{ "paper_id": "2021", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T07:24:31.052256Z" }, "title": "Computing All Quantifier Scopes with CCG", "authors": [ { "first": "Milo\u0161", "middle": [], "last": "Stanojevi\u0107", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Edinburgh", "location": {} }, "email": "m.stanojevic@ed.ac.uk" }, { "first": "Mark", "middle": [], "last": "Steedman", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Edinburgh", "location": {} }, "email": "steedman@inf.ed.ac.uk" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "We present a method for computing all quantifer scopes that can be extracted from a single CCG derivation. To do that we build on the proposal of Steedman (1999, 2011) where all existential quantifiers are treated as Skolem functions. We extend the approach by introducing a better packed representation of all possible specifications that also includes node addresses where the specifications happen. These addresses are necessary for recovering all, and only, possible readings.", "pdf_parse": { "paper_id": "2021", "_pdf_hash": "", "abstract": [ { "text": "We present a method for computing all quantifer scopes that can be extracted from a single CCG derivation. To do that we build on the proposal of Steedman (1999, 2011) where all existential quantifiers are treated as Skolem functions. We extend the approach by introducing a better packed representation of all possible specifications that also includes node addresses where the specifications happen. These addresses are necessary for recovering all, and only, possible readings.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "Quantifiers often introduce a peculiar type of semantic ambiguity.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Take for instance the following sentence: Every farmer owns a donkey. This sentence has two readings: a wide reading where there is one donkey that all farmers share and narrow reading where each farmer has a different donkey. If we express these readings as first-order logic they would look as follows: From these formulas it is clear where the name for different readings come from. In the wide reading the existential quantifier takes the wide scope i.e. it contains the universal quantifier. In the narrow reading the existential quantifier's scope does not cover the universal quantifier.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Any theory of the syntax-semantics interface needs to account for the fact that quantifiers can introduce scope ambiguity. Early approaches to this problem involved either representing the two meanings with distinct logical forms like the above, obtained from the surface string either by treating every farmer and a donkey as generalized quantifiers or \"quantifying in\" in either order to a proposition containing distinguished variables (Montague, 1973) , or via equivalent structure-changing operations of \"quantifier raising\" (May, 1985) . Later approaches decoupled scope from syntactic derivation by the use of \"storage\" to pass scope information (Cooper, 1983; Keller, 1988) . However, all of these approaches overgenerate unattested readings for certain examples involving coordination, first noted by (Geach, 1970) and considered in section 3 below. The approach of (Steedman, 2011) can be thought of as reuniting a storage-like account with surface-compositional syntactic derivation.", "cite_spans": [ { "start": 439, "end": 455, "text": "(Montague, 1973)", "ref_id": "BIBREF7" }, { "start": 530, "end": 541, "text": "(May, 1985)", "ref_id": "BIBREF6" }, { "start": 653, "end": 667, "text": "(Cooper, 1983;", "ref_id": "BIBREF0" }, { "start": 668, "end": 681, "text": "Keller, 1988)", "ref_id": "BIBREF5" }, { "start": 810, "end": 823, "text": "(Geach, 1970)", "ref_id": "BIBREF3" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "2 Computing Scope with CCG Steedman (1999 Steedman ( , 2011 introduces a different view of existential quantifiers, according to which the only true quantifiers are universal quantifiers and that existential quantifiers can be treated as generalized Skolem terms in the following way: Wide:", "cite_spans": [ { "start": 27, "end": 41, "text": "Steedman (1999", "ref_id": "BIBREF10" }, { "start": 42, "end": 59, "text": "Steedman ( , 2011", "ref_id": "BIBREF11" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "\u2200b farmer (b) \u21d2 own (b, sk {} donkey ) Narrow: \u2200b farmer (b) \u21d2 own (b, sk {b} donkey )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Here, sk \u03b1 \u03b2 represents the Skolem function whose arguments are variables of type \u03b1 and whose result is of type \u03b2 .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "In the wide scope reading sk {} donkey is a Skolem constant (Skolem function with no arguments). This means that it will produce only a unique value of type donkey , somewhat like a proper name. In the narrow scope reading sk {b} donkey is a Skolem function that has the variable b bound by the universal as its argument. This function will produce a different value for each b, in other words there will be a different donkey for each farmer .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Other non-universal generalized quantifiers are also treated as Skolem terms. Steedman (2011) also discusses negation which we do not present here, but our approach naturally extends to it. We do not deal with intentionality.", "cite_spans": [ { "start": 78, "end": 93, "text": "Steedman (2011)", "ref_id": "BIBREF11" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "This view of quantifiers allows for a simple Every farmer owns a donkey syntax-semantics interface. CCG derivations for wide and narrow readings are presented in Figure 1, for one of the two derivation trees allowed by CCG. Syntactic component of these two trees is the same, only the semantics differ. Semantic entry for all words are the usual lambda expressions except for the indefinite articles whose entry is \u03bb a.\u03bb b.b(skolem a). Here skolem a is a underspecified Skolem term of type a. An underspecified Skolem term becomes a Skolem function/constant when it is specified. Skolem specification is marked in the derivation tree with a dotted underline, and influences only the logical form, converting an underspecified Skolem terms by giving it as arguments all universally bound variables into whose scope it has been brought by the derivation so far. In Figure 1a that set is empty, so the result of Skolem specification is a Skolem constant, yielding the wide scope reading. In Figure 1b , that set includes the single variable b. By choosing to specify at a different point in the derivation, we get a different narrow-scope reading for the sentence.", "cite_spans": [], "ref_spans": [ { "start": 162, "end": 168, "text": "Figure", "ref_id": null }, { "start": 863, "end": 872, "text": "Figure 1a", "ref_id": null }, { "start": 988, "end": 997, "text": "Figure 1b", "ref_id": null } ], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "S/(S\\NP) (S\\NP)/NP S\\(S/NP) \u03bb a. \u2200b [ f armer (b) \u21d2 (a b)] \u03bb a. \u03bb b. own (b, a) \u03bb a. a(skolem donkey ) >B . . . . . . . . . . . . . . . . . . . . . . . S/NP : \u03bb a. \u2200b [ f armer (b) \u21d2 own (b, a)] S\\(S/NP) : \u03bb a. a sk {} donkey < S : \u2200b f armer (b) \u21d2 own (b, sk {} donkey ) (a) Wide reading. Every farmer owns a donkey S/(S\\NP) (S\\NP)/NP S\\(S/NP) \u03bb a. \u2200b [ f armer (b) \u21d2 (a b)] \u03bb a. \u03bb b. own (b, a) \u03bb a. a(skolem donkey ) >B S/NP : \u03bb a. \u2200b [ f armer (b) \u21d2 own (b, a)] < S : \u2200b [ f armer (b) \u21d2 own (b, skolem donkey )] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S : \u2200b f armer (b) \u21d2 own (b, sk", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "In order to prevent overgeneration of unattested readings, we must impose a further rather natural constraint on Skolem specification requiring that any embedded unspecified Skolem terms are specified at the same time in the same environment. Thus we get the following readings for \"every farmer owns a donkey that ate a hat\": To ensure that all available readings are obtained, it is inefficient to choose all possible specification points in the derivation, because most of them yield duplicate results where there has been no change in the set of scoping variables. To eliminate such redundancy, Steedman (2011) proposed a packed representation presented in Figure 2 where the Skolem term is associated with multiple bindings. At points in the derivation where the binding environment of the function changes, a new argument combination is introduced.", "cite_spans": [], "ref_spans": [ { "start": 661, "end": 669, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "\u2200b[ f armer (b) \u21d2 own (b, sk", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "The proposal of Steedman (2011) was implemented by Kartsaklis (2010) and it works quite well for examples that we have seen so far. However, coordination poses some challenges for the packed representation. Consider coordination of two universal quantifiers in Figure 3a . Here NP \u2191 is a shorthand for a type-raised NP. For a moment ignore additional annotations in the arguments of the Skolem functions. In this example, the specification of an apple will either happen before it is combined with the universals, or after. This means that either it will be in the scope of both or none. The only two readings are given in Figure 3b . However, if we were to unpack the packed formula by computing all combinations of Skolem arguments we would get four readings, including the impossible reading of an apple being within scope of one universal quantifier but not the other. We stress that this is a problem arising from the packed representation, not the theory of scope itself. It may look like the solution to this problem is simple: take all Skolems stemming from the same noun phrase and combine their arguments in order i.e. first arguments of both Skolems go together and second arguments of both Skolems go together. While this solves the example in Figure 3 , it does not work on that in Figure 4 where we coordinate one universal and one existential quantifier. Here there is no clear correspondence between arguments of two Skolem functions: one of them has two different arguments ({} and {a}) while the other one has only one possible argument ({}). Of course, in principle the difference in the number of possible combinations could be significantly larger and it may not be clear how to combine them. To solve this problem we need a more principled solution that directly reflects the mechanism of the non-packed derivations.", "cite_spans": [ { "start": 51, "end": 68, "text": "Kartsaklis (2010)", "ref_id": "BIBREF4" } ], "ref_spans": [ { "start": 261, "end": 270, "text": "Figure 3a", "ref_id": "FIGREF5" }, { "start": 623, "end": 632, "text": "Figure 3b", "ref_id": "FIGREF5" }, { "start": 1256, "end": 1264, "text": "Figure 3", "ref_id": "FIGREF5" }, { "start": 1295, "end": 1303, "text": "Figure 4", "ref_id": "FIGREF6" } ], "eq_spans": [], "section": "Problems with Taking Scope over Coordination", "sec_num": "3" }, { "text": "The packed representation can be seen as a dynamic programming approach to computing all possible orders of specifications of Skolem terms. However, the packed representation of Steedman (2011) that we considered so far is incomplete: from a given packed representation we cannot reconstruct the non-packed representations that are encoded in it. That is caused by the missing information of the location in the tree where the specification was done. We extend the packed representation with this information: whenever a new argument combination is added, together with it we add the Gorn address of the current node. For instance sk {} trrl {a} t apple from Figure 4a signifies that there are two possible arguments for this Skolem function: an empty argument list specified at Gorn address trrrl (top \u2192 right \u2192 right \u2192 right \u2192 left) and a non-empty argument {a} at address t (top). We know that all the Gorn addresses for a given Skolem function will be on a single path from the root of the tree to the determiner that introduced it into derivation. This means that, for a given function, we can sort all addresses by their height in the tree.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "Assume we have a Skolem function with k possible argument sets e 1 , e 2 , . . . , e k sorted by the height of their Gorn addresses g 1 , g 2 , . . . , g k such that g k is closest to the root of the tree. We can say that every argument set e i corresponds to the specializations done on any node g for which it holds g i \u2264 g < g i+1 . 2 In other words g can be any node between g i and g i+1 , including g i but excluding g i+1 . If we take again sk {} trrl {a} t apple as an example we can say that the argument {} corresponds to specialization of Skolem function for nodes trrl, trr and tr.", "cite_spans": [ { "start": 336, "end": 337, "text": "2", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "Additional important point is that we know for certain that g 1 is the address of the leaf of the tree because that is the first point in the derivation where the specification can be done. This is important because in the cases of coordination the logical formula can have copies of the Skolem term that comes from the same noun phrase. We can use the Gorn address of the leaf to identify the Skolem terms that originate in the same noun phrase. Steedman (2011) uses a special index to keep track of this information, but that index is not necessary in our representation due to the existance of Gorn addresses.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "Now we can define unpacking of the new version of the packed representation. We will illustrate it with the example packed formula from the top node of Figure 4a : \u2200a man (a) \u21d2 eat a, sk step 2 For each group of the Skolems extract the unique Gorn addresses where specification changes. In this example that would be {tlrrl} for the first noun phrase and {trrl,t} for the second.", "cite_spans": [], "ref_spans": [ { "start": 152, "end": 161, "text": "Figure 4a", "ref_id": "FIGREF6" } ], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "step 3 Compute the Cartesian product of the sets of Gorn addresses. That will give all possible combinations of specification points. Each combination will correspond to one possible reading of the sentence. In the example that will give {(tlrrl,trrl), (tlrrl,t)}.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "step 4 To transform each entry to a reading we filter the Skolem arguments by the Gorn address. Let us consider how we extract the reading for entry (tlrrl,t woman } with tlrrl is easy because there is only one entry that matches it exactly. Filtering arguments for the second noun phrase is more interesting because there are two copies of it. We need to Every man and every woman eat an apple select for specification on node t. In the first copy sk {} trrl {a} t apple we just select argument {a} since it corresponds to node t. In the second copy sk {} trrl apple we select for {} because it covers all nodes from trrl to the root including t.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "NP \u2191 /N N conj NP \u2191 /N N (S\\NP)/NP NP \u2191 /N N \u03bb", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "NP \u2191 NP \u2191 NP \u2191 \u03bb a. \u2200b [man (b) \u21d2 (a b)] \u03bb a. a sk {} tlrrl woman \u03bb a. a sk {} trrl apple >\u03a6 > NP \u2191 \\NP \u2191 S\\NP \u03bb a. \u03bb b. (a b) \u2227 b sk {} tlrrl woman \u03bb a. eat a, sk {} trrl apple < NP \u2191 \u03bb a. \u2200b [man (b) \u21d2 (a b)] \u2227 a sk {} tlrrl woman < S \u2200a man (a) \u21d2 eat a,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed Solution", "sec_num": "4" }, { "text": "This approach is really just a full dynamic programming representation of the unpacked representations that could easily be extracted from this representation. We do not have to explicitly encode all the nodes where specification happens, but only for the places where that specification changes the existing result and we also encode exactly at which places in the tree this happens.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion", "sec_num": "5" }, { "text": "Here we have described how to get all possible readings from a single CCG derivation. However, in some cases there can be alternative CCG derivations that can provide additional readings. To get those readings we can apply the same method on all alternative derivations either by chart parsing, as described in (Steedman, 2011) , or by recovering alternative derivations with the tree-rotation operation (Niv, 1994; Stanojevi\u0107 and Steedman, 2019) Evang and Bos (2013) show that there is a strong preference for subject to take scope over object.", "cite_spans": [ { "start": 311, "end": 327, "text": "(Steedman, 2011)", "ref_id": "BIBREF11" }, { "start": 404, "end": 415, "text": "(Niv, 1994;", "ref_id": "BIBREF8" }, { "start": 416, "end": 446, "text": "Stanojevi\u0107 and Steedman, 2019)", "ref_id": "BIBREF9" }, { "start": 461, "end": 467, "text": "(2013)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Conclusion", "sec_num": "5" }, { "text": "Our representation of Skolem terms could be extended to encode the information of the type of noun phrase they originate from. With this extension we could rank the extracted readings by subject > object preference.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion", "sec_num": "5" }, { "text": "The implementation of our approach is available at https://github.com/stanojevic/ CCG-Quantifiers.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion", "sec_num": "5" }, { "text": "This condition was inadvertently omitted from the original proposal.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "For simplicity, when g k = t we can consider g k+1 = t in order to have a complete coverage to the root of the tree.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [ { "text": "This work was supported by ERC H2020 Advanced Fellowship GA 742137 SEMANTAX grant.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Acknowledgments", "sec_num": null } ], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Quantification and Syntactic Theory", "authors": [ { "first": "Robin", "middle": [], "last": "Cooper", "suffix": "" } ], "year": 1983, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Robin Cooper. 1983. Quantification and Syntactic The- ory. Reidel, Dordrecht.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Semantics of Natural Language. Reidel", "authors": [], "year": 1972, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Donald Davidson and Gilbert Harman, editors. 1972. Semantics of Natural Language. Reidel, Dordrecht.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Scope Disambiguation as a Tagging Task", "authors": [ { "first": "Kilian", "middle": [], "last": "Evang", "suffix": "" }, { "first": "Johan", "middle": [], "last": "Bos", "suffix": "" } ], "year": 2013, "venue": "Proceedings of the 10th International Conference on Computational Semantics (IWCS 2013) -Short Papers", "volume": "", "issue": "", "pages": "314--320", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kilian Evang and Johan Bos. 2013. Scope Disambigua- tion as a Tagging Task. In Proceedings of the 10th International Conference on Computational Seman- tics (IWCS 2013) -Short Papers, pages 314-320, Potsdam, Germany. Association for Computational Linguistics.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "A program for syntax. Synth\u00e8se", "authors": [ { "first": "Peter", "middle": [], "last": "Geach", "suffix": "" } ], "year": 1970, "venue": "", "volume": "22", "issue": "", "pages": "483--497", "other_ids": {}, "num": null, "urls": [], "raw_text": "Peter Geach. 1970. A program for syntax. Synth\u00e8se, 22:3-17. Reprinted as Davidson and Harman 1972:483-497.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Wide-coverage CCG parsing with quantifier scope", "authors": [ { "first": "Dimitrios", "middle": [], "last": "Kartsaklis", "suffix": "" } ], "year": 2010, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Dimitrios Kartsaklis. 2010. Wide-coverage CCG pars- ing with quantifier scope. Master's thesis, Univer- sity of Edinburgh.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Nested Cooper storage", "authors": [ { "first": "William", "middle": [], "last": "Keller", "suffix": "" } ], "year": 1988, "venue": "Natural Language Parsing and Linguistic Theory", "volume": "", "issue": "", "pages": "432--447", "other_ids": {}, "num": null, "urls": [], "raw_text": "William Keller. 1988. Nested Cooper storage. In Uwe Reyle and Christian Rohrer, editors, Natural Lan- guage Parsing and Linguistic Theory, pages 432- 447. Reidel, Dordrecht.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Logical Form", "authors": [ { "first": "Robert", "middle": [], "last": "May", "suffix": "" } ], "year": 1985, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Robert May. 1985. Logical Form. MIT Press, Cam- bridge, MA.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "The Proper Treatment of Quantification in Ordinary English", "authors": [ { "first": "Richard", "middle": [], "last": "Montague", "suffix": "" } ], "year": 1973, "venue": "Approaches to Natural Language: Proceedings of the 1970 Stanford Workshop on Grammar and Semantics", "volume": "", "issue": "", "pages": "221--242", "other_ids": { "DOI": [ "10.1007/978-94-010-2506-5_10" ] }, "num": null, "urls": [], "raw_text": "Richard Montague. 1973. The Proper Treatment of Quantification in Ordinary English. In K. J. J. Hin- tikka, J. M. E. Moravcsik, and P. Suppes, editors, Approaches to Natural Language: Proceedings of the 1970 Stanford Workshop on Grammar and Se- mantics, pages 221-242. Springer Netherlands, Dor- drecht.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "A Psycholinguistically Motivated Parser for CCG", "authors": [ { "first": "Michael", "middle": [], "last": "Niv", "suffix": "" } ], "year": 1994, "venue": "32nd Annual Meeting of the Association for Computational Linguistics", "volume": "", "issue": "", "pages": "125--132", "other_ids": { "DOI": [ "10.3115/981732.981750" ] }, "num": null, "urls": [], "raw_text": "Michael Niv. 1994. A Psycholinguistically Motivated Parser for CCG. In 32nd Annual Meeting of the As- sociation for Computational Linguistics, pages 125- 132, Las Cruces, New Mexico, USA. Association for Computational Linguistics.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "CCG Parsing Algorithm with Incremental Tree Rotation", "authors": [ { "first": "Milo\u0161", "middle": [], "last": "Stanojevi\u0107", "suffix": "" }, { "first": "Mark", "middle": [], "last": "Steedman", "suffix": "" } ], "year": 2019, "venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies", "volume": "1", "issue": "", "pages": "228--239", "other_ids": {}, "num": null, "urls": [], "raw_text": "Milo\u0161 Stanojevi\u0107 and Mark Steedman. 2019. CCG Parsing Algorithm with Incremental Tree Rotation. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computa- tional Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 228-239, Minneapolis, Minnesota. Association for Computa- tional Linguistics.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Alternating Quantifier Scope in CCG", "authors": [ { "first": "Mark", "middle": [], "last": "Steedman", "suffix": "" } ], "year": 1999, "venue": "Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics on Computational Linguistics, ACL '99", "volume": "", "issue": "", "pages": "301--308", "other_ids": { "DOI": [ "10.3115/1034678.1034728" ] }, "num": null, "urls": [], "raw_text": "Mark Steedman. 1999. Alternating Quantifier Scope in CCG. In Proceedings of the 37th Annual Meet- ing of the Association for Computational Linguistics on Computational Linguistics, ACL '99, pages 301- 308, USA. Association for Computational Linguis- tics.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "Taking Scope: The Natural Semantics of Quantifiers", "authors": [ { "first": "Mark", "middle": [], "last": "Steedman", "suffix": "" } ], "year": 2011, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": { "DOI": [ "10.7551/mitpress/9780262017077.001.0001" ] }, "num": null, "urls": [], "raw_text": "Mark Steedman. 2011. Taking Scope: The Natural Se- mantics of Quantifiers. The MIT Press.", "links": null } }, "ref_entries": { "FIGREF0": { "num": null, "uris": null, "type_str": "figure", "text": "Wide: \u2203a [donkey (a) \u2227 \u2200b [farmer (b) \u21d2 own (b, a)]] Narrow: \u2200b [\u2203a [donkey (a) \u2227 (farmer (b) \u21d2 own (b, a))]]" }, "FIGREF2": { "num": null, "uris": null, "type_str": "figure", "text": ") \u03bb a. \u2200b [ f armer (b) \u21d2 (a b)] \u03bb a. \u03bb b. own (b, a) \u03bb a. a(sk {} donkey ) >B S/NP :\u03bb a. \u2200b [ f armer (b) \u21d2 own (b, a)] < S :\u2200b f armer (b) \u21d2 own (b, sk {} {b} donkey ) Packed representation." }, "FIGREF3": { "num": null, "uris": null, "type_str": "figure", "text": "{}\u03bb a.donkey (a)\u2227ate (a,sk{} hat ) )] \u2200b[ f armer (b) \u21d2 own (b, sk {b} \u03bb a.donkey (a)\u2227ate (a,sk {b} hat ) )] \u2200b[ f armer (b) \u21d2 own (b, sk {b} \u03bb a.donkey (a)\u2227ate (a,sk {} hat ) )] However, we exclude a fourth reading with a wide-scope Skolem constant donkey eating multiple farmer-dependent hats: 1 #\u2200b[ f armer(b) \u21d2 own (b, sk {} \u03bb a.donkey (a)\u2227ate (a,sk {b} hat ) )]" }, "FIGREF4": { "num": null, "uris": null, "type_str": "figure", "text": "a. \u03bb b. \u2200c [(a c) \u21d2 (b c)] man and \u03bb a. \u03bb b. \u2200c [(a c) \u21d2 (b c)] woman \u03bb a. \u03bb b. eat (b, a) \u03bb a. \u03bb b. b sk \u2200b [man (b) \u21d2 (a b)] \u03bb a. \u2200b [woman (b) \u21d2 (a b)]\u03bb a. a sk{} trrl apple >\u03a6 > NP \u2191 \\NP \u2191 S\\NP \u03bb a. \u03bb b. (a b) \u2227 \u2200c [woman (c) \u21d2 (b c)] \u03bb a. eat a, sk {} trrl apple < NP \u2191 \u03bb a. \u2200b [man (b) \u21d2 (a b)] \u2227 \u2200c [woman (c) \u21d2 (a c)] < S \u2200a man (a) \u21d2 eat a, sk{} trrl {a} t apple \u2227 \u2200b woman (b) \u21d2 eat b, sk {} trrl {b} t apple (a) Packed derivation. \u2200a man (a) \u21d2 eat a, sk {} trrl apple \u2227 \u2200b woman (b) \u21d2 eat b, sk {} trrl apple \u2200a man (a) \u21d2 eat a, sk {a} t apple \u2227 \u2200b woman (b) \u21d2 eat b, sk {b} t apple (b) Readings." }, "FIGREF5": { "num": null, "uris": null, "type_str": "figure", "text": "Coordination with two universal quantifiers. NP \u2191 /N N \u03bb a. \u03bb b. \u2200c [(a c) \u21d2 (b c)] man and \u03bb a. \u03bb b. b sk {} tlrrl a woman \u03bb a. \u03bb b. eat (b, a) \u03bb a. \u03bb b. b sk" }, "FIGREF6": { "num": null, "uris": null, "type_str": "figure", "text": "Coordination with one universal and one existential quantifier." }, "TABREF0": { "content": "
\u2227
eat sk {} tlrrl woman , sk {} trrl apple
step 1 Group Skolem terms by the NP they belong
to. For that we can use the first Gorn address
that specifies the leaf node. In the example
that would give {sk {} tlrrl woman } for the first NP and {sk {} trrl {a} t apple , sk {} trrl
", "num": null, "text": "{} trrl {a} t apple", "html": null, "type_str": "table" }, "TABREF2": { "content": "
{} trrl {a} t apple\u2227 eat sk{} tlrrl woman , sk {} trrl apple
(a) Packed derivation.
\u2200a man (a) \u21d2 eat a, sk{} trrl apple\u2227 eat sk{} tlrrl woman , sk {} trrl apple
\u2200a man (a) \u21d2 eat a, sk{a} t apple\u2227 eat sk{} tlrrl woman , sk {} trrl apple
(b) Readings.
", "num": null, "text": "sk", "html": null, "type_str": "table" } } } }