LADChineseRelatives

Some issues raised in analysing Chinese Relative Clauses

ZhenzhenFan (FrancisBond)

We sketch an outline of how we handle relative clauses in Chinese, show that the simplest analysis calls for fewer quantifiers and suggest some strategies for dealing with this.

This is the third Virtual Linguistic Analysis Design Session. It will be held on Wednesday, March 1, 2016 at 00:00:00 UTC/GMT (Seattle 4pm (Tue), Singapore 8am, Berlin 1am)

Details on Overleaf.

1. Introduction We started with a seemingly unambitious task in building Zhong grammar - to model the relative clauses in Mandarin Chinese. Soon we found that we also need to cover other usages of DE. And finally we can’t run away from re-examining the implementation of NP structure in Zhong. 1Hence the need of this VLAD to gather more support in terms of linguistic analysis, HPSG implementation, as well as, moral support. The following topics will be discussed:1. Relative clauses in Mandarin Chinese and the general implementation in Zhong 2. The dilemma of determiner (SPR) in NP 3. Can we adopt a different approach? - Allow unspecified NP and use transfer rule to fix the semantics 4. Clarification seeking: LIGHT +/- as constraint to indicate word/phrase

2. Chinese Relative Clauses

2.1 Brief Overview One of the elements that can occur before a noun in a Chinese NP is a relative clause, which is a nominalized clause placed in front of the head noun, connected by the particle DE. The head noun refers to an omitted argument in the nominalized clause (subject or direct object). Some examples are shown below.

Gapped Object: The head noun refers to the omitted direct object of the nominalization.

Gapped Subject: The head noun refers to the omitted direct object of the nominalization.

What about indirect object? In Chinese sentences with ditransitive verbs (V+Obj1+Obj2), the first object is the indirect object and the second object is the direct object. When the indirect object is actually the head noun, its original position in the relative clause is typically replaced using a pronoun, as illustrated in (3). (Li & Thompson) This phenomenon is to be considered later.

2.2 Implementation: Relativizing DE In Zhong, two rules, ”extracted-comp-phrase” and ”extracted-subj-phrase” perform the task of extracting comp or subj for the verb, respectively. Their HEAD-DTR, the verb, doesn’t have constraint on MC, which allows the extraction to happen in both main clauses and subordinate clauses. The mother node has ”MC +”, since DE is defined to take comp with ”MC +” in its lexical type.

Notes from the Session

**LEGEND:<<BR>><<BR>>**ZZ: Zhenzhen FCB: Francis EMB: Emily LL: Liling Ping: Ping Xue Luis: Luis Joanna: Joanna Woodley: Woodley

Question 1: Is there a neater way to allow the head noun (modified by the relative clause) to be non-specified? Our existing implementation requires changes in many places with "SPR olist" constraint on the noun.

EMB: Regarding question you are posing. I would expect that the easier example is where the Rel. Cl. is closer to the head noun. So I'm surprised your (framework?) comes next in terms of making #5 possible instead of in terms of making #4 possible. ZZ: Yes, looking at the original general order, we have the CL phrase before the noun, and the assoc(iative) phrase is preceding that. The DE phrase is, in a way, similar to the assoc. phrase. So our previous grammar can easily allow that. Because it is expecting the specifier of the noun to be saturated first before combing with the associative phrase. In order to allow #5, we have to take away that constraint that the noun has to have the SPR satuated first. And we started doing lots of changes all over the place. FCB: And that chain that leads to is the complements... In the original grammar, complements must have a saturated specifier, extracted complement... so the thing it extracts must have a saturated specifier, and therefore we get #4 but not #5. EMB: That's probably the part of that chain that I would have done differently at the beginning. Instead of matching the whole SLASH value from the relative clause to the MOD value of DE, I'll just match the HOOK. All you really need to match there is the semantic information. If you look at analysis of relative clauses in other large grammars, you'll find that when you have something whose whole LOCAL value is matched with the SLASH, it's a relative pronoun. But there are no relative pronouns in Chinese, right? DE is not a pronoun. So, it is mediating the relative clause, but it is not actually using head noun as the full syntactic filler of the SLASH.

Joanna: There is a minor difference between #4 and #5. In #4, when you have rel cl. appearing before the DEM CL Noun sequence, you get a restrictive reading. In #5, when you have relative clause between the classifier, you can also have a non-restrictive reading as well. EMB: It will be so nice to have that go the other way! Joanna: When you have different types of relative clauses, when you have a stage-level and individual-level, you prefer to have a individual level between the CL and the Noun, and the stage-level to the left-hand side of the demonstrative. So there is some of preference for the kind of relative clause we like to have. Guy: Can clarify what is stage level? Joanna: Suppose you have two relative clauses, one stage level (talking about temporary properties), and the other individual level (permanent properties), the preference is to have the individual-level one to be in between CL and Noun, and more temporary properties to the left-hand side. So, in general, there is some kind of preference, in terms of the meaning of the relative clause. EMB: I've heard of this for adjectives, but not for relative clauses. So, a stage-level thing is "happy", as a state, versus "optimistic" which is more on an individual level. Joanna: Supposed we have something like "The tall boy I saw yesterday", you would like to have "I saw yesterday" to the left of the demonstrative, and "tall" between the classifer and the noun. EMB: Ok, so it's not the relative clause that's being individual or stage. It's if you have an individual-level adjective, it wants to go closer to the noun? Joanna: Yeah, that's what I meant. So when you have something that is more temporary, the you have want to have it, because adjective and relative clauses, in both cases, you will use DE. So, the order depends on stage- or individual-level. GUy: Will this apply to something you do regularly? "The person who eats rice", something which they do regularly. Does that count as an individual-level property? Joanna: When you have 1, then people say you can have it in either position. Except when you have 2, then people prefer the more temporary one. Guy: So, "The person who eat rice whom I saw yesterday." As in, he eats rice generally, but you saw that person yesterday/ Joanna: Ok, then I would prefer to say "I saw yesterday that eat rice man" Dan: I can provide more motivation for decoupling the value of the MOD LOCAL, and the value of SLASH. I think it is an illusion that they are the same, just because they are both nominal and they have the same referential index, that is, the same HOOK, as Emily as saying. It should at least be clear that in languages where case is marked, we can expect that case is the same in upper noun and the place where there will be a gap. Like when you say "The book that I saw impressed Mary", "the book" in the main clause must be in the nominative case, but the gap must be accusative. I know that Chinese doesn't have case marking, so we don't have that immediate evidence. But in the general design of how relative clauses work, we don't want to have that kind of structural identity between the noun being modified and the gap inside the relative clause. They are different syntactic structures. One is covered in the referential index that has to be identified. I think you will find it most troublesome to handle these constructions as described in #4 and #5.