| { | |
| "paper_id": "1991", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:34:09.579539Z" | |
| }, | |
| "title": "UNIFICATION-BASED DEPENDENCY PAR SING OF GOVERNOR-FINAL LANGUAGES", | |
| "authors": [ | |
| { | |
| "first": "Hyuk-Chul", | |
| "middle": [], | |
| "last": "Kwon", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Pusan National U", | |
| "location": { | |
| "addrLine": ":iiversity 30 changjun-dong, Keumjung-ku", | |
| "postCode": "609-735", | |
| "settlement": "Pusan", | |
| "country": "Republic of Korea" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "This paper describes a unification based dependency parsing method for governor-final languages. Our method can parse not only projective sentences but also non-projective sentences. The feature structures in the tradition of the unification-based formalism are used for writing dependency relations. We use a structure sharing and a local ambiguity packing to save storage.", | |
| "pdf_parse": { | |
| "paper_id": "1991", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "This paper describes a unification based dependency parsing method for governor-final languages. Our method can parse not only projective sentences but also non-projective sentences. The feature structures in the tradition of the unification-based formalism are used for writing dependency relations. We use a structure sharing and a local ambiguity packing to save storage.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "The parsers . of phrase structure grammars face troubles for parsing free word order languages in following respects.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "First, t\ufffdey require a large size of grammatical rules for parsing free word order languages. Second, the free word order often results in discontinuous con stituents (Covington, 1988) . A phrase structure tree of a sentence with discontinuous constituents would have crossing branches. This crossing branches can not be represented by con ventional context free rules. Third, free word order languages feature very rich systems of morphological mark ings (Kwon, 1990) . Word arrangements and morphological markings are ob viously contingent on relations between wordforms rather than on constituen cy(Mel' cuk, 1988).", | |
| "cite_spans": [ | |
| { | |
| "start": 166, | |
| "end": 183, | |
| "text": "(Covington, 1988)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 455, | |
| "end": 467, | |
| "text": "(Kwon, 1990)", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "One approach to parse free word order languages is the principle-based parsing (Berwick, 1987) . The other ap proach is the dependency pars ing(Mel' cuk, 1988). This paper describes a unification based dependency parsing method for governor-final(head-final) languages like Korean and Japanese. We develop the parsing method with special refer ence to Korean but the method can be adapted directly to Japanese parsing. Korean and Japanese are relatively free word order languages (Kwon, 1990 ). Al though their word order is free except that dependents always precede their governor, word order variations lead to different emphasis on the topic and the focus. In contrast, their morpheme order is fixed at the level of words.", | |
| "cite_spans": [ | |
| { | |
| "start": 79, | |
| "end": 94, | |
| "text": "(Berwick, 1987)", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 480, | |
| "end": 491, | |
| "text": "(Kwon, 1990", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In Korean and Japanese, it is quite natural to drop any arguments including a subject and an object if they can be recovered through the context. Null sub jects are also found in Italian and Spanish (Moon, 1989) . Null arguments make it much harder to parse Korean and Japanese using phrase structure gram mars. Because dependency grammars analyze syntactic structure as the relationships between ultimate syntactic units(i.e, morpheme, part of speech), de pendency parsers can easily parse sen tences with null arguments. This paper follows the grammatical formalism of Mel'cuk(1988), but modifies it for computational efficiency and Korean specific characteristics. We try to parse not only projective sentences but also non-projective sentences. Fea ture structures in the tradition of unifica tion-based grammars are used for writing dependency relations. But unification operation is modified f o r parsing non projective sentences. A structure sharing and a local ambiguity packing is used to save storage.", | |
| "cite_spans": [ | |
| { | |
| "start": 199, | |
| "end": 211, | |
| "text": "(Moon, 1989)", | |
| "ref_id": "BIBREF7" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Introduction", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Mel'cuk differentiates three depend ency relations : morphological depend ency, syntactic dependency and semantic dependency(Mel'cuk, 1988).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The syntactic dependency is binary relations between wordforms, which are anti-symmetric, anti-reflexive and anti transitive. The syntactic relations are represented by arcs : X -> Y: where X governs Y; X is called the governor of Y; and Y is called the dependent of X. The syntactic relations are best represented by a connected directed labeled graph.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Mel' cuk gives additional restrictions on the syntactic structure. First, a syntac tic structure contains exactly one node(root) that does not depend on another node. Second, in a syntactic structure, no node may simultaneously depend on two or more other nodes. The syntactic structure becomes a rooted tree, specifically a D-tree by these two restrictions.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In Korean and Japanese, there are two different morphemes: free( content) morphemes and bound(function) mor phemes. Bound morphemes include postpositions and verbal endings. A free morpheme can depend on another mor pheme directly. But a bound morpheme can depend on another morpheme after it governs other morphemes. This means that the leaf nodes of the D-tree are always free morphemes.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We use feature structures in the tradi tion of unification-based grammars for writing dependency relations (Sells, 1985 < Dictionary 1 > is a sample Korean dictionary. The feature \"bound\" is used to differentiate between bound mor phemes and free morphemes. When a bound morpheme governs another mor pheme, the value of \"bound\" become \"nil\". As ''bound\" is not controlled by the unification operation, the change of the value of \"bound\" does not destroy the monotonicity of the unification. More explanation will be found in chapter III. In (1) and (2), the subject marker(\"i\") governs \"John\" and the object marker(\"ul\") governs \"Susan\". \"Po\" governs both the nominative construc tion (\"John-i\") and the accusative con struction (\"Susan-ul\"). Because of no dependency between \"J ohn-i\" and \"Susan-ul\", there is no precedence rela tion between them. \"da\" governs \"John-i Susan-ul po\". As a result, both (1) and 2are grammatical sentences and they have the same meaning as \"John sees Susan\".", | |
| "cite_spans": [ | |
| { | |
| "start": 107, | |
| "end": 119, | |
| "text": "(Sells, 1985", | |
| "ref_id": "BIBREF9" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(3) John-i Bill-kwa Susan-ul po-da Although \"kwa\" is a postposition, it can depend on a verb stem or a noun, but not both: When it depends on a verb stem, its meaning is \"with\". But its mean ing is \"and\" if it depends on a noun. < Dictionary 2 > shows the lexical information of \"kwa\". From < Table 1 > and < Dictionary 2 >, we conclude that (3) has two dif ferent interpretations. The decision of null arguments also requires the subcategorization. As the \ufffd verb stem \"malha\" subcategorizes a sub ject and a complementizer(\"ko\"), and \"po\" subcategorizes a subject and an ob ject, two subjects are required in (7). But there is only one nominative construc tion. The nominative construction can be governed by \"po\" or \"malha\", but not both. As a result, we can conclude that one subject is dropped. (7) has two dif ferent interpretations as below. \"kwon\" is a noun but a bound mor pheme. We call it an incomplete noun. \"kwon\" can govern a numeral and a noun but there are restrictions in the govern ing order. \"kwon\" can govern a noun only after it governs a numeral, but the op posite is not true. This additional precedence restrictions can be formu lated as < The second row of < Table 2 > shows that a noun which is modified by a numeral (determiner) can govern a noun. The dictionary also shows that \"kwon\" is an incomplete noun and is a unit for counting books._ There is a mor phological dependency between \"kwon\" and \"sajen\". The above shows how our system deals with the morphological de pendencies and additional precedence restrictions using feature structures.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 293, | |
| "end": 300, | |
| "text": "Table 1", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 1183, | |
| "end": 1192, | |
| "text": "Table 2 >", | |
| "ref_id": "TABREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "II. Dependency Relations and Fea ture Structures", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Using dependencies for parsing natural languages, the projectivity is an extremely important property of the word order. A sentence is called projec tive if and only if the arcs of dependency links satisfy fallowing restric tions(Mel'cuk, 1988). The algorithm scans a sentence from left-to -:-right for searching a governor. If a governor is found, it tries to make all the dependency links between the gover nor and the constructions whose head is the morpheme which immediately precedes the governor. The term head is used in the sense of top node of a construction as Mel'cuk(1988).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Structure Sharing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In a projective sentence, a governor can govern a wordform if and only if the governor governs directly or indirectly all the wordforms between them. Let < m1,m2, ... ,mn > be an ordered list of morphemes. If \u2022 mi governs mj and mj governs mk, then mi indirectly governs mk. The morpheme mi can govern Illj if and only if all the morphemes between mj and mi are governed directly or in directly by mi where j < i. A head governs directly or indirectly all the other mor phemes in a construction.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Structure Sharing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Our parsing strategy is as follows.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Structure Sharing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "First) The parser gets a morpheme mi from the lexical analyzer until an end-of sentence marker is encounted.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Structure Sharing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Second) The parser searches construc tions whose head is mi-1. When there exist dependency relations between mi and some of them, the parser generates new constructions and stores them in the queue.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Structure Sharing", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "When some constructions exist in the queue, the parser gets one of them from the queue. Otherwise, goto first) . Let that construction contain all the mor phemes from mj to mi where j < i and mi is its head. The parser searches construe-tions whose head is mj-1. When there exist dependency relations between mi and some of them, the parser generates new constructions, stores them in the queue and repeats third) .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Third)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We implement the algorithm by chart. < Fig.1 > shows the architecture (iv)", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 37, | |
| "end": 69, | |
| "text": "< Fig.1 > shows the architecture", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Third)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "<Fig. 2> The state of the pool (i) is the pool after processing \"arum dap-n Mary-eui\". As the inactive edge pool is empty, the parser gets \"chinku\" from the lexical analyzer as (ii). When the processing of (#7) is finished, the pool become (iii). (iv) shows the pool when {#10) is generated. As bound mor phemes (\"n\",\"eui\") can not depend on other morphemes by themselves, it is not necessary to store bound morphemes at the pool.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Third)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We use a structure sharing(Tomita, 1986) and a local ambiguity pack ing (Shieber, 1986) to save storage. Al though the order of the features is not important in the unification formalism, we always place the \"bound\" feature first.", | |
| "cite_spans": [ | |
| { | |
| "start": 72, | |
| "end": 87, | |
| "text": "(Shieber, 1986)", | |
| "ref_id": "BIBREF10" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Tqe storage for parsing grows ex ponentially as ambiguities are increased\ufffd", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "dcp l(ov <Fig. 3>", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Tqe storage for parsing grows ex ponentially as ambiguities are increased\ufffd", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "< Fig. 3 > shows that {#8) shares the structures of {#6) and {#7). {#6) shares the structure of \"eui\" except for the \"bound\" feature. As the \"bound\" feature is excluded, the monotonicity of the unification is not destroyed.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 2, | |
| "end": 26, | |
| "text": "Fig. 3 > shows that {#8)", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Tqe storage for parsing grows ex ponentially as ambiguities are increased\ufffd", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We state that two or more subtrees represent a local ambiguity if they have the same starting point and the same en-ding point and if their top nodes are the same wordform. That is, (#9) and {#10) of the < Table 3 > represent a local am biguity. If a sentence has many local am biguities, the total ambiguities would grow exponentially. To avoid this, we use a technique called local ambiguity pack ing which is suggested by Tomita(1986). < Table 4 > is the content of the pools after (9) is parsed with a structure sharing and a local ambiguity packing. #(9, 10) in < Table 4 > is the result of the local ambiguity packing of ( #9) and {#10) in <Table 3 >.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 206, | |
| "end": 213, | |
| "text": "Table 3", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 441, | |
| "end": 448, | |
| "text": "Table 4", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Tqe storage for parsing grows ex ponentially as ambiguities are increased\ufffd", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Non-projective sentences give serious difficulties in parsing natural languages. The first row of < Table 5 > shows that a verb stem temporarily governs an adverb which has [ negative : c + ]. When the verb stem depends on a construction which has [ negative : + ] and the depend ency does not pass over the sentence . boundary, the temporary depende\u2022 ncy link is removed and a new dependency link between the adverb and the construction is connected. Two construc tions are not unified when their dependency is temporary. We handle the co-occurrence feature similar to the \"bound\" feature. When \"an\" governs \"po-ji\", a new link between \"an\" and \"kulko\" replaces the temporary link between \"po\" and \"kulko\". It is important that [ negative: c +] is removed in (ii). If some co-occurrence fe atures remain after the parsing, the sentence is incorrect.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 100, | |
| "end": 107, | |
| "text": "Table 5", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "IV. Parsing Non-Projective Sen tences", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The other type of non-projectivity oc curs by non-local dependencies. Some constructions which are the dependents of an embedded verb can be placed at outer sentences in Korean. We can also find non-local dependencies in Fin nish (Karttunen, 1986) .", | |
| "cite_spans": [ | |
| { | |
| "start": 230, | |
| "end": 247, | |
| "text": "(Karttunen, 1986)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "IV. Parsing Non-Projective Sen tences", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "NomR _LmR ci ( 1 I) Susan\ufffd John\ufffd Tom-i po -ass -da-ko malha -da (John sa ys that Tom saw Susan)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "j[d ccR", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "As stated above, \"po\" subcategorizes a subject and an object, and \"malha\" sub categorizes a subject and a complemen tizer. ( 11) has the cross arcs because \"J h \u202211 d II II \"malha\" governs o n-1 an po governs \"Susan-ul\". Karttunen shows that this problem can be solved by functors with a floating type in Finish (Karttunen, 1986) . The same framework also works in Korean. The framework can yield more than one results, but most of them are only accept able at extraordinary situations. There fore, our system strengthens the framework as a construction can be com bined only with the nearest verb stem which can govern it when there is no projective governor of it.", | |
| "cite_spans": [ | |
| { | |
| "start": 312, | |
| "end": 329, | |
| "text": "(Karttunen, 1986)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "j[d ccR", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We have shown a unification-based dependency parsing method for gover nor-final languages like Korean and Japanese. Feature structures in the tradi tion of unification-based grammars have been used for writing dependency rela tions. Our method can parse non-projec ti ve sentences as well as projecitve sentences.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "V. Conclu\u2022 sion", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We implement a Korean parser by the method presented in this paper using C language. The first version parser only used a structure sharing. But the current version uses a structure sharing and a local ambiguity packing. The local am biguity packing saves about 35 % of storage for parsing sample sentences.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "V. Conclu\u2022 sion", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "More efficient structure sharing method and the dictionary structure are under study. We plan to use our method for parsing fixed word order languages.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "V. Conclu\u2022 sion", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "[12] Tomita, M. 1986 Efficient Parsing for Natural Language, Kluwer Academic Pub.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "But almost all languages have some sorts of non-projectivity(Mel'cuk, 1988).There are two types of non-projec tivity in Korean. The first one is related to the feature co-occurrence where the dependency links do not pass over the sentence boundary. \"kulko\" is used only in negative sentences. In (10), \"po\" governs \"John-i\" and \"Susan-ul\", but the auxiliary verb stem \"an\" governs \"kulko\" and \"po-ji\". \"kulko\" can be placed anywhere before \"an\" at (10). ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "annex", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Principle-Based Parsing", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Berwick", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "A.I. T.R", | |
| "volume": "", | |
| "issue": "972", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Berwick, R. C. 1987 Principle-Based Parsing, A.I. T.R. No. 972, MIT AI Lab.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Parsing Vari able Word Order Langu age with Unification-Based Dependency Gram mar", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Covington", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Covington, M. A. 1988 Parsing Vari able Word Order Langu age with Unification-Based Dependency Gram mar, ACMC Research Report 01-0022, University of Georgia.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Dependency Unification Grammar", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Hellwig", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "Proc. of Colling", | |
| "volume": "86", | |
| "issue": "", | |
| "pages": "195--198", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hellwig, P. 1986 Dependency Unification Grammar, Proc. of Colling 86, pp.195-198.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Radical Lexi calism, Alternative Conceptions of Phrase Structure", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Karttunen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "44--65", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Karttunen, L. 1989 Radical Lexi calism, Alternative Conceptions of Phrase Structure, The University of Chicago Press., pp44-65.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "A Government Binding Based Parser for Warlpiri, TR 993", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "B" | |
| ], | |
| "last": "Kashket", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kashket, M. B. 1987 A Government Binding Based Parser for Warlpiri, TR 993, MIT AI Lab.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "A Korean Analysis System Based on Unification and Chart", | |
| "authors": [ | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Kwon", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Yoon", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Kim", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Proc. of Pacific Rim International Conference on Artifi cial Intelligence", | |
| "volume": "90", | |
| "issue": "", | |
| "pages": "251--256", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kwon, H., Yoon, A., Kim, Y. 1990 A Korean Analysis System Based on Unification and Chart, Proc. of Pacific Rim International Conference on Artifi cial Intelligence '90, pp251-256.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Dependency Syn tax : Theory and Practice", | |
| "authors": [ | |
| { | |
| "first": "I", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Mel'cuk", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mel'cuk, I.A. 1988 Dependency Syn tax : Theory and Practice, State of University of New York Press.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "The Syntax of Null Arguments with Special Refer--ence to Korean", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Moon", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "Ph.D.Diss", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Moon, G. 1989 Ph.D.Diss.,The Syntax of Null Arguments with Special Refer- -ence to Korean, The University of Texas at Austin.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "A Common Parsing Scheme for Left-and Right-Branching Languages", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [ | |
| "T" | |
| ], | |
| "last": "Sato", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "J. of Computational Lin guistics", | |
| "volume": "14", | |
| "issue": "1", | |
| "pages": "20--30", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sato, P. T. 1988 \"A Common Parsing Scheme for Left-and Right-Branching Languages,\" J. of Computational Lin guistics, Vol. 14, No.1, pp.20-30.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Lectures on Contem porary Syntactic Theories, CSLI", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Sells", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1985, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sells, P. 1985 Lectures on Contem porary Syntactic Theories, CSLI.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "A Compila tion of Papers on Unification-Based Grammar Formalisms : Part II", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Shieber", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Shieber,S.M et al. 1986 A Compila tion of Papers on Unification-Based Grammar Formalisms : Part II, Report No. CSLI-86-48", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF2": { | |
| "uris": null, | |
| "num": null, | |
| "text": ") John -1 Bill -kwa Susan -ul po -da (John sees Susan with Bill) I NomR j_ _ e\ufffd fAccR 7 I (3\") John -i Bill -kwa Susan -ul po -da ( John sees Susan and Bill) < NomR : Nom i native Relation, AccR : Accusative R, ConjR :Conjunctive R, ComR :Comminative R> But ( 4) has only one interpretation. John-i Susan-ul Bill-kwa po-da (John sees Susan with Bill) < Table 1 > and < Dictionary 2 > also show that the possessive \u2022 postposi tion \"eui(of)\" only depends on a noun. The subcategorization of a verb gives additional constraints on the depend ency relations. The subcategorization is used for a case assignment, the decision of null arguments and a filter on govern ing patterns. When a subject and an ob-: ject are topicalized, the subject marker and the object marker are replaced to topic markers. John-i Susan-un po-da TM < 1M :Topic Marker, TopR : Topical R > Postpositions do not provide the suf ficient information for the case assignment of topicalized constructions in ( 5) and (6).In (5), \"po\" governs the topicalized construction and the accusative conver stem po sucategorizes both a subject and an object. So, the noun of the topicalized construc tion is the subject of (5). (5) and (6) have the same meaning as (1) except that the subject and the object are topicalized respectively.In Korean, the noun of a nominative construction is always the subject of a verb, and the noun of an accusative con struction is the object of a verb, but not vice-versa. Therefore, we separate the case marking operation and the case as signment operation. The case of a topi calized construction is assigned when a verb stem is governed by a verbal ending.(7) John-i Susan-ul po-ass -da -ko malha -da VE VE VE VS VE (past) (DEC) (COMP) (say) (DEC) <COMP : COMPiementizer>", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF3": { | |
| "uris": null, | |
| "num": null, | |
| "text": ") John-i Susan-ul po -ass da-ko malha-da (? says that John saw Susan) i) sajen se kwon(three dictionaries) NOUN DET NOUN (dictionary)(three)Book.Form +o:i ii) -;-k won (three book-like materials) iii) *sajen kwon(not allowed) iv) * se sajen kwon(not allowed) < \"kwon H : a unit for counting book-like materials, DET : determiner, ModR : Modificative Relation, ClassR : Classificative Relation>", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF4": { | |
| "uris": null, | |
| "num": null, | |
| "text": "(i) No arc crosses another arc (ii) No arc covers the root of D-tree Although most sentences of natural languages are projective, there exist several types of non-projective senten ces. Non-projective sentences have dis continuous constituents. This chapter gives a parsing algorithm for projective sentences. The algorithm will be modified for non-projective sentences in the next chapter.", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF5": { | |
| "uris": null, | |
| "num": null, | |
| "text": "The parser joins one dependent to one governor at a time. Each edge has a starting point and an ending point. (([arumdap,n],Mary],eui],chinku] friend of beautiful Mary 10 0 5 ([[arumdap,n],Mary,eui]),chinku) beautiful friend or Mary <Table 3> Stem, VE : Verb Ending, PP[poss] : Possesive Postposition> < Table 3 > shows the content of the pool while (9) is parsed. (9) means \"a/the friend of Mary who is \u2022beautiful\" and has two different interpretations as (#9) and ( # 10). < Fig .2 > shows the state of the pool when \"chinku\" is processed. I \" 2 I ... I \" 5 I \" 6 1 #? I #B I#", | |
| "type_str": "figure" | |
| }, | |
| "TABREF0": { | |
| "num": null, | |
| "text": "", | |
| "html": null, | |
| "type_str": "table", | |
| "content": "<table><tr><td/><td/><td>\ufffd:a\ufffd; \ufffd\ufffd\ufffdsi\ufffdo \ufffd case : accusative</td><td>e r :==::\ufffd stem subcat->subj,obj} \ufffd</td><td>e [ :/:e\ufffdl -endin \ufffd modal : declarativ</td></tr><tr><td/><td/><td>bound : +</td><td>ubj : [animate : +</td><td>bound : +</td></tr><tr><td/><td/><td/><td>< Dictionary 1 ></td></tr><tr><td>\ufffd x : \"' i</td><td>\"'</td><td>J</td></tr><tr><td colspan=\"3\">case : nominative</td></tr><tr><td colspan=\"2\">bound :+</td><td/></tr></table>" | |
| }, | |
| "TABREF1": { | |
| "num": null, | |
| "text": "", | |
| "html": null, | |
| "type_str": "table", | |
| "content": "<table><tr><td>governor</td><td/><td>re lation</td><td>dependent</td></tr><tr><td>[cat :noun]</td><td/><td>mod i ficative</td><td>[ cat:detJ</td></tr><tr><td>': noun</td><td>j</td><td/></tr><tr><td colspan=\"2\">[ modifier lex:det [numeral : +</td><td>classificative</td><td>[cat:nounj</td></tr><tr><td/><td colspan=\"2\">< Table 2 ></td></tr></table>" | |
| } | |
| } | |
| } | |
| } |