Romance languages, Italian: deciding between multiple functions of _che_ (i.e. subordinating conjunction VS relative pronoun VS determiner, etc.)

[francescomass]

Italian: deciding between multiple functions of che

In Italian che expresses multiple functions (i.e. che 'subordinating conjunction' e.g. voglio che...; che 'non-oblique relative pronoun' e.g. la persona che amo; che 'interrogative determiner' e.g. che libro compri?; che 'interrogative pronoun' e.g. che dici?; che 'adverbial modifier' e.g. che bello!), so its disambiguation can cause problems with automatic parsing.

While reviewing our treebank containing texts with idiomatic expressions in Arborator, my colleagues (@ellepannitto & @fla-pi) and I came across rather common pattern in spoken Italian, in which deciding among the different functions of che is especially challenging (not only with respect to automatic parsing, but also in terms of making a theoretically grounded decision). As other Romance languages share a similar situation (e.g. Spanish que, despite orthographic norms), this issue is mostly a question to anybody who's already come across this kind of problem in syntactic parsing, to anybody who's observed similar patterns in other languages, or to anybody willing to provide us with their insight!

che X che Y Construction

I'll be referring to the pattern we came across as the che X che Y construction, in instances such as che caldo che fa!, che bella che sei! or che bene che ti voglio!

While the propositional meaning for intances such as these ones could be expressed respectively as something like 'it is very hot', 'you are very pretty', 'I love you very much', it is clear that the costruction conveys further information (be it pragmatic or else, this is not the aspect we're mostly concerned with at the moment), which, along with its form, seems to account for the idiomaticity of the construction.

As for information structure, the construction appears to me as focus initial expression with postponed presupposition, i.e. in che bella che sei! 'you are very pretty!', the compliment comes first and the person to which the compliment is being payed follows (literally being something like 'how pretty that (you) are!').

Examples

Depending on what fills up the X and Y slots in the construction, we have identified at least 2 sub-patterns, which in my opinion call for a different syntactic analysis (at least for formalization purposes).

In pattern #1, X is a NOUN and Y is a VERB (other than essere 'to be') or a whole clause. This scenario doesn't seem to cause significant problems, as both the parser and our review seem to find reasonble resorting to relative clauses (much like for cleft-sentences). Here is the CoNLL-U of one instance revised by me, where the first che is treated as a determiner of the the noun in the X slot (sole), and the second che is treated as a relative pronoun, subject of the clause containing the verb in the Y slot (splende), all modifying the noun in the X slot (sole).

Does this solution appear reasonable?

# sent_id = cxn_135_example_2_2
# text = Guarda che sole che splende sopra di noi Nicole meraviglioso non trovi?
# translation attempt = "Look at how shiny the sun above us is, Nicole... Wonderful, don't (you) agree?"
# literal translation = "Look at what sun (the one) that shines above us, Nicole... Wonderful, don't you find (it)?"
# validated_by = francescomass
1	Guarda	guardare	VERB	V	Mood=Imp|Number=Sing|Person=2|Tense=Pres|VerbForm=Fin	0	root	_	_
2	che	che	DET	DQ	PronType=Int	3	det	_	_
3	sole	sole	NOUN	S	Gender=Masc|Number=Sing	1	obj	_	_
4	che	che	PRON	PR	PronType=Rel	5	nsubj	_	_
5	splende	splendere	VERB	V	Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin	3	acl:relcl	_	_
6	sopra	sopra	ADV	B	ExtPos=ADP	8	case	_	_
7	di	di	ADP	E	_	6	fixed	_	_
8	noi	noi	PRON	PE	Number=Plur|Person=1|PronType=Prs	5	obl	_	_
9	Nicole	Nicole	PROPN	SP	_	12	vocative	_	_
10	meraviglioso	meraviglioso	ADJ	A	Gender=Masc|Number=Sing	12	xcomp	_	_
11	non	non	ADV	BN	PronType=Neg	12	advmod	_	_
12	trovi	trovare	VERB	V	Mood=Ind|Number=Sing|Person=2|Tense=Pres|VerbForm=Fin	1	parataxis	_	SpaceAfter=No
13	?	?	PUNCT	FS	_	1	punct	_	_

In pattern #2, X is usually an ADJ and Y is the VERB essere 'to be'. This scenario is where doubts start to rise. Here's an example of the automatic parsing for instances like this (this is NOT the tree validated by me, as the purpose of this issue is mainly to help me decide how to finally validate such trees). The first che was once again treated as a determiner, however it appears to me as more of an adverbial modifier of the ADJ in the X slot (grande), given that the meaning of che grande would be equivalent to quanto grande, where quanto is definitely an interrogative adverbial form (like English 'how'). The treatment of the second che is even less acceptable to me, as it is assigned the edge "nominal subject", implying that the form of essere in the Y slot is its copula, coordinated with grande in the X slot.

Not only there does not appear to be any coordination here, but a syntactic subject of essere could be expressed along with che (i.e. che grande che è quello), making it unreasonable to consider che as the subject and therefore as a pronoun (in fact, in our repository we do have instances such as che brutto che è quel maglione 'that sweater is really ugly/how ugly that it is that sweater' where an "actual" subject naturally occurs, i.e. quel maglione 'that sweater').

What could be a better solution here?

# sent_id = cxn_84_example_2_1
# text = che grande che è
# translation attempt = "How big is it?!? (rethorical)", "Wow, it's really big!"
# literal translation = "How big, that (it) is!"
1	che	che	DET	DQ	PronType=Int	2	det	_	_
2	grande	grande	ADJ	A	Number=Sing	0	root	_	_
3	che	che	PRON	PR	PronType=Rel	4	nsubj	_	_
4	è	essere	AUX	VA	Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin	2	cop	_	SpaceAfter=No

So Far

Likely biased by the compositional meaning of the sentence and by the means I would use to express in English the vibe it gives off, so far I have unelegantly been treating instances like this one as a single sentence (i.e. no subordinate clauses), where the ADJ is the root, essere is the copula and the two che are both adverbial modifiers, the first one modifying the ADJ in the X slot, and the second one modifying the copula itself...

Alternatively, influenced by cases such as che peccato che sia gay! 'what a pity (the fact) that he is gay', where the second che seems to simply be the subordinating conjunction introducing a clause called for by the semantics of the element in X peccato 'pity' (or, arguably, a full clause, argument of an additional noun before che, such as il fatto as in 'the fact that'), the edge "mark" could be used for the second che, as to allow us to cosider it as a SCONJ and not as a PRON, which still feels unacceptable to me here. This solution is nonetheless unelegant, since using "mark" implies the presence of some kind of clausal entity, which fits well the example che peccato che sia gay! but not the example che grande che è, still in need for a better formalization.

Please share any insight or link your project here if you have already observed similar constructions in other languages!

Finding More Examples

The name of our treebank on Arborator is ItCon-examples.
I've isolated the trees concerning this issue in the sample che_X_che_Y.

[Stormur]

The first observation is that I would say that we observe two che in Italian: a DET and an SCONJ, which is a complementiser. They are sufficient to explain all its occurrences. There could be other, "reductionist" analyses (i.e. there is only one che), or they could be seen as two identical, but etymologically different lexemes, but sticking to this I see two main points:

• the "interrogative" and the "adverbial" che are really the same thing. They follow the same distribution of other relative elements like chi, quale, etc. Being "interrogative" or "exclamative" (I would argue this is a better description than "adverbial") is not a feature of the che, but of the clause;
• it has been shown at length in other issues and in literature that the che in "relative clause" is not a pronoun but a generic complementiser which is pretty pervasive in modern Italian. It has no other particular semantic or syntactic values.

The first point seems to be more or less in line with the analysis you suggest.

---

(If I can ramble on about this...

Many traditional distinctions are not principled on morphosyntax, but on pragmatics, postulated semantics, and so on, and they look pretty aprioristic to me (e.g. "a so-called relative clause must have a pronoun referring to its head" - which by the way cannot be less true universally and is based on Latin where you do have an inflectable relative element qui, from which che is probably derived). Also the distinction made by some between a generic che and a "ché" staying for perché is a kind of hallucination. The che is simply a generic complementiser and as such is semantically underdefined (similarly to how in Latin an ablativus absolutus can be interpreted in different ways). It can also be noticed that perché is really just per + che (I had a further reference for this but fail to remember the author and the book...). )

---

In che caldo che fa and similar expressions we see both in use: the first is a DET of caldo, the second is a complementiser caldo. I agree with the "postponed presupposition" analysis, which boils down to a cleft clause: cf. nsubj:cleft (guidelines are only for Latin). I would warn against treating it as a relative clause of the first element: this is an error I see often with cleft clauses. The relative structure is just used to be able to use a clause as a subject for what remains after the focalisation, but I do not see it modifying anything.

In an example like guarda che sole che splende... I would venture saying that che sole che splende... is a ccomp of guarda. It is like considering a "basic" guarda che splende il sole, with the focalisation in the ccomp. In thsi the structure is parallel to something like guarda quanto splende il sole (or even guarda quant'è che splende il sole, or guarda quanto che splende il sole...).

The first che was once again treated as a determiner, however it appears to me as more of an adverbial modifier of the ADJ in the X slot (grande), given that the meaning of che grande would be equivalent to quanto grande, where quanto is definitely an interrogative adverbial form (like English 'how').

In principle you can use a DET with advmod (meaning that the validator will not complain). But probably there is no reason for it: the adjective can stand on its own without a modified head. The same could go for quanto. I am not sure that che grande and quanto grande are the same thing. For example, you cannot say che è grande as quanto è grande. This makes me actually doubt that the first che is a DET at all, and not again just a generic, pervasive complementiser.

The treatment of the second che is even less acceptable to me, as it is assigned the edge "nominal subject", implying that the form of essere in the Y slot is its copula, coordinated with grande in the X slot.

That is for sure a complementiser, as in all "relative" clauses. In your following analysis you do highlight how an interpretation as PRON fails.

What could be a better solution here?

# sent_id = cxn_84_example_2_1
# text = che grande che è
# translation attempt = "How big is it?!? (rethorical)", "Wow, it's really big!"
# literal translation = "How big, that (it) is!"
1	che	che	DET	DQ	PronType=Int	2	det	_	_
2	grande	grande	ADJ	A	Number=Sing	0	root	_	_
3	che	che	PRON	PR	PronType=Rel	4	nsubj	_	_
4	è	essere	AUX	VA	Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin	2	cop	_	SpaceAfter=No

The various cases are not different, the analysis can be always the same. Here the corresponding clause is è grande, and when you focus on grande there is nothing left for the cleft clause than the auxiliary è, promoted to head because of ellipsis. It can happen!

---

Final main takeaway point: the "relative che" is an SCONJ. There is no such thing as a "pronominal che".