6
  1. Go to this website: http://parts-of-speech.info/
  2. Now go to this Wikipedia article: https://en.wikipedia.org/wiki/Evolutionary_psychology
  3. Grab the first paragraph and feed it into POS tagger to see the tags. This is an image of the result:

parts of speech tagging result

However, as human beings, to understand this text, we need to understand some concepts that are denoted by more than one word:

  • Evolutionary psychology
  • Natural science
  • Natural selection
  • ...

These are technical terms that mean more than simply a bunch of words and they are not tagged. For example, if you know natural and selection, you don't necessarily know natural selection.

Is there a technical term to refer to concepts which are composed of more than one word? Also, is there a way to identify them in a given text? In other words, what is the name for technical terms in Linguistics, and how does find technical terms using NLP algorithms and techniques?

  • 1
    As an aside, for the specific examples you've given, "compound noun" is the common colloquial usage. – JAB Jun 9 '17 at 19:14
  • You know, there's more than one correct answer. You should accept one of them as "correct" – Belphegor Jun 13 '17 at 10:05
4

There are several terminologies for such words. Here are their definitions found in Wikipedia:

Multi-word Expressions (MWEs): A multiword expression (MWE), also called phraseme, is a lexeme made up of a sequence of two or more lexemes that has properties that are not predictable from the properties of the individual lexemes or their normal mode of combination. [URL]

Terminology: the study of terms and their use. Terms are words and compound words or multi-word expressions that in specific contexts are given specific meanings—these may deviate from the meanings the same words have in other contexts and in everyday language. [URL]

Collocations: In corpus linguistics, a collocation is a sequence of words or terms that co-occur more often than would be expected by chance. [URL]

Note that all of these are similar. Both "collocations" and "terms" are "MWEs". The terms are usually MWEs that are specific for a given topic, while collocations are MWEs which co-occur together in a corpus more frequently than by chance. Collocations and terms often coincide (you can always use a predefined MWE dictionary with terms/collocations).

If you want to discover MWEs in an unsupervised manner by using the corpus statistics - you should use collocation extraction techniques. There have been plenty of work on collocation extraction, so it should be easy for you to find software or some APIs.

Here's a list of tools (feel free to edit the list):

In my opinion, a nice reading resource is the paper "50-something years of work on collocations".

| improve this answer | |
4

You are dealing with terminology (a discipline of applied linguistics) and term extraction. You are looking for terms. The quoted wikipedia articles will give you some starters for further reading.

| improve this answer | |
2

These are called multi-word expressions (MWEs). Before we put things into a POS tagger, we need to perform another process known as tokenisation to separate characters into words. This is especially important in Chinese, where there are no whitespaces at all, but is also needed in English, where strings surrounded by word boundaries (in the sense of regex) are not always words either.

For solutions, you can search on Google scholar; there is a huge literature on that. :)

| improve this answer | |
1

You may also want to look into Idioms. An Idiom is a phrase (series of words, usually) whose meaning is not composed of its parts, at least not without a layer of metaphorical or (usually lost) historical meaning. For example,

  • The fly [kicked the bucket] (to mean the fly died)

is an idiom since knowing what [kick] and [the bucket] mean does not allow one to derive what [kick the bucket] means. It may be productive to look into whether or not [natural science] can be derived from what [natural] and [science] means in the technical sense.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.