What are the fundamental differences between Natural Language Processing and Computational Linguistics?

Question

I have a vague knowledge regarding those two fields, but I admit there are some fundamental concepts that I lack.

So, if we had to write down the actual differences between these two fields, what would they be?

I'll suggest some points that I think the answers should cover for successfully describing each field in a complete and comprehensive way:

• What it does and what it is about (also what it's not);
• Common misconceptions of the field (with consequent debunking);
• Aims/objective of the field;
• Tools/instruments and methods adopted by the field;
• Subfields of each field (if any);
• Any other points I might have forgotten.

Answer 1

I have a PhD in computational linguistics. I can tell you that NLP and CL are not two separate fields. Rather, CL is the superset that encompasses NLP.

In everyday CL practice, NLP focuses on the building of NL parsers and as such it is central to the CL field. CL as a field includes a lot more than NLP. For instance, you can study machine translation, knowledge representation, ontology engineering, text mining, information extraction, etc. all within the CL field. CL is a pretty broad thing and (unlike CS) is not primarily focused on theory. It is highly hands-on. Most theories in CL come from theoretical CS. When it comes to the nitty-gritty CL is the practical application of various algorithms for purposes of natural language processing.

You may occasionally encounter a reference to NLP (sans CL) within the field of CS. This is due to the fact that -originally- the generation of parsers served purposes beyond the confines of natural language (the way we mean "natural language" within CL). So, one could argue that NLP within CS is a slightly different animal than NLP within CL. In essence, it's the same kind of object seen under slightly different light.

Answer 2

The above answers are all good. I'd like to offer another perspective that I learned while teaching digital libraries that draws on the analogy used in biology:

Computational biology = the study of biology using computational techniques. The goal is to learn new biology, knowledge about living sytems. It is about science.

Bioinformatics = the creation of tools (algorithms, databases) that solve problems. The goal is to build useful tools that work on biological data. It is about engineering.

To make the analogy for any field X, we thus have "Computational X" and "X-omatics". In NLP/CL, NLP is the equivalent of "Linguamatics".

I don't really subscribe to the notion that CL encompasses NLP or vice versa. They both have a purpose. CL studies human language to computationally understand how we as humans have the capacity to produce and understand language. NLP takes a more pragmatic perspective and says that we wish to build systems that facilitate some language interface.

Answer 3

The difference is that Computational Linguistics tends more towards Linguistics, and answers linguistic questions using computational tools. Natural Language Processing involves applications that process language and tends more towards Computer Science.

However, the distinction between the two terms is fading and they are being used more and more interchangeably.

Answer 4

The terms are often used interchangeably. More often than not*, NLP is used when there is actual processing involved, rather than abstract theoretical topics. Comprehensive websites such as ACL wiki page write both separately, or combine them into CL/NLP or NLP/CL.

* - just my impression

Perhaps Computational Psycholinguistics is part of Computational Linguistics, but not a part of Natural Language Processing.

Answer 5

Some time has passed since the original question. For people looking at this question now, I'd say "Computational Linguistics" certainly encompasses more researches related to humanities, such as historical linguistics. A parallel would be the computational approaches employed in sociology or history research - Fundamentally they are still sociology and history nonetheless. "NLP" is much more closely related to computer science. Especially now in 2017, NLP is almost exclusively about how deep learning helps with various practical tasks such as speech recognition, machine translation etc.

Another illustration of the matter would be that NLP is mostly a research direction under various Computer Science departments, especially in the US, while many "Computational Linguistics" research is actually under the Linguistic department/Philosophy faculty and the degree program confers a Bachelor of Arts degree, especially in many European universities.

Of course the boundary might not be that clear and many universities which offer Computational Linguistics degrees have long been involved in more "NLP-esque" aspects of research as well and have been hiring professors with CS background. But the above generalization still applies.

Answer 6

A parts of the Linguistics community considers "Computational Linguistics" to have a very narrow scope about how the human brain computes during the processing of language, what are the limits of this processing, and how can we create tests that measure these limits of the human brain. Sometimes this involves computer simulation to test a theory. This community would say that Artificial Intelligence that uses computers to process language should be called "Computer Linguistics" or "Natural Language Processing".

The mainstream point of view, however, is the one already given in previous answers: "Computational Linguistics" is the same as "Computer Linguistics" and contain "Natural Language Processing", with the difference that the latter operates on a much more shallow representational level than the former.