I have been running through some examples of DG and from those, I have realized that we can safely remove modifiers from a sentence and still preserve the meaning of the sentence.

For Example, consider the sentence:-

1) took forever, 3 calls in 30 minutes for the waiter to come pick up the table after we finished eating room service

2) after placing my order, I had to wait another 10 minutes for my drink to arrive.

In the first sentence, the tokens or phrases: 3, 30, forever, "after we finished eating room service" ---- are nummod, nummod, advmod and advcl respectively in DG. removing them from the sentence furnishes - "took calls in minutes for waiter to come pick up the tables"

and similarly in second sentence, the tokens or phrases: "to arrive"--- are acl in its DG removing them from the sentence furnishes - "after placing my order, I had to wait another minutes for my drink"

we can observe that removing the modifiers still address the underlying concern that "responsiveness of the waiter was poor"

can we assume this thing in majority??


The question actually concerns adjuncts. Adjuncts are optional constituents that can be removed from a sentence without rendering the sentence ungrammatical. The modifiers mentioned in the question are adjuncts for the most part and can hence be removed from the sentence without causing the sentence to become ungrammatical. The term modifier and the term adjunct overlap in meaning and use. However, the extent of the overlap may or may not be complete.

Another example can help make these notions clearer:

(1) Yesterday, we repeatedly attempted to visit the new brew pub in our neighborhood.

The bolded constituents are all adjuncts or parts of adjuncts. They can be omitted from the sentence without causing ungrammaticality:

(2) We attempted to visit the pub.

If any of the other words are omitted, however, the sentence becomes incomplete and hence ungrammatical, e.g.

(3) *We to visit the pub.

(4) *We attempted visit the pub.

(4) *We attempted to the pub.

(5) *Attempted to visit the pub.

(6) *We attempted to visit.

(7) *We attempted to visit pub.

Based on the examples, we can conclude that each of the words We, attempted, to, visit, the, and pub is NOT an adjunct or part of an adjunct.

Note how the determiner the behaves in this example; it cannot be omitted, which indicates that it should NOT be viewed as an adjunct. I think many grammarians would, however, view determiners such as the as modifiers, so we see that for these grammarians, the overlap (adjunct vs. modifier) is indeed not complete.

Concerning the DG analysis of these matters, many DGs distinguish between arguments/complements and adjuncts/modifiers in their inventories of syntactic functions. We see an example of this in the question where the functions nummod, nummod, advmod, and advcl are mentioned. These designations are clearly those of adjuncts.

A word or two of caution are also appropriate. The answer here is simplifying matters somewhat, for there is a further important distinction, that between obligatory and optional arguments/complements. An optional argument can be overlooked and incorrectly taken to be an adjunct. Yet another important point is that the adjunct vs. argument/complement distinction is not unique to DG; it exists in most theories of grammar.

In sum, the answer to the question is "Yes", although there are certain caveats to be aware of concerning the use of terminology.

  • so simply put, can we safely say that most modifiers come under adjuncts or not? Jun 9 '20 at 18:39
  • I think many view most modifiers as adjuncts. However, it may not be the case that all adjuncts are modifiers. Which DG are you reading? It should provide some guidance about this. Jun 10 '20 at 1:32
  • Here is the link of dependency parser I am reading :- github.com/clir/clearnlp-guidelines/blob/master/md/… Jun 10 '20 at 3:34
  • Thanks, I've taken a quick look. The inventory you have linked to does not seem to use the term adjunct. But I think most of the mod labels do indeed correspond to adjuncts, so your assumption in the question is mostly correct. I can comment, though, that the dependency parses produced by ClearNLP are going to be controversial from a linguistic point of view. I took a quick look at one of the academic papers linked to in the Github page for ClearNLP. The parse I saw there is not good, i.e. not well-motivated from a linguistic point of view. Their are certainly better parsers out there. Jun 10 '20 at 3:56
  • Understood, right now I will have to make do with this parser only, but I will keep your points with me and gonna use them in future. Thanks a lot Jun 10 '20 at 4:20

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.