| { | |
| "paper_id": "W98-0135", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T06:05:53.282882Z" | |
| }, | |
| "title": "Whrislands in TAG and Related Formalisms", | |
| "authors": [ | |
| { | |
| "first": "Owen", | |
| "middle": [], | |
| "last": "Rambow", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "CoGenTex, Inc", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Vijay-Shankert", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "W98-0135", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "1 Introduction: TAG and wh-Movement", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The analysis of wh-movement given within TAG is a very convincing argument for the use of a constrained tree-rewriting formalism in syntax, since wh-movement does not require any special mechanism in TAG. wh-movement can be localized to elementary trees, and island effects are obtained naturally. This situation contrasts with approaches based on string-rewriting formalisms such as CFG, which require extensions (mathematical or at any rate definitional) to the basic mathematical formalism (resulting in theories such as GPSG, HPSG, LFG, or transformational grammar).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "However, the question arises how other treerewriting formalisms such as D-Tree Grammar (Rambow et al., 1995) can handle whmovement. Specifically, the question arises whether an equally elegant solution to the problem of wh-movement can be found. In this paper, we propose to study e:icactly which what features of the formal (mathematical) definition of TAG contribute to the correct analysis of whmovement (in English). We will mainly concentrate on TAG, but occasionally mention treelocal MC-TAG.", | |
| "cite_spans": [ | |
| { | |
| "start": 87, | |
| "end": 108, | |
| "text": "(Rambow et al., 1995)", | |
| "ref_id": "BIBREF3" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The paper is structured as follows. In Section 2, we present the relevant elements of the definition of TAG. We then proceed to discuss specific island types and how these can be expressed in TAG: relative ciause and other adjunct ishmd:s in Section 3, sentential subject islands in Section 4, and wh-islands in Section 5.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In this paper, we will distinguish the following elcments of the definition of TAG. (For a full mathematical definition, see (Vijay-Shanker, 1987) .)", | |
| "cite_spans": [ | |
| { | |
| "start": 125, | |
| "end": 146, | |
| "text": "(Vijay-Shanker, 1987)", | |
| "ref_id": "BIBREF4" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 The extended domain of locality (EDL). In TAG, the elementary structures are trees (rather than strings), so we can state extensive linguistically motivated restrictions on the shape of the elementary trees of a grammar. In fact, any such linguistic restriction on the shape of elementary structures exploits EDL.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 The geometry of adjunction (GA). By this term, we mean the specific, mathcmatical defiuition of the adjunction operation in TAG and, especially, the shape of the resulting derived tree. Specifically, an auxiliary tree \u00df has a designated footnode; when \u00df is adjoined in a tree a at node v, it is inserted in its entirety into a. In the process, \u00df remains intact, but a is divided in two subtrees at node v, with \u00df now attached at v and the subtree formerly rooted in v now attached to the footnode of \u00df.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 The factoring of recursion (FR). By definition, in an auxiliary tree \u00df, the footnode and the root node must have the same }abel, A. Furthermore, \u00df can only be adjoined at a node labeled A. We observe that this aspect of the definition of TAG is not essential in the sense that the restrictions could be lifted without affecting the remainder of the definition, in particular the geometry of adjunction. The crucial part for the geometry of adjunction is the presence of a footnode; its labe! does not a prior\u2022i matter.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We observer that tree-local MC-TAG has the same notion of EDL as TAG, and it ha.s it own notion of GA. FR is limited to those cases in which adjuncti\u00f6n of on eof the component trees takes place.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "By definition, any other tree-rewriting system 1 will also have EDL, while GA and FR are specific to TAG. Thus, we are in particular interested in the extent to which GA and FR are used in deriving island constraints, since such use would not necessarily carry over to other tree-rewriting systems.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "In the following, we will be making an important assumption. Because of the EDL of the elementary structures of TAG, it is possible to lexicalize TAG in a straightforward manner (Schabes, 1990) , meaning that each elementary tree in a grammar is associate<l with exactly one lexical item. Furthermore, we can require that each tree corresponding to a lexical item has positions (substitution nodes or a footnode) corresponding to each syntactic argument of that lexical item, and that the derivation thus refl.ects the syntactic relation between the lexical items involved (Rambow and Joshi, 1996) { the \"lexical derivation constraint\"). In this paper, we will only be interested in lexicalized grammars and in derivations that conform to the lexical derivation constraint.", | |
| "cite_spans": [ | |
| { | |
| "start": 178, | |
| "end": 193, | |
| "text": "(Schabes, 1990)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 573, | |
| "end": 597, | |
| "text": "(Rambow and Joshi, 1996)", | |
| "ref_id": "BIBREF2" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Elements of the Definition of TAG", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Sentence-initial extraction from certain adjuncts such as relative clauses modifying nonfronted object NPs or VP sentential adjuncts is ruled out simply by GA (in conjunction with the lexical derivation constraint). lt is simply impossible to adjoin (or substitute) a tree into a (non-fronted) object, or adjoin a tree at a VP node (in a tree which has a subject NP to the left of the VP node), and obtain a derived tree in w hich some part of the adjoined tree is now in sentence-initial position.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "In contrast, it is quite possible to adjoin a relative clause to a s~bject or adjoin an S-adjunct to a clausal tree (i.e., and adjunct phrase rooted in S), and obtain a wh-extraction to sentenceinitial position. A sample auxiliary tree that would result in illicit extraction is shown in Fig-1 \\Vc includc in this category systems which operate on trcc-like structures. . This tree can be ruled out in several different ways resorting to linguistic arguments. For example, one could exclude the tree by saying that extraction beyond the root node of an adjunct is impossible since the root node is not part of the projection of the lexeme anchoring the adjunct, or one could say that the tree in Figure 1 is illicit because of independently formulated constraints on node labels. In any case, one would be exploiting the EDL to express linguistically motivated constraints on the shape of elementary structures in the grammar. But, crucially, these constraints would carry over to the case of the relative clause modifying an object NP, and to the case of the VP-adjunct: it is not plausible that the linguistic constraints would be formulated in such a way that they only apply to subject relative clauses (or S adjuncts), but not to object relative clauses (or VP adjuncts). Thus, these cases are redundantly rules out by GA.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 288, | |
| "end": 293, | |
| "text": "Fig-1", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 696, | |
| "end": 704, | |
| "text": "Figure 1", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Furthermore, there is a point that is easily overlooked. While object relative clauses with sentence-initial fronting are ruled out by GA, we also need to rule out non-initial fronting:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "(1) *I saw whatithe man who was wearing ti", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "While these kinds of sentences may be pathologically bad, they still need to be ruled out in a TAG grammar", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "We close by observing that if we are using treelocal MC-TAG, an argument very similar to the one above can be made to demonstrate that any predictive power obtained from the geometry of tree-local multicomponent adjunction is redundant with respect to independently required linguistic restrictions on the shape of the elementary tree sets. We omit the details.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "4 Sentential Subjects lt would be possible to derive extraction from sentential subjects in the same manner that we derive extraction from sentential objects, namely by adjoining a matrix clause of the type shown in Figure 2 into the subordinate clause. In order to exclude such a derivation, we must say that the subject position, even when labeled S, cannot be a footnode. Tims, simply saying that we have factoring of recursion does not limit the extraction patterns: we must, in addition, make a linguistically motivated choice among possible footnodes. Designating a footnode is equivalent to allowing extraction from that position. However, the designation of the footnode is not sufficient. This is because of a wellknown asymmetry in extractiou from picture-NPs: while extraction from certain object NPs is possible, extraction from subject NPs never IS.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 216, | |
| "end": 224, | |
| "text": "Figure 2", | |
| "ref_id": "FIGREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "(2) a. Whati did you buy a picture of ti? b. * Wltati did a picture of ti fall on your head?", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Tlrns, if we use tree-local multicomponent MC-TAG to derive picture-NP cxtraction by sub-stituting the main NP and substituting or adjoining the extracted wh-element, we must still specifically rule out extraction from subject position in some manner. 2 Furthermore, the same problem arises when we want to distinguish between verbs that allow picture-NP extraction and those that do not ( as readily). Therefore, we will need some formal device (say, a feature EXTRACT on frontier nodes which regulates multicomponent derivations across them) for blocking extraction from certain positions in addition to the choice of footnodes. (This will also exclude extraction from sentential subjects if these are analyzed as projecting to NP.) The use of the device will need to be linguistically motivated. Some sort of equivalent device with similar linguistic motivation for its use can be used in tree rewriting systems which do not. have FR or GA.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Relative Clause Islands and Other Adjunct Islands", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "In English, we can exclude some wh-islands by restricting the shape of elementary trees.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "(3) *Whati do you know whomi Mary gave ti ti?", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "(3) is excluded because the elementary tree for give, which would need t have two wh-moved elements, is already excluded (we never have multiple wh-movement in English elementary trees). This analysis exploits the EDL and transfers to other tree-rewriting formalisms.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "But that does not cover all cases of wh-islands.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "(4) *Whati do you know whomj Mary told ti that she had bought ti?", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "In 4, there is only one wh-extraction per elemeutary verbal tree. These cases can be excluded in several ways, but they all use FR. We take (Frank, 1992) as the most advanced example. There, trees in which wh-extraction from below takes place are footed in C' and (hence by FR) are rooted in C', while those without wh-extraction from below are both footed and rooted in CP. This ensures that if there is a wh element below ( and assuming wh elements are always in SPEC(CP)), then the tree below must project to CP, and then the foot node must be CP, and hence the root node as well. Therefore, there is no room for a further wh element up front that would come from below. Note that if there is a single wh-movement at any depth of embedding, then because of the recursion part of FR, all trees above it must be CP-footed-androoted as well.", | |
| "cite_spans": [ | |
| { | |
| "start": 140, | |
| "end": 153, | |
| "text": "(Frank, 1992)", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Frank's analysis makes use of several linguistic constraints on elementary structures (exploiting EDL), among which:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "1. In an elementary tree, a C' may never dominate a CP.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "2. An elementary tree may not have two CP nodes one immediately dominating the other (the \"anti-CP-recursion stipulation\").", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "3. Each tree can only contain a single lexical item and its projection and ( crucially) no part of a different lexical item 's projection. Otherwise, we could have (did) (john} wonder whether in one tree which is rooted and footed in C'. Such a tree would allow sentences such as *Who did John wonder whether Sue saw?.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Given these linguistic constraints as well as FR, it is impossible to obtain a node labeled CP immediately dominating a wh-element on the path separating a \"moved\" wh-element from the rest of its tree.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "In tree rewriting systems that do not have FR, it \\Vill be necessary to derivc thc path constra.int in some other manner. In DTG, it is possible to include path constraints explicitly in the elementary structures. In such an approach, the linguistic restrictions can be relaxed; it is not necessary to assume the anti-CP-recursion constraint, for example, and it would even be possible to allow an inversion of CP and C'.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Wh-Islands", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "In conclusion, we have seen tliat for relative clause islands and clausal adjunct islands, and for sentential subject islands, the TAG analysis exploits EDL but not GA or FR. These analyses would therefore carry over to other treerewriting systems. In the case of wh-islands, FR is exploited in conjunction with several linguistic EDL-type constraints in order to limited the occurrence of certain nodes on the path of wh-\"movement\". While this can not be replicated exactly in a system without FR, any other device to restrict the path has the same effect.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "6" | |
| }, | |
| { | |
| "text": "Kroch (1989) suggests instead that the traces of picture-NP extractions are found in the eiementary structures of the main verb. They are not license<l in the verbal tree because not bound; an index is adjoined through multi-component a<ljunction (along with the whelement), which provides the binding. However, unlike traces in object position, traces in subject position are never licensed to begin with. This analysis exploits the EDL and could be expressed in other tree-rewriting formalisms as weil.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Syntactic Locality and Tree Adjoining Grammar: Grammatical, Acquisition and Processing Perspectives", | |
| "authors": [ | |
| { | |
| "first": "Robert", | |
| "middle": [], | |
| "last": "Frank", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Frank, Robert (1992). Syntactic Locality and Tree Adjoining Grammar: Grammatical, Acqui- sition and Processing Perspectives. PhD thesis, Department of Computer and Information Sci- ence, University of Pennsylvania.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Asymmetries in long distance extraction in a Tree Adjoining Grammar", | |
| "authors": [ | |
| { | |
| "first": "Anthony", | |
| "middle": [], | |
| "last": "Kroch", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "Alternative Conceptions of Phrase Structure", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "66--98", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kroch, Anthony (1989). Asymmetries in long distance extraction in a Tree Adjoining Gram- mar. In Baltin, Mark and Kroch, Anthony, ed- itors, Alternative Conceptions of Phrase Struc- ture, pages 66-98. University of Chicago Press.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A formal look at dependency grammars and phrase-structure grammars, with special consideration of word-order phenomena", | |
| "authors": [ | |
| { | |
| "first": "Owen", | |
| "middle": [], | |
| "last": "Rambow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Aravind", | |
| "middle": [], | |
| "last": "Joshi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1996, | |
| "venue": "Current Issues in Meaning-Text Theory", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Rambow, Owen and Joshi, Aravind (1996). A formal look at dependency grammars and phrase-structure grammars, with special con- sideration of word-order phenomena. In Wan- uer, Leo, editor, Current Issues in Meaning- Text Theory. Pinter, London.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Mathematical and Computational Aspects of Lexicalized Grammars", | |
| "authors": [ | |
| { | |
| "first": "Owen", | |
| "middle": [ | |
| ";" | |
| ], | |
| "last": "Rambow", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Vijay-Shanker", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Weir", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "33rd Meeting of the Association for Computational Linguistics (ACL'95}", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "151--158", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Rambow, Owen; Vijay-Shanker, K.; and Weir, David (1995). D-Tree Grammars. In 33rd Meet- ing of the Association for Computational Lin- guistics (ACL'95}, pages 151-158. ACL. Schabes, Yves (1990). Mathematical and Com- putational Aspects of Lexicalized Grammars. PhD thesis, Department of Computer and In- formation Science, University of Pennsylvania.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "A study of Tree Adjoining Grammars", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Vijay-Shanker", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Vijay-Shanker, K. (1987). A study of Tree Ad- joining Grammars. PhD thesis, Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "type_str": "figure", | |
| "text": "Relative clause with wh-moved element ure 1", | |
| "uris": null, | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "type_str": "figure", | |
| "text": "Matrix clause with sentential subject", | |
| "uris": null, | |
| "num": null | |
| } | |
| } | |
| } | |
| } |