it does work, I remember MS Word having a fairly decent grammar checker when I was using it in school - which predated LLMs by many years!
I suspect an LLM wouldn’t be the most optimal choice
In this case, both fair and fare are words in English. Which shows that spell checking needs to know a lot about grammar and context to work in general. Basically you need an LLM. Or if not a 'large language model', perhaps at least a small language model.
I wonder how it does work, I remember MS Word having a fairly decent grammar checker when I was using it in school - which predated LLMs by many years!
I suspect an LLM wouldn’t be the most optimal choice
Depends on what you mean by 'optimal'. Ie what are you optimising over?
In terms of 'can I run it locally on an early 2000s machine?' LLMs are definitely the wrong choice.
In terms of 'what can I quickly hack together in 2025 regardless of variable cost?' LLMs might be the right choice.
> I wonder how it does work, I remember MS Word having a fairly decent grammar checker [...]
You can get pretty far with some lookup tables and some heuristics.
farely decent?