Fingerprint for Relative Clause

Hello everyone! I am a postgraduate student with a computational background, and I just happened to learn anything about linguistics and the ERG. It has really taken me a long time to understand different basic linguistical knowledge, and even the more advanced ERG by going through page by page of and also Finishing documentation is a long way to go, and I would appreciate your help if professors or researchers here could advise me on and guide me through the learning process.

Here is one of my confusion. I read that the fingerprint of finite relative clause can be identified by (from:
h1:[ARG0 x]
h1:[ARG0 e{TENSE tensed}]
However, I found an example that I am confused:
“She is the girl who hadn’t arrived.” with the DMRS:
As seen from the DMRS, after the neg predicate (or generally any scopal modifier?) is added, the neg takes the same label as _girl_n_1. As a result, _arrive_v_1 does not share the same label (which means different scope?) with _girl_n_1. Moreover, neg is untensed so the fingerprint is nowhere to be found in this example.

Currently, I find it the hardest to understand the concept of “scoping”, esp. when a normal sentence can have up to thousands of readings with different scoping and it is just impossible to understand each of them. I have read about “quantifier scoping”, but when it comes to the hole-handle relations between non-quantifiers, or even verbs in MRS, things just complicate. Any guidance on these? Thanks in advance for your help!

1 Like

This fingerprint doesn’t match all relative clauses. (In fact, the page you’ve linked to also mentions that it doesn’t match untensed relative clauses.)

There are two important points about this fingerprint. Firstly, the two predicates share a label. In non-scopal modification (which is the normal kind of modification), there will be a shared label. In DMRS, you will see a /EQ link. Secondly, there is a tensed predicate. This rules out adjective modification. For example, the difference between “the big dog barked” (no relative clause) and “the dog that is big barked” (relative clause) is that _big_a_1 is untensed for “big dog” and tensed for “dog that is big”.

In your example, both of the above points are true: there is a sharing of labels (for _girl_n_1 and neg), and there is a tensed predicate (_arrive_v_1). The reason that the fingerprint doesn’t work is that the tensed predicate is not the one sharing the label. However, the two predicates are connected: we have neg scoping over _arrive_v_1.

neg is a scopal modifier because it’s a logical operator. There are other scopal modifiers like “probably” and “never”.

If you want to understand scope, start with simple sentences, so that you don’t have thousands of readings! For verbs which scope over their complements, see section 6.4 of Copestake et al. (2005).


Thanks for the reading suggestion! I will definitely take a look of it. Meanwhile here are two more questions I have in mind:

This fingerprint doesn’t match all relative clauses.
Do you mean that the fingerprint is reverse-engineered to account for relative clauses so it does not match all; or do you mean that it matches only finite relative clause, but not untensed ones, or extrapositioned ones (which I think is addressed by relative_mod)?

However, the two predicates are connected: we have neg scoping over _arrive_v_1 .
Can we say that then, if there exists one scopal modifier sharing the label with another predicate of type x, and that scopal modifier outscopes a predicate of type e with TENSE: tensed, then it is a finite relative clause? An example would be “She is the girl who probably hadn’t never arrived.”, where _probable_a_1 is the associated scopal modifier that connects (through multiple scopes) to the tensed predicate _arrive_v_1:
Thank you!

Yes, I guess you could say it’s reverse-engineered. A semantic fingerprint is not explicitly defined in the grammar.

Yes, that sounds right.

1 Like