{ "paper_id": "O06-5002", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:07:13.411401Z" }, "title": "Multiply Quantified Internally Headed Relative Clause in Japanese: A Skolem Term Based Approach", "authors": [ { "first": "Rui", "middle": [], "last": "Otake", "suffix": "", "affiliation": { "laboratory": "", "institution": "Tohoku University", "location": { "addrLine": "41 Kawauchi, Aoba-ku", "postCode": "980-8576", "settlement": "Sendai", "country": "Japan" } }, "email": "otake@linguist.jp" }, { "first": "Kei", "middle": [], "last": "Yoshimoto", "suffix": "", "affiliation": { "laboratory": "", "institution": "Tohoku University", "location": { "addrLine": "41 Kawauchi, Aoba-ku", "postCode": "980-8576", "settlement": "Sendai", "country": "Japan" } }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "This paper presents an analysis of Internally Headed Relative Clause (IHRC) construction in Japanese within the framework of Combinatory Categorial Grammar [Steedman 2000]. Shimoyama [1999] argues that when an IHRC appears within the scope of a universal quantifier, the interpretation of the IHRC exemplifies E-type anaphora and that the LF representation of the IHRC should have a variable bound by the quantifier in the matrix clause. To accommodate this argument Shimoyama posits a free variable of a functional type to which the bound variable is applied, and whose denotation is determined by the context-dependent assignment function. However, since there is in principle no limit to the number of quantifiers in the matrix clause (and accordingly that of bound variables in the IHRC), the semantic type of the free variable would be highly ambiguous if the IHRC occurs within the scope of multiple quantifiers. The current analysis assumes that the interpretation of IHRCs exhibits an instance of generalized Skolem term [Steedman 2005], a term whose denotation varies with the value of bound variables introduced by scope-taking operators, but which is interpreted as a constant in the absence of such operators. This paper provides a straightforward account for the semantics of the construction without invoking the complexities of the type ambiguity of free variables.", "pdf_parse": { "paper_id": "O06-5002", "_pdf_hash": "", "abstract": [ { "text": "This paper presents an analysis of Internally Headed Relative Clause (IHRC) construction in Japanese within the framework of Combinatory Categorial Grammar [Steedman 2000]. Shimoyama [1999] argues that when an IHRC appears within the scope of a universal quantifier, the interpretation of the IHRC exemplifies E-type anaphora and that the LF representation of the IHRC should have a variable bound by the quantifier in the matrix clause. To accommodate this argument Shimoyama posits a free variable of a functional type to which the bound variable is applied, and whose denotation is determined by the context-dependent assignment function. However, since there is in principle no limit to the number of quantifiers in the matrix clause (and accordingly that of bound variables in the IHRC), the semantic type of the free variable would be highly ambiguous if the IHRC occurs within the scope of multiple quantifiers. The current analysis assumes that the interpretation of IHRCs exhibits an instance of generalized Skolem term [Steedman 2005], a term whose denotation varies with the value of bound variables introduced by scope-taking operators, but which is interpreted as a constant in the absence of such operators. This paper provides a straightforward account for the semantics of the construction without invoking the complexities of the type ambiguity of free variables.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "This paper presents an analysis of Internally Headed Relative Clause (IHRC) construction in Japanese paying particular attention to the effect of quantification on its interpretation. (1) illustrates the basic form of the construction: 1", "cite_spans": [ { "start": 184, "end": 187, "text": "(1)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Taroo-ga [Hanako-ga ringo-o muita] no-o tabeta.", "cite_spans": [ { "start": 9, "end": 34, "text": "[Hanako-ga ringo-o muita]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Taro-NOM Hanako-NOM apple-ACC peeled NML-ACC ate 1'Taro ate the apple that Hanako peeled.'", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "The bracketed clause Hanako-ga ringo-o muita 'Hanako peeled an apple' is followed by the nominalizer no and the accusative particle -o, thereby construed as the object of the matrix verb tabeta 'ate'. Since the verb requires as its semantic restriction that the accusative argument be an edible thing, it anaphorically picks up the referent of ringo 'apple' from the embedded clause. This kind of construction is often contrasted with the Externally Headed Relative Clauses (EHRC), which is illustrated in (2).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Taroo-ga [Hanako-ga muita] ringo-o tabeta.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Taro-NOM Hanako-NOM peeled apple-ACC ate 2'Taro ate the apple that Hanako peeled.'", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "As can be seen from the translation, (1) and (2) have almost the same meaning. But we hasten to add that IHRCs are not always paraphrasable to the EHRC version, for the former construction is subject to some pragmatic condition for its felicitous use, which we will not attempt to specify. A terminological note: we use the term the antecedent of an IHRC to mean the referent of the IHRC which functions as the argument of the matrix predicate. And we also use the term the head of the IHRC to refer to the linguistic element in the IHRC which describes the antecedent. For example, the antecedent of the IHRC in (1) is the apple that Hanako peeled, and the head is the noun ringo 'apple'. It is important to notice that we define IHRC construction in terms of the nominal character of its anaphoric referent, despite the name suggesting that the presence of the head noun inside the relative clause is the defining feature of the construction. In fact, there are cases where the IHRC has no explicit nominal head. Such examples will be dealt with in section 2.2.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "This paper is organized as follows. In section 2, we review the observation made by Shimoyama [1999] and her E-type analysis of IHRC. We then address the problem that the E-type analysis would raise focusing on multiply quantified IHRCs. We also discuss the A Skolem Term Based Approach interpretational characteristics of IHRC, drawing on the study of Kikuta [2000] . Section 3 introduces the notion of generalized Skolem term, which provides a straightforward account for multiply quantified IHRC. Finally, section 4 concludes. Shimoyama [1999] claims that when IHRC appears within the scope of a universal quantifier, the interpretation of IHRC exemplifies E-type anaphora:", "cite_spans": [ { "start": 84, "end": 100, "text": "Shimoyama [1999]", "ref_id": "BIBREF7" }, { "start": 353, "end": 366, "text": "Kikuta [2000]", "ref_id": "BIBREF2" }, { "start": 530, "end": 546, "text": "Shimoyama [1999]", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Dono gakusei i -mo [soitu i -ga kongakki peepaa-o san-bon kaita] no-o", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "Every student he-NOM this:semester paper-ACC three-CL wrote NML-ACC kesa teisyutusita.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "this:morning turned:in 'This morning every student turned in the three term papers he or she wrote this semester.'", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "In (3), the subject of the embedded clause is bound by the universal quantifier in the matrix clause. Note that \"[t]he matrix object [...] does not refer to any particular set of term papers [Shimoyama 1999] ,\" and the interpretation of the matrix object can be paraphrased by the definite description the term papers he or she wrote this semester. The interpretation of the relative clause is an instance of E-type anaphora [Evans 1980] . Given this observation, Shimoyama [1999] proposes the LF structure of IHRC as schematized in (4). According to this analysis, CP in the Spec of DP corresponds to the relative clause, which moves up to some higher position and is interpreted independently because of type mismatch. The DP is headed by the nominalizer no. The complement NP indicated by e is an empty pronoun. Basically, the interpretation of this empty pronoun determines the argument of matrix predicate. For example, in the LF representation (5) of the IHRC in (1), the empty pronoun is represented as P, a free variable of type e\u2192t. Now, it is assumed that free variables in general are assigned a value by the assignment function g relative to the context c. In this case, P is assigned the value in (6): the set of apples which Hanako peeled. x apple x peel x hanako", "cite_spans": [ { "start": 133, "end": 138, "text": "[...]", "ref_id": null }, { "start": 191, "end": 207, "text": "[Shimoyama 1999]", "ref_id": "BIBREF7" }, { "start": 425, "end": 437, "text": "[Evans 1980]", "ref_id": "BIBREF1" }, { "start": 464, "end": 480, "text": "Shimoyama [1999]", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\u2032 \u2032 \u2032 = \u03bb \u2227", "eq_num": "(6)" } ], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "Shimoyama assumes that the nominalizer no is interpreted as the function from a set to the maximality of the set, adopting Link's [1983] result, the denotation of IHRC is equivalent to the English definite description the apples that Hanako peeled. In the case of (3), the LF representation of the IHRC is something like (7), in which the empty pronoun is further divided into pro and a free variable R.", "cite_spans": [ { "start": 123, "end": 136, "text": "Link's [1983]", "ref_id": "BIBREF4" } ], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "pro is a variable of type e bound by the universal quantifier in the matrix clause. On the other hand, R is a free variable of type e\u2192(e\u2192t), to which an assignment function g c assigns the value in (8) as a salient two-place predicate in the context.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "\u03bb .\u03bb . c g R", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "x y paper y wrote yx", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\u2032 \u2032 = \u2227", "eq_num": "(8)" } ], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "The latter takes pro as its argument to yield the interpretation shown in (9).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "( ) \u03bb . c c g g R pro y paper y wrote yx \u2032 \u2032 = \u2227", "eq_num": "(9)" } ], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "In words, this is the set of y such that y is a set of papers that x wrote, where x is bound by the universal quantifier. Then the result serves as the argument for no as before. In the end, (3) is interpreted as 'for every student x, x wrote three papers this semester, and this morning x turned in the papers x wrote this semester.' We thus get correct semantics. However, this analysis poses two problems. First, as is noted by Shimoyama herself, the assignment function is not properly constrained. It just picks up a property or relation which is 'salient' in the context. We will discuss this problem in section 2.2. We will assume that the appropriate constraints can be captured by the predicate result\u2032 or abt\u2032 although we will not be determinate on how to decide between these two options. The second problem is the ambiguity in the semantic type of the free variable. As we have just seen, the assignment function g c assigns a value to a free variable relative to the context. In order to assign a value, at least the semantic type of the variable needs to be known. However, the semantic type of that free variable can be determined only after the context is available. Then, the problem is that it is unclear as to how the context can be available before the context-dependent interpretation comes in. Furthermore, the semantic type of the free variable can be arbitrarily complex according to the number of universal quantifiers in the matrix clause. This is illustrated in (10-11): In principle, there is no limit to the number of quantifiers in the matrix clause. Therefore, a mechanism that does not invoke complexity of this sort would be preferred. In section 3, we propose an analysis that can derive the interpretation of such multiply quantified IHRC in the same way as the quantifier-free cases like (1) and singly quantified cases like (2) by introducing the concept of the generalized Skolem term proposed in Steedman [2005] .", "cite_spans": [ { "start": 1935, "end": 1950, "text": "Steedman [2005]", "ref_id": "BIBREF9" } ], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "Dono sensei-", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Shimoyama's [1999] E-type analysis", "sec_num": "2.1" }, { "text": "In order to state a proper constraint on the possible antecedent of an IHRC, let us examine the semantics of IHRC in more detail. The anaphoric nature of IHRC is best illustrated by the fact that the antecedent of an IHRC is occasionally not expressed as a linguistic element. (12) (13) (14) are such examples of 'headless' IHRC from Nomura [2000 ", "cite_spans": [ { "start": 277, "end": 281, "text": "(12)", "ref_id": null }, { "start": 282, "end": 286, "text": "(13)", "ref_id": null }, { "start": 287, "end": 291, "text": "(14)", "ref_id": null }, { "start": 334, "end": 346, "text": "Nomura [2000", "ref_id": "BIBREF6" } ], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "'I dug the soil, and looked into (the hole).'", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "Determining the antecedent of a headless IHRC obviously requires inference of some sort. Kikuta [2000] addresses the question of exactly how much information is needed for this inference within the framework of Generative Lexicon (GL). To recapitulate Kikuta's argument, if the inference always requires unconstrained pragmatics, the theory would overgenerate IHRCs with an illicit antecedent. Accordingly, there should be a more restricted way for determining the antecedent. And indeed she shows that the antecedent can be identified only by the linguistically specified information. Kikuta's conclusion is that the possible antecedent of the headless IHRC must meet the following conditions: (i) necessary involvement in the event described by the main predicate in the IHRC; (ii) recoverability of the denotation from linguistically pre-specified information; and (iii) presence in the resultant state of the process described by the main predicate in the IHRC. Note that for example (12), the antecedent is the water which is necessarily involved in the overflowing event, lexically specified by the predicate ahureta 'overflowed', and exists in the resultant state. Other examples can also be shown to satisfy these conditions. In contrast, violating these conditions leads to unacceptability. Following Kikuta's result (but simplifying the matter somewhat), in our analysis we will use a predicate result\u2032 which is defined such that result\u2032 px means that an element x is involved in the resultant state of the process or event described by the proposition p. However, the three conditions given above cannot straightforwardly be applied to the IHRCs in general. Consider (15), the example taken from a news article (asahi.com, June 22, The main predicate of the IHRC is nakizyakutteita 'sobbing', which denotes an activity, apparently lacks a lexical specification of the resultant (or consequent) state. Therefore, there is no way to satisfy condition (iii) . And yet the antecedent betuno otoko 'the other man' is an entity necessarily involved in the described event, in accordance with the above condition (i).", "cite_spans": [ { "start": 89, "end": 102, "text": "Kikuta [2000]", "ref_id": "BIBREF2" }, { "start": 1960, "end": 1965, "text": "(iii)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "In such cases, we will use a predicate abt\u2032 for 'aboutness' relation defined informally such that abt\u2032 px means that an element x is necessarily involved in the event described by the proposition p. Now we have two distinct relations (i.e. result\u2032 and abt\u2032) to describe the semantic constraint on the possible antecedent of IHRCs. Note that result\u2032 is a subtype of abt\u2032 since the former entails the latter. We will define the nominalizer category as initially having the interpretation containing the predicate abt\u2032, which is on occasion replaced with result\u2032. In the following, we give an informal sketch of how this replacement is done. However, in the analysis in section 3, we will take this replacement for granted, treating result\u2032 as if it is given lexically. Given the fact that an IHRC is syntactically a complete sentence, we can assume that an IHRC constitutes a separate information unit from the matrix clause. And we adopt the claim of Segmented Discourse Representation Theory [Asher and Lascarides 2003 ] that every information unit or proposition is connected via some rhetorical relation in order for the entire discourse to be coherent. Then the IHRC and the matrix clause must be connected via some rhetorical relation. By way of illustration, let us consider (12) again. Here, we have two clauses, namely the IHRC and the matrix clause, and the two corresponding propositions: the fish tank upstairs overflowed (\u03c0 1 ), and something leaked to downstairs (\u03c0 2 ). In the latter proposition, something corresponds to the nominative argument IHRC. Being anaphoric, this must be resolved in some way. In addition, the two propositions need to be connected via some rhetorical relation. Suppose that the latter requirement is somehow fulfilled by inferring Result(\u03c0 1 ,\u03c0 2 ) as the relevant rhetorical relation. Then the semantics of Result entails that the event of \u03c0 1 caused that of \u03c0 2 . Now, something in \u03c0 2 can plausibly be equated with the water, making the discourse (consisting of two clauses) coherent. We assume that the relation symbol abt\u2032 originating from the lexical information is replaced with the more specific relation symbol result\u2032 in the process of such inference. Note, incidentally, that this kind of approach may provide a way to account for the Relevancy Condition on IHRC discussed in Kuroda [1975-6] :", "cite_spans": [ { "start": 992, "end": 1018, "text": "[Asher and Lascarides 2003", "ref_id": "BIBREF0" }, { "start": 2328, "end": 2343, "text": "Kuroda [1975-6]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "For a p.-i. relative clause [IHRC] to be acceptable, it is necessary that it be interpreted pragmatically in such a way as to be directly relevant to the pragmatic content of its matrix clause.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "As the effect of the Relevancy Condition, events described by an IHRC and the matrix clause are typically related in any of the following terms: (i) temporal overlap, (ii) relevance of purpose, (iii) relevance of motivation, or (iv) spatial proximity. Note that Kuroda stated the Relevancy Condition as a mere descriptive generalization, whereas in our approach sketched above, this can be viewed as a consequence of the principle of discourse coherence in general, since if IHRC cannot be connected by some rhetorical relation, there would be no way to attach the information of IHRC to the matrix clause in a coherent manner. The effects of the Relevancy Condition mentioned above can also be regarded as entailments of the inferred rhetorical relation.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The Interpretational Characteristics of IHRC", "sec_num": "2.2" }, { "text": "In this section, we will introduce the concept of the generalized Skolem term proposed by Steedman [2005] . We briefly sketch the basic idea here and we will show in section 3.3 how to apply it to the analysis of (multiply quantified) IHRCs. The basic idea is this: a generalized Skolem term is a term whose denotation varies with the value of bound variables introduced by scope-taking operators such as universal quantifiers, but which is interpreted as a constant in the absence of such operators. One of the main motivations for such a mechanism is to give an analysis of the alternation of quantifier scope. By way of illustration, let us consider the sentence Everybody loves somebody. The narrow reading of somebody can be translated as ( )", "cite_spans": [ { "start": 90, "end": 105, "text": "Steedman [2005]", "ref_id": "BIBREF9" } ], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "( ) ( ) x person x person sk x love sk x x \u2032 \u2032 \u2032 \u2032 \u2032 \u23a1 \u23a4 \u2200 \u2192 \u2227 \u23a3 \u23a6", "eq_num": "(17)" } ], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "Note that sk\u2032 here is a Skolem function, and the referent of sk\u2032 x is dependent on who the variable x refers to. On the other hand, we get the wide scope reading of somebody by letting the Skolem term be a constant sk\u2032.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "( )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "x person x person sk love sk x \u2032 \u2032 \u2032 \u2032 \u2032 \u23a1 \u23a4 \u2200 \u2192 \u2227 \u23a3 \u23a6", "eq_num": "(18)" } ], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "The above transformations of Logical Form illustrate the standard Skolemization. In the current framework, however, indefinite noun phrases are interpreted as Skolem terms right from the start, and the nominal properties such as person\u2032 will be directly associated with them. More specifically, when introduced as a Logical Form element, a Skolem term is indicated as skolem\u2032p where p designates a nominal property. At this stage, this is unspecified as to its arguments. At some later step in the derivation, an operation called Skolem specification is applied to this unspecified Skolem term to yield a generalized Skolem term, designated as E p sk , where E is the environment, an ordered set consisting of the bound variables of the universal quantifiers that take scope at that point of the derivation. Since p sk is a function with E as its argument, its reference varies depending on the values of bound variables. And if E is the empty set, the E p sk will be a constant. The notion of environment is incorporated in the grammatical rule of Combinatory Categorial Grammar (CCG) in the following way. First, we have two function application rules: 4", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "X/Y : f \u2032 Y : a\u2032 \u21d2 X : f \u2032a\u2032 (>) Y : a\u2032 X\\Y : f \u2032 \u21d2 X : f\u2032a\u2032 (<)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "The environment is the operator bound variable identifier and is associated with the propositional body of the interpretation. For example, the transitive verb loves has the following category with its environment being the empty set: Application of the rule induces environment passing in the following way: if a function with environment F is applied to an argument with environment A, the environment of the argument in the resulting Logical Form is the union of the two (F\u222aA). We often omit the environment from the notation where it is of little interest. Let us now turn to the way this works. Expressions traditionally analyzed as an existential quantifier such as somebody are analyzed here as an unspecified Skolem Term skolem\u2032person\u2032. In the sentence Everybody loves somebody, for example, if specification applies at a point of derivation in which the Skolem term has not yet been in the scope of the universal quantifier as in (20), the resulting generalized Skolem term will be person sk \u2032 , hence we get the wide scope reading of somebody. In the derivation shown below, Skolem specification is indicated by the dotted underline.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "In contrast, if specification applies after it enters within the scope of the universal quantifier, the resulting generalized Skolem term is", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "( ) y person sk \u2032 , where person sk", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "\u2032 is a function which takes the variable y bound by the universal quantifier as its argument, hence the narrow scope reading of somebody. The following illustrates the derivation of this reading.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "(20) 21done by distributivizing verb categories. We assume that daremo 'everybody' denotes a set of individuals whose members are all the people in the universe of discourse, indicated as all-people\u2032. (25) illustrates the derivation of (24), in which dareka 'somebody' is interpreted as having narrow scope.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "Here, we defined the category of the verb so that each member of the set all-people\u2032 distributes over the Skolem term introduced by the subject NP. If we further assume that distributivizing as well as scrambling is realized by a lexical rule, then the fact that distributive reading is absent in a non-scrambled sentence can be thought of an accidental lack of such lexical rule. The derivation (26) illustrates the only possible reading for (23).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Generalized Skolem Term", "sec_num": "3.1" }, { "text": "The current analysis views the interpretation of IHRC as an instance of generalized Skolem term, and provides a straightforward account for the semantics of the construction. In order to capture the restriction on the interpretation of IHRC, the nominalizer category is defined as in (27), where R stands for either abt\u2032 or result\u2032 (see section 2.2 for the discussion of this treatment):", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "no : \\ : \u03bb . (\u03bb . ) NP S p skolem x Rpx \u2032 =", "eq_num": "(27)" } ], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "The derivation of (1) is shown in (28).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "A Skolem Term Based Approach", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "The semantics of the embedded clause is fairly obvious. This is then applied to the function defined in (27). The 'result' predicate result\u2032 px indicates that the event described by p yields an individual x. Thus the IHRC is analyzed as a Skolem term with the property of being involved in the result state of Hanako's peeling an apple. It is further subject to Skolem specification (indicated by the dotted line as before). Since there is no universal quantifier, the resulting generalized Skolem term is a constant whose referent is the apple that Hanako peeled. It should be obvious that this analysis also accommodates to the examples of headless IHRCs discussed in section 2.2. Interpretation of an IHRC is always determined via the abt\u2032 or result\u2032 relation, regardless of whether the antecedent is explicitly expressed or not. Still, we have to admit that the nominal property of the Skolem term \u03bbx. result\u2032(peel\u2032apple\u2032hanako\u2032)x is not a sufficient characterization of the antecedent, as it can also be applied to the apple skin, rather than only to the peeled apple fruit. Basically this \u03bb-expression should function as a constraint on the possible referent of the IHRC, and it is subject to the process of anaphora resolution. The resolved Skolem term in this case would be skolem\u2032\u03bbx. (result\u2032(peel\u2032apple\u2032hanako\u2032) x \u2227 fruit\u2032x). However, we will gloss over this problem here, and the question of how this mechanism works is left open for future research. Now, let us examine the interpretation of the IHRC that occurs within the scope of a universal quantifier. The derivation of (3) is shown in (29):", "cite_spans": [ { "start": 1293, "end": 1321, "text": "(result\u2032(peel\u2032apple\u2032hanako\u2032)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "Here, soitu in the IHRC is interpreted as pro\u2032u. As the notation may suggest, it is introduced as an ordinary pronoun. This will later get bound by the universal quantifier. 5 The last line is the result of Skolem specification. Since the Skolem term is in the scope of the universal quantifier, it takes the bound variable u as its argument. Therefore, the resulting generalized Skolem term refers to different sets of term papers, according to the value of u. This is the desired result, and this is achieved without invoking any complexity such as the type ambiguity of the empty pronoun. Multiply quantified cases like (10) can be derived in a similar way: 5 We take this process as anaphora resolution of the usual kind. This is because soitu could also be interpreted as deictic pronoun, in which case there is a ghostwriter (referred to by soitu) who wrote all the papers that every student turned in (a highly implausible situation, though).", "cite_spans": [ { "start": 174, "end": 175, "text": "5", "ref_id": null }, { "start": 661, "end": 662, "text": "5", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "IHRC as Generalized Skolem Term", "sec_num": "3.3" }, { "text": "Since Skolem term automatically takes the bound variables of the environment, the problem of type ambiguity that the E-type analysis suffered does not arise.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "(29)", "sec_num": null }, { "text": "We developed an analysis of Internally Headed Relative Clause (IHRC) construction in Japanese within the framework of Combinatory Categorial Grammar [Steedman 2000 ]. In section 2, we first looked at the argument made by Shimoyama [1999] and her E-type analysis of IHRC. We addressed the problem of type ambiguity that her E-type analysis would raise focusing on multiply quantified IHRCs. We then discussed the interpretational characteristics of IHRC, drawing on Kikuta's [2000] study of headless IHRCs. We generalized her idea by partly adopting Asher and Lascarides's [2003] theory, and argued that semantics of rhetorical relation would also help to determine the antecedent. And we also suggested that Kuroda's [1975-6] Relevancy Condition on IHRC can be viewed as a consequence of the principle of discourse coherence in general. In section 3 we introduced the notion of generalized Skolem term, and discussed the problem of scope alternation in Japanese, which is quite different from English, and motivated the distributivizing verb category adopting the Steedman's [2005] suggestion. Finally we integrated the whole discussion into the analysis of IHRCs. The main point is that, if we take the interpretation of an IHRC as a generalized Skolem term, we can attain the uniformed analysis for the several kinds of IHRCs, namely, headless IHRC, simple (quantifier-free) IHRC, and (singly or multiply) quantified IHRC. Our focus here was exclusively on the IHRC, but of course the approach here also applies to other kinds of noun phrases. However, the theoretical implication of this approach in other nominal construction is not entirely clear at this point, and is left for future work. 30", "cite_spans": [ { "start": 149, "end": 163, "text": "[Steedman 2000", "ref_id": "BIBREF8" }, { "start": 221, "end": 237, "text": "Shimoyama [1999]", "ref_id": "BIBREF7" }, { "start": 465, "end": 480, "text": "Kikuta's [2000]", "ref_id": "BIBREF2" }, { "start": 549, "end": 578, "text": "Asher and Lascarides's [2003]", "ref_id": "BIBREF0" }, { "start": 708, "end": 725, "text": "Kuroda's [1975-6]", "ref_id": null }, { "start": 1064, "end": 1081, "text": "Steedman's [2005]", "ref_id": "BIBREF9" } ], "ref_spans": [], "eq_spans": [], "section": "Conclusion", "sec_num": "4." }, { "text": "Abbreviations used: ACC = accusative, ALL = allative, CL = classifier, COMP = complementizer, COP = copula, GEN = genitive, LOC = locative, NML = nominalizer, NOM = nominative, TOPIC = topic.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "In the translation, antecedents are given in brackets.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "In this paper, we use only the application rules, which (apart from the notational convention) is common to all variants of Categorial Grammars. The reason for this choice is just the simplicity of presentation. For other rules of CCG, we refer the reader toSteedman [2000].", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [ { "text": "The research reported here was supported in part by the Tohoku University 21st Century Center of Excellence (COE) Program in Humanities, Research and Strategic Center for an Integrated Approach to Language, Brain, and Cognition (http://www.lbc21.jp/). We are grateful to two anonymous reviewers for their comments.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Acknowledgements", "sec_num": null }, { "text": "The scope alternation is also observed if the universal quantifier noun phrase appears in the object position. This is accounted for in a quite similar way to that shown above. The derivation below illustrates the inverse scope reading.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A Skolem Term Based Approach", "sec_num": null }, { "text": "Before going into the Skolem term approach to IHRC, let us consider how the scope alternation is achieved in Japanese. One striking fact about scope alternation in Japanese is that its availability is more restricted than in English. While in English the universal quantifier in the object position can take the inverse scope over the subject indefinite as in Somebody loves everybody, the Japanese counterpart does not seem to accept such reading. In 23 ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Distributivizing Verb Category in Japanese", "sec_num": "3.2" } ], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Logics of Conversation", "authors": [ { "first": "N", "middle": [], "last": "Asher", "suffix": "" }, { "first": "A", "middle": [], "last": "Lascarides", "suffix": "" } ], "year": 2003, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Asher, N., and A. Lascarides, Logics of Conversation. Cambridge: Cambridge University Press, 2003.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Pronouns", "authors": [ { "first": "G", "middle": [], "last": "Evans", "suffix": "" } ], "year": 1980, "venue": "Linguistic Inquiry", "volume": "11", "issue": "", "pages": "337--362", "other_ids": {}, "num": null, "urls": [], "raw_text": "Evans, G., \"Pronouns,\" Linguistic Inquiry, 11, 1980, pp. 337-362.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Qualia Structure and the Accessibility of Arguments: Japanese Internally-headed Relative Clauses with Implicit Target", "authors": [ { "first": "C", "middle": [ "U" ], "last": "Kikuta", "suffix": "" } ], "year": 2000, "venue": "The Proceedings of the 14th Pacific Asia Conference on Language, Information and Computation", "volume": "", "issue": "", "pages": "153--164", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kikuta, C. U., \"Qualia Structure and the Accessibility of Arguments: Japanese Internally-headed Relative Clauses with Implicit Target,\" The Proceedings of the 14th Pacific Asia Conference on Language, Information and Computation, Akira Ikeya and Masahito Kawamori (eds.), 2000, pp. 153-164.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Pivot-Independent Relativization in Japanese II", "authors": [ { "first": "S.-Y", "middle": [], "last": "Kuroda", "suffix": "" } ], "year": 1975, "venue": "Japanese Syntax and Semantics", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kuroda, S.-Y., \"Pivot-Independent Relativization in Japanese II,\" 1975-1976, Reprinted in S.-Y. Kuroda, Japanese Syntax and Semantics. Dordrecht: Kluwer, 1992.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "The Logical Analysis of Plurals and Mass Terms: A Lattice-theoretical Approach", "authors": [ { "first": "G", "middle": [], "last": "Link", "suffix": "" } ], "year": 1983, "venue": "Meaning, Use, and Interpretation of Language", "volume": "", "issue": "", "pages": "302--323", "other_ids": {}, "num": null, "urls": [], "raw_text": "Link, G., \"The Logical Analysis of Plurals and Mass Terms: A Lattice-theoretical Approach,\" in Meaning, Use, and Interpretation of Language, R. B\u00e4uerle et al. (eds.), de Gruyter, Berlin, 1983, pp. 302-323.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Scrambling and Scope in Japanese", "authors": [ { "first": "M", "middle": [], "last": "Nakamura", "suffix": "" } ], "year": 1993, "venue": "Japanese/Korean Linguistics", "volume": "2", "issue": "", "pages": "283--298", "other_ids": {}, "num": null, "urls": [], "raw_text": "Nakamura, M., \"Scrambling and Scope in Japanese.\" In Patricia M. Clancy (ed.), Japanese/Korean Linguistics, vol. 2, 1993, pp. 283-298. Stanford Linguistics Association, Stanford CA: CSLI.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "The Internally-Headed Relative Clause Construction in Japanese: A Cognitive Grammar Approach", "authors": [ { "first": "M", "middle": [], "last": "Nomura", "suffix": "" } ], "year": 2000, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Nomura, M., The Internally-Headed Relative Clause Construction in Japanese: A Cognitive Grammar Approach. Ph.D. dissertation, University of California, San Diego, 2000.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Internally Headed Relative Clauses in Japanese and E-type Anaphora", "authors": [ { "first": "J", "middle": [], "last": "Shimoyama", "suffix": "" } ], "year": 1999, "venue": "Journal of East Asian Linguistics", "volume": "8", "issue": "", "pages": "147--182", "other_ids": {}, "num": null, "urls": [], "raw_text": "Shimoyama, J., \"Internally Headed Relative Clauses in Japanese and E-type Anaphora,\" Journal of East Asian Linguistics 8, 1999, pp. 147-182.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "The Syntactic Process", "authors": [ { "first": "M", "middle": [], "last": "Steedman", "suffix": "" } ], "year": 2000, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Steedman, M., The Syntactic Process. Cambridge, MA: MIT Press, 2000.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "Surface-Compositional Scope-Alternation without Existential Quantifiers. Draft", "authors": [ { "first": "M", "middle": [], "last": "Steedman", "suffix": "" } ], "year": 2005, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Steedman, M., Surface-Compositional Scope-Alternation without Existential Quantifiers. Draft, 2005. http://www.iccs.informatics.ed.ac.uk/\u02dcsteedman/papers.html", "links": null } }, "ref_entries": { "TABREF1": { "num": null, "type_str": "table", "html": null, "text": "Every prof, in all his classes, woke up a student who is sleeping before his eyes.' Dono bando-no dono gitarisuto-mo subeteno suteezi-de dono kyoku-demo", "content": "
mo subeteno zyugyoo-de [menomae-de all class-LOC before:eyes-LOC student-NOM sleeping gakusei-ga neteiru ] tatakiokosita NML-ACC woke:up:roughly every prof no-o all stages-LOC every song-LOC [gitaa-no tyuuningu-ga kurutteiru] no-o sonobade tatakikowasita guitar-GEN tuning-NOM wrong NML-ACC on:the:spot broke 'Every guitarist of every band smashed the guitar which was out of tune in every song (10) (11) 'Every band-GEN every guitarist on all stages.' |