The acomp and ncomp indices in TAASSC

[Maria Goldshtein]

According to the Kyle 2016 paper, these are the complements of a sentence that are adjectives and nouns. e.g., :

She [looks]gov [beautiful]acomp
He [is]gov a [teacher]ncomp

My question is: would the complements of those adjectives and nouns also be within the acomp/ncomp brakcets for the analysis?

e.g., would TAASSC analyze the scope of the acomp below as (a) or (b)?

(a) Anna is [delighted]acomp that she no longer has to take any classes
(b) Anna is [delightedthat she no longer has to take any classes]acomp

Thanks in advance!

maria

[Maria Goldshtein]

I tried using  https://corenlp.run to test this, and all adjective-headed clausal structures were tagged as ccomp, and not acomp. 

[Masaki Eguchi]

Hi Maria,

I am posting this again because I keep forgetting to reply to all (and additional content).

The version (and the tag sets) of the Core NLP in TAASSC will be different from the demo I shared. You will be able to see how TAASSC is tagging your example by outputting the parsed XML, I believe.

You are right in saying the adjective-headed clausal structure is ccomp. The adjective “delighted” would receive acomp (as dependent) if it is the acomp of the overall clausal construction, but tag sets can be updated and there are different flavors in how these are operationalized. Sometimes they differ depending on the tag scheme the specific Core NLP version had been trained on. The current version of the demo likely uses Universal Dependency version 2.

https://universaldependencies.org/u/dep/index.html

Edits:

If I remember correctly, the version of Core NLP TAASSC is using relies on Stanford dependency representation. Then it makes sense that acomp is not shown because UD does not have it, but the old CoreNLP does.

https://downloads.cs.stanford.edu/nlp/software/dependencies_manual.pdf

I hope that clarifies!

Thank you,

Masaki

--
You received this message because you are subscribed to the Google Groups "Suite of automatic linguistic analysis tools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linguistic-analysi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/linguistic-analysis-tools/b91d611e-8fe9-4927-843d-d05ba136a949n%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Maria Goldshtein]

Thank for the response, I just want to make sure that I understand correctly. 

Regardless of the details, the acomp and ncomp indices should represent an adjective/noun with dependents as different from one without, right?