{
    "paper_id": "W98-0141",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T06:03:09.388748Z"
    },
    "title": "Translating the XTAG English Grammar to HPSG",
    "authors": [
        {
            "first": "Yuka",
            "middle": [],
            "last": "Tateisit",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "UMIST' u. K",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Kentaro",
            "middle": [],
            "last": "Torisawat",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "UMIST' u. K",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Yusuke",
            "middle": [],
            "last": "Miyaot",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "UMIST' u. K",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Jun",
            "middle": [
                "'"
            ],
            "last": "Ichi",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "UMIST' u. K",
                "location": {}
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "",
    "pdf_parse": {
        "paper_id": "W98-0141",
        "_pdf_hash": "",
        "abstract": [],
        "body_text": [
            {
                "text": "We describe the development of XHPSG, a large-scale English grammar in the HPSG formalism translated from the XTAG grammar (The XTAG Research Group, 1995) . Our goal is to obtain a large-scale, linguistically sound grammar for our HPSG parser (Makino et al\" 1998) with a relatively small workload. For this purpose, we try to make an HPSG grammar equivalent to the XTAG grammar in the strong sense where we preserve the structures and the linguistic analysis of the XTAG grammar.",
                "cite_spans": [
                    {
                        "start": 123,
                        "end": 154,
                        "text": "(The XTAG Research Group, 1995)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "To guarantee the equivalence of the XHPSG and XTAG grammars, the following conditions must be satisfied: 1) An XTAG elementary tree is translated to an XHPSG lexical item that translates back to the original elementary tree by applying the schemata and principles; 2) No XHPSG lexical item translates back to a tree other than the original XTAG elementary tree; 3) Substitution and adjunction allowed in the original grarnmar, and no other opeations, are simulated in the XHPSG parsing.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "We not only use the HPSG formalism to express the linguistic analyses of the XTAG grammar, but also preserve, as much as possible, the general framework of the linguistic analyses given in the standard HPSG (Pollard and Sag, 1994) . We use the standard HPSG schemata and the principles that are concerned with syntax, and translate the XTAG elementary trees into lexical feature structures so that they satisfy the conditions 1), 2) and 3) with them. Given that the XTAG features are used for controlling the substitution and adjunction, the condition 3) is reduced to the problem of whether or not all the XTAG features can be rnapped to HPSG feature structures so that their values are properly propageted by the application of those schemata and principles.",
                "cite_spans": [
                    {
                        "start": 207,
                        "end": 230,
                        "text": "(Pollard and Sag, 1994)",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "vVe start with the standard HPSG feature structure and schemata 1 with slight modification and addition. As for principles, we use phonology-, head-feature-, valence-, non-local feature-, specand marker-principles.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation",
                "sec_num": "2"
            },
            {
                "text": "We separate the translations to two steps. First, we translate the tree structure of elementary trees to HPSG feature structures. Second, we map the XTAG feature into the HPSG structure.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation",
                "sec_num": "2"
            },
            {
                "text": "In most initial trees, labels of the nodes on the trunk (the path from the anchor to the root) are the projections of that of the lexical anchor. On the other hand, in HPSG, labels are expressed by apart ofthe HEAD and the VALENCE features. The HEAD feature corresponds to the projection of a category. For example, VP is expressed as a structure whose HEAD is verb and COMPS is saturated and S is expressed as a structure whose HEAD is verb and both COMPS and SUBJ is saturated (Figure 1 ). Thus, if no features are concerned, the nodes on the trunk corresponds to the HEAD feature 2 and we can construct the lexical feature structure corresponding to an initial tree by translating the label of the nodcs on 1 ReCer to (Pollard and Sag, 1994) Ior the meaning oC the standard HPSG Ceatures and schemata: We use the following abbreviations for feature names: SS=SYllSEH, LOC=LDCAL, NONLDC=NDNLOCAL, VAL=VALENCE, HARK=HARKING.",
                "cite_spans": [
                    {
                        "start": 721,
                        "end": 744,
                        "text": "(Pollard and Sag, 1994)",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 479,
                        "end": 488,
                        "text": "(Figure 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "'\"\"'[:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "5 ll 0J ..._ S ssJLOCICA111UJl verb J A / w..i.rs.,.,, l\u2022\u2022l] NP VP l.!:aas IJ S$ILOCICATtHV.ll verb J /\"-,_ w..i.rstl8J C\u2022\u2022iJ ---v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "Lco.>s !In NP Figure 1 : Correspondence between node labels and feature structures the trunk into the HEAD features and rernaining nodes into SUBJ, COMPS or SLASH features according to the syntactic role of the nodes.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 14,
                        "end": 22,
                        "text": "Figure 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "For exarnple, a1 in Fig 2 shows the tree for a transitive verb like. As an HPSG scherna cor- responds to one branching of the tree, the tree can be re-constructed if a proper schema is selected for each branching. In this case, applying head-cornplement, and head-subject in this order will restore the shape of a tree and properly project the information contained in the anchor to the root of the tree 3 \u2022",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 20,
                        "end": 25,
                        "text": "Fig 2",
                        "ref_id": "FIGREF0"
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "LTl\\G ol s-. Jubjtct/'-., ~ Nil> VP 1\u00c4 1 V NPt HPSG S (_ lil<e _l Jubjw ~htad NP Yl' (lil<e _) htad~ complrmt11t rt J campltmer.1 V (llke) NP projtcrion like .................\u2022..\u2022...... \" ... . ~/ . q [~IBONlli~1\u2022~~~,,JjJj :' //;\\,_ VAI{\"SUBJ(NPJJ ~ :: V/ NP l COMPS:tlP J . \u2022\u2022... t\u00b0 \u2022\u2022\u2022\u2022\" ..... \".. CONT~&X : '- \u2022\u2022 \u2022 Jj~e \" ..... \u2022 \u2022 \u2022... NONT..OC SI.As!{ ] i \u2022\u2022\u2022-\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022 .............................. :' \u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022-\u2022\u2022\u2022\u2022\u2022\"-\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "In auxiliary trees, we translate the anchor to the HEAD and leaf node other than the foot node to an element of either SUBJ, CDMPS, or SLASH list, as in initial trees. However, as an auxiliary tree adjoins to a node of another tree, the labe! and the structure of the adjoined tree must be preserved after the adjunction. Therefore, in the application of a schema corresponding to a branching just above the foot node, the HEAD, VAL and SLASH must be preserved ( Figure  3 ). Considering this property and the fact that Figure 3 : Adjunction the auxiliary tree selects the node it adjoins, it seerns natural to put the foot node into the MDD or SPEC feature so that the head-adjunct or head-marker schema will be applied to form the branching just above the foot node in a way that the auxiliary tree becomes the adjunct or the marker (Figure 4 ). As adjunction involves the ",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 463,
                        "end": 472,
                        "text": "Figure  3",
                        "ref_id": null
                    },
                    {
                        "start": 520,
                        "end": 528,
                        "text": "Figure 3",
                        "ref_id": null
                    },
                    {
                        "start": 835,
                        "end": 844,
                        "text": "(Figure 4",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "~ A NP VP VP /A..... Ad,\u2022",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "VP \u2022\u2022\u2022\u2022 /\"-,_ ...... q Adv vp\u2022 1 realty ... \u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022\u2022 \u2022\u2022\u2022\u2022\u2022 \u2022\u2022 .. ;~o[i;;a[~~ a~~: t l i J]] CONT INDEX NONLOC SLASH [ )",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation of the tree structure",
                "sec_num": "2.1"
            },
            {
                "text": "Generally, the HPSG features propagatc from the daughters to the mother as iri Table 1 4 . We put the XTAG features that have an equivalent in the HEAD features of the standard HPSG into XHPSG HEAD features. \\\\'c call these XTAG features HEAD' features. \\Ve observed that the values of HEAD' features are propagated from the node defined as the hea<ldaughter in section 2.1 on the branching that is H:the value propagates from head-daughter to the mother n:the value propagates from head-daughter to the mother e:the value propagates from either daughter A:the value must agree between both daughters U:the value of the feature on the head-daughter unifies the nonhead daughter not the one just above the foot node of an auxiliary tree. We put the agr feature into INDEX feature and trace into the SLASH feature.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 79,
                        "end": 86,
                        "text": "Table 1",
                        "ref_id": "TABREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the feature and the auxiliary trees",
                "sec_num": "2.2"
            },
            {
                "text": "Arnong the remaining features, the ones whose value propagates in the same way as the HEAD' feature values are also put into the XHPSG HEAD feature, and the others are put into the MARKING feature. Table 2 shows where the XTAG features are put into in the XH-PSG feature structure. Now, we re-examine the translation of tree structure regarding foot nodes. We deterrnine the schema to be applied to the branching above the foot node as follows: If the auxiliary tree changes the value of the HEAD feature on adjunction, we apply the head-cornplernent scherna on the branching just above the foot node. In this case, the foot node becomes a complement just as in the case of substitution node 5 ( Figure 5) ; If the auxiliary tree does not change the value of the HEAD node but changes the value of the MARKING feature on adjunction, we apply the head-rnarker schema on the branching just above the foot node. In this case, the foot node becomes the head to be selected by the SPEC feature of the lexical feature structure corresponding to the auxiliary tree ( Figure 6) ; Otherwise, we apply the headsome Figure 6 : Translation of determiners rnodifier schema on the branching just abon\u2022 the foot node. In this case 1 the foot node bC'\u2022 comes the head to be selected by the MOD featun\u2022 of the lexical feature structure corresponding to the auxiliary tree 6 (Figure 4 ).",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 198,
                        "end": 205,
                        "text": "Table 2",
                        "ref_id": "TABREF2"
                    },
                    {
                        "start": 696,
                        "end": 705,
                        "text": "Figure 5)",
                        "ref_id": "FIGREF2"
                    },
                    {
                        "start": 1060,
                        "end": 1069,
                        "text": "Figure 6)",
                        "ref_id": null
                    },
                    {
                        "start": 1105,
                        "end": 1113,
                        "text": "Figure 6",
                        "ref_id": null
                    },
                    {
                        "start": 1357,
                        "end": 1366,
                        "text": "(Figure 4",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Translation of the feature and the auxiliary trees",
                "sec_num": "2.2"
            },
            {
                "text": "Though in most cases the abovementiom\u2022cl translation works, there . are a few exceptional cases. In this section, we mention two casc~. The first one is a treatment of bar level, tlll' second is predicatives.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Problems",
                "sec_num": "3"
            },
            {
                "text": "As mentioned in section 2.1, bar-level is not explicitly marked in the standard HPSG ( SC'C' Figure 1 ), but implicitly stated in VALENCE fcatures. In consequence, there is no distinction between a word who has no arguments and the phrase just consists of that word.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 93,
                        "end": 101,
                        "text": "Figure 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Bar level",
                "sec_num": "3.1"
            },
            {
                "text": "This caused a problem when modifiers are involved. For example, there is no way to prevent a noun-modifying adjective from modifying an NP as there is no distinction between N with no arguments and NP.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Bar level",
                "sec_num": "3.1"
            },
            {
                "text": "To solve the problem we introduced features named XP and ASSIGN...XP. XP is used for restrict the modifiee's bar-level. ASSIGN...XP is used by a modifier to assign a bar-level to a phrase generated as a result of modification.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Bar level",
                "sec_num": "3.1"
            },
            {
                "text": "In XTAG analysis, a predicative noun 7 has a tree whose root is labeled S and the copula be has an auxiliary that adjoins to the tree. We assigned a head feature verb to a predicative noun (see the footnote in section 2.1). However, we could not allow the extraction of the predicative noun, because it would be the head that is extracted. We splitted the lexical entry of be to handle the extraction.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Predicatives and small clause",
                "sec_num": "3.2"
            },
            {
                "text": "We have translated the syntactic lexicon of the XTAG grammar version 1.1 and implemented the translated grammar in LiLFeS language (Makino et al., 1998) . We assumed only binary branching, and splitted the schemata according to whether the head is on the left or on the right.",
                "cite_spans": [
                    {
                        "start": 131,
                        "end": 152,
                        "text": "(Makino et al., 1998)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Implementation",
                "sec_num": "4"
            },
            {
                "text": "Currently we have verified our grammar partially in the sense that XHPSG grammar generates the structures equivalent to the elementary trees and the trees constructed with one or less adjunction. For the general cases, we are currently working on ~onstructing a structure equivalent to the derivation tree for XTAG parsing in XHPSG. The derivation trees willenable us to easily compare the parsing results between the original and the translated grammars to check the validity of XHPSG in a practical sense.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Implementation",
                "sec_num": "4"
            },
            {
                "text": "We optimized the grammar by pre-compiation (Torisawa and Tsujii, 1996) and measured the parsing time of the ATIS corpus using the two-phased parsing of the precompiled XHPSG. The average user time was 1.12 seconds on Alpha Station (400MHz CPU, 4GB main memory). vVe expect a futher speedup of the parsing by packing feature structures (Miyao et al., 1998 We translated the XTAG grammar to get a wide-coverage grammar in the HPSG formalism. By assigning an HPSG schema to a branching of XTAG trees, we have shown that the branching in XTAG trees can be licensed by the standard HPSG schemata and principles.",
                "cite_spans": [
                    {
                        "start": 43,
                        "end": 70,
                        "text": "(Torisawa and Tsujii, 1996)",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 335,
                        "end": 354,
                        "text": "(Miyao et al., 1998",
                        "ref_id": "BIBREF3"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Implementation",
                "sec_num": "4"
            },
            {
                "text": "\u2022we are also interested in comparing our result to the HPSG English grammar being developed at Stanford University (CSLI, 1998) and to the CCG English grammar converted from XTAG grammar (Doran and Srinivas, to appear).",
                "cite_spans": [
                    {
                        "start": 115,
                        "end": 127,
                        "text": "(CSLI, 1998)",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Implementation",
                "sec_num": "4"
            },
            {
                "text": "We are grateful to Prof. Aravind Joshi and the members of the XTAG group for their valuable helps and suggestions.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Acknowledgment",
                "sec_num": "6"
            },
            {
                "text": "'This work is partially Iounded by Japan Society for the Promotion oC Science (JSPS-RFTF96P00502) .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "In a Cew initial trees whose labe! oC the root is different from that of the anchor, we set the HEAD feature according to the root node because oC the substitution .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "The specification of schemata presented in(Pollard and Sag, 1994) ensures that the schemata are applied in proper order. For example, the condition on headsubject schema that the COHPS of the head daughter must be empty ensures that the subject comes above the complements on the tree structure.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "In exceptional cases, the propagation is explicitly marke<l in the lexical item",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "~The tree structu re of the tree that this kind of au xiliary tree adjoins to is kept by letting the appropiriate part of the VALElfCE feature structure-share betweeu the head and the complement.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "In the latter two cases, the tree structure of the trec\u2022 that the auxiliary tree adjoins to is kept by the Yalenc<\u2022 principle and the head-marker (head-adjunct) schema.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "English Resource Gramrnar Online",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "Csli",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "CSLI. 1998. English Resource Gramrnar Online. available in http://hpsg.stanford.edu/hpsg/ergo.html.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Developing a wide-coverage CCG system",
                "authors": [
                    {
                        "first": "Christine",
                        "middle": [],
                        "last": "Doran",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Srinivas",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Christine Doran and B. Srinivas. to appear. Devel- oping a wide-coverage CCG system. CSLI lecture notes edited by A. Abeille and 0. Rarnbow.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "LiLFeS-towards a practical hpsg parser. to appear in",
                "authors": [
                    {
                        "first": "Takaki",
                        "middle": [],
                        "last": "Maldno",
                        "suffix": ""
                    },
                    {
                        "first": "Minoru",
                        "middle": [],
                        "last": "Yoshida",
                        "suffix": ""
                    },
                    {
                        "first": "Kentaro",
                        "middle": [],
                        "last": "Torisawa",
                        "suffix": ""
                    },
                    {
                        "first": "Jun'ichi",
                        "middle": [],
                        "last": "Tsujii",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "Proc. COLING-ACL '98",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Takaki Maldno, Minoru Yoshida, Kentaro Torisawa, and Jun'ichi Tsujii. 1998. LiLFeS-towards a practical hpsg parser. to appear in Proc. COLING-ACL '98.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Packing of feature structures for optirnizing the hpsg-style grarnmar translated from TAG",
                "authors": [
                    {
                        "first": "Yusuke",
                        "middle": [],
                        "last": "Miyao",
                        "suffix": ""
                    },
                    {
                        "first": "Kentaro",
                        "middle": [],
                        "last": "Torisawa",
                        "suffix": ""
                    },
                    {
                        "first": "Yuka",
                        "middle": [],
                        "last": "Tateisi",
                        "suffix": ""
                    },
                    {
                        "first": "Jun'ichi",
                        "middle": [],
                        "last": "Tsujii",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yusuke Miyao, Kentaro Torisawa, Yuka Tateisi, and Jun'ichi Tsujii. 1998. Packing of feature structures for optirnizing the hpsg-style grarnmar translated from TAG. subrnitted to TAG+ work- shop.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Head-Driven Phrase Structure Grammar",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Carl",
                        "suffix": ""
                    },
                    {
                        "first": "Ivan",
                        "middle": [
                            "A"
                        ],
                        "last": "Pollard",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Sag",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Carl J. Pollard and Ivan A. Sag. 1994. Head- Driven Phrase Structure Grammar. University of Chicago Press.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "A lexicalized tree adjoining grammar for english",
                "authors": [
                    {
                        "first": "Xtag Research",
                        "middle": [],
                        "last": "The",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Group",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "The XTAG Research Group. 1995. A lexicalized tree adjoining grammar for en- glish. Technical Report IRCS Research Re- port 95-03, IRCS, University of Pennsylvania. available in http: //WJw. cis. upenn. edu;-xtag/ tech-report/tech-report.html.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Cornputing phrasal-signs in HPSG prior to parsing",
                "authors": [
                    {
                        "first": "Kentaro",
                        "middle": [],
                        "last": "Torisawa",
                        "suffix": ""
                    },
                    {
                        "first": "Jun'ichi",
                        "middle": [],
                        "last": "Tsujii",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "Proc. Coling 96",
                "volume": "",
                "issue": "",
                "pages": "949--955",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kentaro Torisawa and Jun'ichi Tsujii. 1996. Corn- puting phrasal-signs in HPSG prior to parsing. In Proc. Coling 96, pages 949-955.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "type_str": "figure",
                "uris": null,
                "text": "Translation of verb like",
                "num": null
            },
            "FIGREF1": {
                "type_str": "figure",
                "uris": null,
                "text": "Translation of modifiers propagation of other feature values, the translation of an auxiliary trees is re-examined in section 2.2.",
                "num": null
            },
            "FIGREF2": {
                "type_str": "figure",
                "uris": null,
                "text": "Translation of verbs that take a sentential complement",
                "num": null
            },
            "TABREF1": {
                "content": "<table/>",
                "html": null,
                "text": "Propagation of HPSG features",
                "num": null,
                "type_str": "table"
            },
            "TABREF2": {
                "content": "<table><tr><td>IJ\\.TA\u00fc feature</td><td>XH .. \" -leature</td></tr><tr><td colspan=\"2\">assign-ca.se, a.ss1gn-comp, ca.se\"', ex-HEAD</td></tr><tr><td>tracted\u2022, inv*, mainv\u2022, mode*, passive,</td><td/></tr><tr><td>perfect, pred, progressive, pron, tense*</td><td/></tr><tr><td>card, comp, const, decrease, detimte,</td><td>HARKillG</td></tr><tr><td>gen, neg, quan, sub-conj, wh</td><td/></tr><tr><td>agr</td><td>IllDEX</td></tr><tr><td>trace</td><td>SLASH</td></tr><tr><td colspan=\"2\">The features marked w1th * has a counterpart HEAD</td></tr><tr><td colspan=\"2\">feature in tbe standard HPSG analysis.</td></tr></table>",
                "html": null,
                "text": "Correspondence ofXTAG and XHPSG features",
                "num": null,
                "type_str": "table"
            },
            "TABREF3": {
                "content": "<table><tr><td>5 Conclusion and Future Work</td></tr><tr><td>7 adjective and preposition also</td></tr><tr><td>175</td></tr></table>",
                "html": null,
                "text": ").",
                "num": null,
                "type_str": "table"
            }
        }
    }
}