Grammatical inference: strengths and weaknesses

Grammatical inference (GI) refers to the inference of a set of grammatical rules that defines a formal language from queries or data about that language. Most prominently, these
grammars are either regular or context-free, and most often defined on either strings,
trees, or graphs – discrete, structured data. Though there are several advantages to this
form of machine learning, including generally small, transparent and efficient models,
there are also several drawbacks that impede their use outside some very specific contexts, including sensitivity to noise and inference processes that are hard to parallelise.
This extended abstract explores some of these strengths and weaknesses.