NLP researchers catalog how language describes language — a catalog of an emerging research direction

This paper surveys metalanguage research in NLP: how machine learning systems understand language that talks about language itself. The work maps four dimensions of metalinguistic tasks and identifies gaps where nobody has looked yet.

Most NLP research focuses on language that does something in the world — classify emails, translate text, answer questions. This paper asks a different question: what happens when language describes its own rules, meanings, or structure? That matters because understanding metalanguage is a prerequisite for systems that can reason about language, explain their own decisions, or learn from feedback about how they're using words. Right now, the field doesn't know which metalinguistic tasks actually scale.