The landscape of NLP was very different in the beginning of the field.

“But it must be recognized that the notion ‘probability of a sentence’ is an entirely useless one, under any known interpretation of this term 1968 p 53. Noam Chomsky.

Probability was not seen very well (Chomsky has said many wrong things indeed), and linguists were considered useless. Recently deep learning and computational papers are ubiquitous in major conferences in linguistics, e.g. ACL.

One of the main aims of linguistics is understanding the structure of human language. How are we able to speak it so naturally, when we are not able to formally describe it? Other natural phenomenon are clearly described, but not language. And another thing is that nobody exposes these idea to young people during middle or high school! Even if it’s so ubiquitous. In Prof. Cotterell’s opinion, Linguistics is a science, even if it’s catalogued in humanities. Mathematics is very useful to study language (this is not very intuitive). For Cotterell, linguistics can be as formal as physics from a mathematical point of view. A weird thing is that linguistics PhD people usually have no college-math experience, and this is as if you are starting a physics PhD without any maths, in Cotterell’s point of view. The main difference is that the maths of linguistics is mostly discrete.

Chomsky proposed the idea of competence and performance. (Similar thing explained in (Mahowald et al. 2023)).

Competence asserts that there is a true grammar for a language. We have those grammars for programming languages, but we don’t clearly know for human languages. This is much studied especially for the study of compilers. And Chomsky 1943 McCullochs paper for perceptrons invented finite-state automata in the same paper!?!? The whole objective is trying to build this function that says yes or no for a syntactical correct sentence.

Performance is studied by psycholinguistists. They want to know how humans produce language. But humans usually don’t use the competence to produce their sentences.

What is NLP?

A set of methods and algorithms for making natural languages accessible to computers.

E.g. autocorrect, grammarly, machine translation, question answering, many many things, so a quite broad field.

What is linguistics?

Linguistics studies properties of languages. Computational linguistics uses techniques from computer science to study language. This is the main difference with NLP. In NLP computers are central, in computational linguistics they are aiding devices.

References

[1] Mahowald et al. “Dissociating Language and Thought in Large Language Models: A Cognitive Perspective” 2023