Automating QA rule creation

Good morning,

Today I would like to ask a language-independent question. Not a development question per se, but probably more aimed at developers than linguists.

Is there any effort or initiative to develop a system to create this kind of rules automatically based on a corpus of corrected texts? Does anyone follow or do research in this domain?

I suppose machine or deep learning could be used to infer rules based on a corpus where you have the original version of texts and their corrected version.

Have a good day!

Cheers, Manuel

You can Google for “grammatical error correction” to find recent research. The goal is usually not to create rules this way, but to directly get the correction. Basically the same way Google Translate or DeepL works with e.g. English/German pairs, but this time with incorrect/correct pairs.

A related approach is described here, but it doesn’t rely on incorrect/correct sentence pairs.