| { |
| "paper_id": "2020", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T15:39:02.533104Z" |
| }, |
| "title": "The Subregular Complexity of Syntactic Islands", |
| "authors": [ |
| { |
| "first": "Nazila", |
| "middle": [], |
| "last": "Shafiei", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Brook University Stony Brook", |
| "location": { |
| "postCode": "11794", |
| "region": "NY", |
| "country": "USA" |
| } |
| }, |
| "email": "nazila.shafiei@stonybrook.edu" |
| }, |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Brook University Stony Brook", |
| "location": { |
| "postCode": "11794", |
| "region": "NY", |
| "country": "USA" |
| } |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "We provide a formal framework for analyzing syntactic island effects from a subregular perspective. Key aspects of the syntactic representation are encoded as strings where precedence represents containment. Island effects then are expressed as constraints on the shape of these strings. The constraints fit in the class IBSP (Interval-Based Strictly Piecewise), which has been previously explored in subregular phonology. Consequently, the characterization of islands in terms of IBSP string constraints not only provides a computational upper bound on the inventory of feasible island effects, but also establishes a surprising link between syntax on the one hand and phonology on the other.", |
| "pdf_parse": { |
| "paper_id": "2020", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "We provide a formal framework for analyzing syntactic island effects from a subregular perspective. Key aspects of the syntactic representation are encoded as strings where precedence represents containment. Island effects then are expressed as constraints on the shape of these strings. The constraints fit in the class IBSP (Interval-Based Strictly Piecewise), which has been previously explored in subregular phonology. Consequently, the characterization of islands in terms of IBSP string constraints not only provides a computational upper bound on the inventory of feasible island effects, but also establishes a surprising link between syntax on the one hand and phonology on the other.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "The subregular program is concerned with analyzing the complexity of linguistic dependencies that are at most regular. The program has found great success in computational phonology (see Heinz 2018 and references therein), where it has resulted in a computational typology of phonological patterns and corresponding learning algorithms. Syntax, by virtue of being mildly context-sensitive, may seem far beyond the purview of the subregular program. But syntax is also subregular once one considers more suitable representations. Two routes have been explored: lifting subregular classes from strings to trees (Graf, 2018b; Vu et al., 2019) , and putting string constraints on particular path languages of syntactic trees . Whereas the former has been mostly used in the analysis of structure building operations, the latter has been applied to syntactic constraints such as NPI-licensing. This paper focuses on an area where these two aspects of syntax meet: island constraints. Island constraints impose additional restrictions on displacement, which in the tradition of Transformational grammar is equated with the operation Move. The shape of islands is narrowly circumscribed, indicating that they are very limited from a computational perspective. In this paper, we confirm this intuition. Island constraints are expressed as constraints over a path language where linear precedence in the string encodes (a specific notion of) containment. Given such a string representation, island constraints fall into the subregular class Interval-Based Strictly Piecewise (IBSP), which has been argued to play a central role in phonology (Graf, 2017 (Graf, , 2018a . At the same time, IBSP is sufficiently weak to rule out many unattested island constraints. Our paper thus makes several contributions: it deepens our understanding of subregular syntax, establishes parallels to phonology, and provides linguists with a computational theory of islands.", |
| "cite_spans": [ |
| { |
| "start": 609, |
| "end": 622, |
| "text": "(Graf, 2018b;", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 623, |
| "end": 639, |
| "text": "Vu et al., 2019)", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 1632, |
| "end": 1643, |
| "text": "(Graf, 2017", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 1644, |
| "end": 1658, |
| "text": "(Graf, , 2018a", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Due to space constraints, we focus largely on strong islands, and only on the canonical cases for most of them. We also investigate the that-trace constraint and the coordinate structure constraint, and we show that they cannot be handled in the system proposed here. This paper thus marks but the first step towards a fully articulated, empirically grounded theory of islands.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The discussion proceeds as follows: the preliminaries section ( \u00a72) discusses Minimalist grammars ( \u00a72.1), our string representation format ( \u00a72.2), and the subregular class IBSP ( \u00a72.3). Section 3 presents the central result that a number of (strong) island constraints follow a uniform IBSP pattern of very low complexity. We start with the adjunct island constraint ( \u00a73.1) and then generalize the analysis to wh-islands, the complex np constraints, the subject condition, and freezing effects ( \u00a73.2). Section 4 then explores the limits of IBSP over a-strings. On the one hand this allows us to correctly rule out many unattested island con-straints, but it also means that the approach cannot handle all aspects of the that-trace constraint and the coordinate structure constraint. In addition, our approach currently lacks any notion of linguistic naturalness, which allows for some very odd (albeit computationally simple) island constraints ( \u00a75).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The paper rests on several research traditions, which are briefly sketched in this section: Minimalist grammars as a formal model of syntax ( \u00a72.1), string representations for syntax ( \u00a72.2), and the subregular class of IBSP string languages ( \u00a72.3).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Preliminaries", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Since island constraints have mostly been studied in the generative tradition, we adopt Minimalist grammars (MGs; Stabler, 1997 ) as a formal model of syntax. MGs are a derivational grammar formalism for building tree structures by combining feature-annotated lexical items via the operations Merge and Move. Figure 1 gives a concrete example of this process. Only a few key aspects of MGs matter for this paper, in particular their feature system (see Stabler 2011 for a full discussion).", |
| "cite_spans": [ |
| { |
| "start": 114, |
| "end": 127, |
| "text": "Stabler, 1997", |
| "ref_id": "BIBREF12" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 309, |
| "end": 317, |
| "text": "Figure 1", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Minimalist Grammars", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "Each lexical item consists of a phonetic exponent and a string of features. There are four distinct types of features. Category features (X ) and selector features (X + ) establish head argument relations via Merge. The other two feature types drive the operation Move. A licensee feature f indicate that the phrase headed by the lexical item undergoes f-movement, and the matching licensor feature f + indicates the landing site of fmovement. As in Minimalist syntax, movement is a mechanism for displacing subtrees of an already assembled tree, and movement always targets the closest available landing site (encoded in MGs via licensor features).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Minimalist Grammars", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "Given the special role of adjuncts in island constraints, we also adopt the adjunction mechanism of (Frey and G\u00e4rtner, 2002) . Instead of a category feature, a lexical item l may carry an adjunction feature X \u21e0 which allows it to adjoin to an XP. An MG's structure building process is usually represented as a derivation tree like the one in Fig. 1 . But we will frequently represent derivation trees with the more compact format of dependency trees. The rightmost tree in Fig. 1 presents a concrete example.", |
| "cite_spans": [ |
| { |
| "start": 100, |
| "end": 124, |
| "text": "(Frey and G\u00e4rtner, 2002)", |
| "ref_id": "BIBREF1" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 342, |
| "end": 348, |
| "text": "Fig. 1", |
| "ref_id": null |
| }, |
| { |
| "start": 473, |
| "end": 479, |
| "text": "Fig. 1", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Minimalist Grammars", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "Our investigation of island constraints will not operate directly over trees, but rather over strings that represent specific aspects of the tree structure. This follows recent work by , who analyze syntactic constraints such as NPI-licensing and Principle A as operating over strings that encode asymmetric c-command relations. A tree is well-formed iff it holds for every node n in the tree that the relevant string representation for n is well-formed with respect to the syntactic string constraints.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "String Representations for Syntax", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Graf and Shafiei (2019) choose a string representation that encodes both containment and a limited form of c-command (cf. Frank and Vijay-Shanker, 2001 ). These augmented command strings (or simply c-strings) can be defined in various ways, but the easiest option uses MG dependency trees. We adopt this definition but simplify it so that the resulting string representation only keeps track of containment. For this reason, we call these strings ancestor strings (or simple astrings).", |
| "cite_spans": [ |
| { |
| "start": 122, |
| "end": 151, |
| "text": "Frank and Vijay-Shanker, 2001", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "String Representations for Syntax", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Definition 1 (A-strings). Let t be an MG dependency tree. If n is the root of t, then as(n) := n. If n has mother m, then as(n) := n as(m). y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "String Representations for Syntax", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Example. In Fig. 1 , as(Mary :: D nom ) = Mary :: D nom buy :: D + D + V \" \" :: V + nom + T \" did :: T + wh + C . For increased readability, we may omit features and replace empty heads by their category. Then as(Mary) = Mary buy T did. y", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 12, |
| "end": 18, |
| "text": "Fig. 1", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "String Representations for Syntax", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "The use of strings is a matter of mathematical convenience. The results obtained this way can be backported to subregular machinery that operates directly on dependency trees or derivation trees (Graf and De Santo, 2019 ). This will be discussed further in \u00a75.", |
| "cite_spans": [ |
| { |
| "start": 195, |
| "end": 219, |
| "text": "(Graf and De Santo, 2019", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "String Representations for Syntax", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Formal language theory has a rich tradition of studying proper subclasses of the regular string languages (McNaughton and Papert, 1971; Pin, 1997; Yli-Jyr\u00e4, 2005, a.o.) . More recently, this line of work has been picked up and extended by computational phonologists (see Heinz 2018 and references therein). The class Interval-Based Strictly Piecewise (IBSP) was proposed as a linguistically natural unification of previously pro- posed classes for subregular phonology (Graf, 2017 (Graf, , 2018a . IBSP constitutes an approximate upper bound on string dependencies in phonology.", |
| "cite_spans": [ |
| { |
| "start": 106, |
| "end": 135, |
| "text": "(McNaughton and Papert, 1971;", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 136, |
| "end": 146, |
| "text": "Pin, 1997;", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 147, |
| "end": 168, |
| "text": "Yli-Jyr\u00e4, 2005, a.o.)", |
| "ref_id": null |
| }, |
| { |
| "start": 271, |
| "end": 281, |
| "text": "Heinz 2018", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 469, |
| "end": 480, |
| "text": "(Graf, 2017", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 481, |
| "end": 495, |
| "text": "(Graf, , 2018a", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "CP DP o D 0 D which NP N 0 N car C 0 C did TP DP s D 0 D Mary T 0 T \" VP VP t s V 0 V buy t o", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Definition 2 (k-val). A segmented k-interval (k 0) over alphabet \u2303, or simply segmented k-val, is a tuple hL, R, F i i 0\uf8ffi\uf8ffk such that \u2022 L, R \u2713 \u2303[{\"", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "} specify the left edge and right edge, respectively, and", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "\u2022 F i \u2713 \u2303 specifies the i-th filler slot. y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Definition 3 (IBSP-k). Let \u2303 be some fixed alphabet and o, n / 2 \u2303 two distinguished symbols. An IBSP-k grammar over \u2303 (k 0) is a pair", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "G := hi, Si, where i is a segmented k-val over \u2303 [ {o, n} and S \u2713 (\u2303 [ {o, n}) k is a set of forbidden k-grams. A string s 2 \u2303 \u21e4 is generated by G iff there is no k-gram u 1 \u2022 \u2022 \u2022 u k 2 S such that o k sn k is a member of the language (\u2303 [ {o, n}) \u21e4 \u2022 L \u2022 F \u21e4 0 \u2022 {u 1 } \u2022 F \u21e4 1 \u2022 {u 2 } \u2022 . . . \u2022 F \u21e4 k 1 \u2022 {u k } \u2022 F \u21e4 k \u2022 R \u2022 (\u2303 [ {o, n}) \u21e4", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "The language L(G) is the set of all s 2 \u2303 \u21e4 that are generated by", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "G. A stringset L is IBSP-k iff L = L(G) for some IBSP-k grammar G. y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "In the definition above, \u21e4 represents the usual Kleene closure. The symbol \u2022 denotes string concatenation, lifted to sets:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "A \u2022 B := {ab | a 2 A, b 2 B}.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Following , we can use IBSP grammars over strings to regulate the shape of trees.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Definition 4 (IBSP over trees). Let G be an IBSP grammar, t an MG dependency tree, and f t a total function from nodes of t to strings. Then t is well-formed with respect to G iff it holds for all lexical items l in t that f t (l) is generated by G. y Example. Suppose that nom-movement is forbidden out of VPs. Over a-strings, this corresponds to the requirement that no V-head may occur between a node with nom and the closest node with nom + (because movement in MGs always targets the closest head with a matching feature). This can be expressed as the following IBSP-1 grammar:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "L := l | l carries nom F 0 := l | l does not carry nom + F 1 := l | l does not carry nom + R := l | l carries nom + S := l | l carries V", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Then as(Mary) = Mary buy T did will be deemed illicit because it matches a forbidden pattern with L := Mary, F \u21e4 0 := \", F \u21e4 1 := \", R := T, and buy 2 S. Consequently, the dependency tree is not well-formed, either. y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "It is often convenient to represent IBSP grammars in a more visual format. The example grammar above corresponds to the diagram below.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "(1) Graphical representation of an IBSP grammar", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "nom V nom + \u00acnom + \u00acnom +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "The outermost vertical boxes represent the left and right edge, respectively. The square in the middle represents a position of the forbidden k-grams -since the example grammar uses forbidden unigrams, there is only one such square. The vertically offset boxes represent the fillers, in this case F 0 and F 1 . We use features to as a shorthand for the set of lexical items that carry this feature. For instance, nom denotes the set of all lexical items carrying nom . The expression \u00acnom + denotes the of all lexical items that do not carry the relevant feature, in this case nom + . This visual format can also be used to show that a string is ill-formed.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Example. Recall that as(Mary) = Mary buy T did is illicit. We can show this by giving a specific instantiation of the interval and the k-gram in the string.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "Mary buy T did \" \"", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "This disagram conveys the same information as the formal description in the previous example. y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "In the next section, we use this machinery to analyze syntactic island effects from a subregular perspective. We show that strong islands follow a fixed IBSP pattern over a-strings that is exceedingly simple.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Subregular Complexity", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "The notion of syntactic islands originates from Transformational Grammar (Ross, 1967) . From the perspective of MGs, a constituent C is an island iff no phrase contained by C may have a licensee feature checked by a matching licensor feature outside C. A distinction is commonly made between strong islands and weak islands. Strong islands limit movement irrespective of whether the mover is an argument or an adjunct. Weak islands, on the other hand, limit adjunct movement but not argument movement. We will focus mostly on strong islands in this paper. We first analyze the adjunct island constraint ( \u00a73.1) as an IBSP perspective over a-strings, and then show how the same template can be used for several other strong island effects ( \u00a73.2).", |
| "cite_spans": [ |
| { |
| "start": 73, |
| "end": 85, |
| "text": "(Ross, 1967)", |
| "ref_id": "BIBREF11" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Strong Islands over A-Strings", |
| "sec_num": "3" |
| }, |
| { |
| "text": "The adjunct island constraint is arguably the most robust case of a strong island. It is illustrated by the contrast in (2).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "(2) Adjunct island constraint a. In both (2a) and (2b) the wh-phrase which car moves out of an embedded clause. But in (2a) the embedded clause is an argument of the verb complain, whereas it is an adjunct in (2b). That movement is allowed out of the argument clause but not the adjunct clause is referred to as the adjunct island constraint. The adjunct island constraint can be easily expressed as an IBSP constraint on a-strings. In fact, it uses the template we already encountered in \u00a72.3, except that the set of forbidden unigrams consists of all adjuncts rather than all verbs. As explained in \u00a72.1, we adopt the proposal of Frey and G\u00e4rtner (2002) that every adjunct carries an adjunct feature X \u21e0 that allows it to adjoin to XPs. The IBSP grammar for the adjunct island constraint thus corresponds to the following template.", |
| "cite_spans": [ |
| { |
| "start": 632, |
| "end": 655, |
| "text": "Frey and G\u00e4rtner (2002)", |
| "ref_id": "BIBREF1" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "(3) IBSP-1 grammar for adjunct islands", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "f X \u21e0 f + \u00acf + \u00acf +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Note that this template actually represents multiple IBSP grammars as f must be correctly instantiated for each movement feature: nom for subject movement, wh for wh-movement, top for topicalization, and so on. If all of those were put inside a single IBSP grammar, then one lose the fact that the left edge and the right edge must be opposite polarities of the same feature -a nom for the left edge could be paired up with an wh + as the right edge. Since IBSP lacks a direct means of coordinating left and right edges like this, we instead have to posit a separate grammar for each movement feature in order to correctly enforce the adjunct island constraint for that specific movement type.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Example. Figure 2 shows the MG dependency tree for which topic did you leave because Mary talked about . This sentence contains illicit wh-movement out of an adjunct. Now consider the a-string for which, with the relevant features indicated in square brackets: as(which) = which[wh ] talked about T because", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 9, |
| "end": 17, |
| "text": "Figure 2", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "[V \u21e0 ] leave T did[wh + ].", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "As shown by the diagram below, this astring is ill-formed with respect to the IBSP grammar in (3) (assuming f := wh). As the a-string is illicit, the whole sentence is illformed. y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The reader may wonder why the template explicitly forbids f + as fillers. This ensures that the right edge is always the closest f + , which is the one targeted for movement by the lexical item with f in the left edge. Without this restriction, the IBSP grammar would incorrectly rule out wellformed movement patterns.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Example. Consider once more the example sentence which topic did you leave because Mary talked about as depicted in Fig. 2 . This sentence contains two instances of nom-movement, both of which are well-formed. But now consider the IBSP grammar regulating nom-movement. Suppose that this grammar allows for lexical items with nom + to appear in the filler slots. Then this grammar would incorrectly rule out as(Mary)", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 116, |
| "end": 122, |
| "text": "Fig. 2", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "= Mary[nom ] talked about T[nom + ] because[V \u21e0 ] leave T[nom + ] did. Mary because T did talked about T leave y", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The reader should also keep in mind that the use of X \u21e0 is just a notational shorthand for specifying a list of lexical items. One can remove some items from this set to allow for exceptions to the adjunct island constraint, such as the ones noted by Truswell (2007 Assuming a distinction between finite T-heads (T ) and other T-heads (T inf ), we can account for this by excluding while :: T + inf V \u21e0 from the list of forbidden lexical items.", |
| "cite_spans": [ |
| { |
| "start": 251, |
| "end": 265, |
| "text": "Truswell (2007", |
| "ref_id": "BIBREF14" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "In sum, the adjunct island constraint can be handled by a very simple and intuitive IBSP grammar (or rather, a collection of such grammars, one for each movement type). From a formal perspective, this IBSP grammar looks very similar to the IBSP treatment of blocking effects in phonology. In phonology, an intervening consonant cluster may block long-distance harmony. In syntax, an intervening head with an adjunction feature interrupts the dependency between an f and an f + . The existence of the adjunct island constraint thus becomes a bit less mysterious: it is very simple from a computational perspective, and it employs a general blocking mechanism that also seems to be active in other parts of language.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adjunct Island Constraint", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Besides the adjunct island constraint, the class of strong islands also includes wh-islands, complex NPs, and subjects. The corresponding constraints are illustrated below. These all use minor variations of the template for the adjunct island constraint. Let us start with the wh-island constraint. Here it suffices to make two changes. Since most types of movements, e.g. topicalization, are not affected by this constraint, we limit the possible instantiations for f and f + to just wh and wh + , respectively. Then the list of blockers is changed from adjuncts to all elements that induce wh-islands. These are commonly taken to be all C-heads that have some kind of question semantics, including whether, how, and if. We denote this set C [Q].", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "(8) IBSP-1 grammar for wh-islands", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "wh C [Q] wh + \u00acwh + \u00acwh +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Next we turn to the complex NP constraint. This one, too, uses the basic template of the adjunct island constraint, but we once again have to change the list of blockers. In the complex NP constraint, the blocking is not done by an adjunct, but by a more complex structural configuration: movement out of a CP is illicit if the CP is the argument of a noun. Thanks to the MG feature calculus, we can rephrase this as a ban against moving out of an NP that selects a CP, 1 which means that the set of blockers contains all lexical items, and only those, that contain a selector feature C + and a category feature N . We denote this set of lexical items by", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "C + \u2022 \u2022 \u2022 N .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "(9) IBSP-1 grammar for complex NP constraint", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "f C + \u2022 \u2022 \u2022 N f + \u00acf + \u00acf +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "The reader is invited to verify that this grammar correctly rules out the sentence what did you hear rumors that John bought, which is depicted in Fig. 3 . This leaves us with the subject condition, which can actually be regarded as an instance of what is known as freezing effects. This describes the phenomenon that once a phase XP has undergone movement, it becomes opaque to extraction. Any mover inside XP has to move out of the phrase before it starts moving. From the perspective of MGs, this can be rephrased as a constraint on the distribution of movement features. Let f 1 and g 1 1 Our feature-based interpretation of the complex NP constraint is actually stronger than the original version. Suppose that the NP selects a CP as its complement and some XP as its specifier. The complex NP constraint as originally stated would allow the XP to be extracted, whereas our version does not. As far as we have been able to determine, there are no nouns that take two arguments in this configuration, let alone one where the XP then is allowed to undergo movement. denote lexical items whose first movement feature is f and g , respectively. If the phrase headed by g 1 contains f 1 , then the target of f 1 must be contained by the target of g 1 . We can capture this generalization by moving from an IBSP-1 grammar to an IBSP-2 grammar (or rather, a collection of such grammar for every possible choice of f and g ).", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 147, |
| "end": 153, |
| "text": "Fig. 3", |
| "ref_id": "FIGREF3" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "(10) IBSP-2 grammar for freezing effects", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "f 1 g 1 g + f + \u00acf + \u00acf + , \u00acg + \u00acf +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "The step up from IBSP-1 to IBSP-2 makes freezing effects appear more complex. But it is actually possible to get the same effect just with an IBSP-1 grammar. The trick is to make g + the right edge of the k-val rather than f + . (11) IBSP-1 grammar for freezing effects", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "f 1 g 1 g + \u00acf + \u00acf + , \u00acg +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Example. Consider the abstract a-string f 1 a b g 1 m n g +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "x y f + z. Both grammars correctly rule it out as illicit.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "f 1 g 1 g + f + z a b m n x y f 1 g 1 g + x y f + z a b m n", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Similarly, both grammars agree that the minimally different f 1 a b g 1 m n f + x y g + z is wellformed. y Both grammars also agree that the tree in Fig. 4 is illicit because of the ill-formed a-string of who.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 149, |
| "end": 155, |
| "text": "Fig. 4", |
| "ref_id": "FIGREF4" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Note that we can apply the same kind of truncation strategy to the IBSP grammars for the other island constraints. This effectively reduces their complexity of IBSP-1 to IBSP-0. As laid out in Def. 3, an IBSP-0 grammar consists only of the left edge L, the right edge R, and a single filler F 0 inbetween. The set of forbidden k-grams is immaterial as every string is ruled out that matches", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "(\u2303 [ {o, n}) \u21e4 \u2022 L \u2022 F \u21e4 \u2022 R \u2022 (\u2303 [ {o, n}) \u21e4 . (12) IBSP-0 grammar for adjunct islands f X \u21e0 \u00acf +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "(13) IBSP-0 grammar for wh-islands", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "wh C [Q]", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "\u00acwh + (14) IBSP-0 grammar for complex NP constraint", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "f X \u21e0 \u00acf +", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "In sum, all four island constraints can be captured with very simple IBSP-1 grammars (or even IBSP-0 grammars) over a-strings. Adjunct islands, wh-islands, and the complex NP constraint all follow the very same pattern. Subject islands, as a specific subcase of freezing effects, have a slightly higher complexity in that they are either IBSP-2 or IBSP-1. This depends on whether one requires the left and right edge of the k-val to be tied to the same feature f. Since freezing effects are widely considered to be more complex than standard island constraints and depend on the interaction of multiple movements, it is unsurprising that their IBSP complexity should be slightly higher. Nonetheless the IBSP approach with astring provides a unified perspective of several movement restrictions that highlights their computational simplicity and treats them as a natural syntactic counterpart of blocking effects in phonology.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Other Strong Islands", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "The previous section has argued that IBSP grammars over a-strings provide an insightful perspective on movement constraints that highlights their simplicity and their formal parallels to blocking effects in phonology.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "It is also noteworthy just how limited the machinery is. For instance, it is now unsurprising that no language has island constraints such as \"you may move out of as many adjuncts as you have movement features\". This simply cannot be expressed with IBSP-1 or IBSP-0. Similarly, we correctly predict that no language has complex structural conditions like \"an adjunct is an island iff it is c-commanded by another adjunct\". Not only would this require a larger k-val than IBSP-1 and IBSP-0 provide, the use of a-strings makes it completely impossible to refer to c-commanders. By adopting a string representation that only keeps track of containment, c-command conditions become inexpressible. While every Y in as(X) ccommands X, not every c-commander of X appears in as(X) -only those that are heads of phrases containing X do so. The absence of some c-commanders in a-strings thus makes them unsuitable to express c-command conditions.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "The limits of a-strings with respect to ccommand is both a curse and a blessing. As just discussed, it has the advantage of greatly limiting the predicted typology of island constraints. At the same time, it also means that the current approach is entirely incapable of handling some well-known restrictions on movement: the that-trace effect, and the coordinate structure constraint.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Let us first consider the that-trace effect, the core cases of which are illustrated below: 15 The that-trace filter forbids a subject to move across the head of the smallest containing sentential CP if that head is empty. This adds several new complications, but these can all be handled with IBSP. The restriction to subjects amounts to the requirement that the left edge of the k-val must be a mover whose first movement feature is nom , followed by some f . Similarly, the limitation to sentential CPs can be expressed in terms of the MG feature calculus. The complementizer in the examples above has the feature make-up T + C , whereas the complementizer of a relative clause, for instance, would have T + N \u21e0 (under an analysis of relative clauses as NP-adjuncts; other analyses require different features, but it will never be T + C ). So this aspect of the that-trace effect does not challenge the IBSP perspective either. Finally, the requirement that the constraint only applies to the closest such complementizer can be captured by restricting the appropriate filler. Overall, the typical instances of the that-trace constraint can be handled by an IBSP-1 grammar that uses the same truncation trick as our IBSP-1 treatment of freezing effects in (11).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "(16) IBSP-1 grammar for the that-trace effect nom f nom + that", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "\u00acnom + \u00acf + , \u00acC", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "For the core cases, then, the that-trace effect exceeds the strong island constraints in complexity, but is comparable to freezing effects. However, there are cases where that-trace violations are repaired, and these cannot be handled in our approach. For instance, the that-trace effect does not apply when the gap is c-commanded by additional material. Here under no circumstances is an adjunct that attaches to TP or some other position below the complementizer and above the subject gap. This adjunct does not contain the gap, it only ccommands it. As a result, it is not present in the relevant a-strings, which makes it impossible for us to suspend the that-trace constraint. In order to handle this case, one needs a representation that encodes both c-command and containment, e.g. the c-strings of . But the addition of c-command actually undermines the whole approach because it becomes impossible to determine a mover's landing site. Recall that our grammars block f + from occurring in the fillers so that we can correctly pick out the landing site for f-movement, i.e. the closest containing head with f + . Crucially, heads that ccommand the mover but do not contain it are not viable landing sites. For instance, if we are looking at the c-string of some f-mover that is the complement of some head H, the specifier of H may be some lexical item carrying f + . This specifier should be allowed to go into a filler slot. At the same time, a c-commander that both carries f + and contains the f-mover should not be allowed to go into a filler slot. Since fillers are specified as lists of lexical items, there is no way to distinguish in their specification between containing c-commanders and all other c-commanders. Either we allow both in the filler or neither, and in each case we end up with an unsuitable grammar. IBSP is too weak to make the relevant distinctions with a representation format that encodes both ccommand and containment.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "While that-trace repair points out a limitation of IBSP, the coordinate structure constraint challenges the very notion of string-based representations for movement. This constraint forbids extraction from a conjunct, except if movement takes place across-the-board from all conjuncts. Since there is no c-command or containment relation between the gaps in (18b), neither one appears in the other's c-string or a-string. Consequently, the c-strings for the object of drink do not differ at all between the two sentences, which makes it impossible to give a c-string account of this island constraint irrespective of how powerful one's computational apparatus is. These two constraints show that IBSP over astrings does not provide a fully exhaustive theory of islands or movement constraints. But the IBSP approach does highlight the structural uniformity of many islands, their computational simplicity, and their parallels to blocking effects in phonology. While our findings are still preliminary and need to be vetted by detailed analysis of a much wider range of constraints across many languages, it is encouraging that they closely mirror previous findings in phonology and yield rigorous claims about the possible shapes of islands.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Limits of A-Strings", |
| "sec_num": "4" |
| }, |
| { |
| "text": "The previous section focused on some shortcomings of our approach with respect to expressivity, but there is also the issue of linguistic naturalness. First, the choice of string representations is unusual. Second, the reliance on lists of lexical items for specifying the components of an IBSP grammar means that there is no notion of naturalness. We acknowledge both issues, but we think that they can be insightfully addressed in future work.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linguistic Naturalness", |
| "sec_num": "5" |
| }, |
| { |
| "text": "As was briefly mentioned in \u00a72.2, a-strings are just a convenient abstraction and the findings of this paper can be restated in terms of formal machinery that operates over trees instead of strings. This includes the tree tiers of Graf (2018b) and the sensing tree automata of Graf and De Santo (2019) . But in both cases the necessary math is more likely to obfuscate the simplicity of the underlying principles, and the use of tree structures hides that the simple notion of containment is already enough to state many conditions on movement. We thus maintain that a-strings are methodologically useful even if they may not be cognitively real.", |
| "cite_spans": [ |
| { |
| "start": 231, |
| "end": 243, |
| "text": "Graf (2018b)", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 277, |
| "end": 301, |
| "text": "Graf and De Santo (2019)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linguistic Naturalness", |
| "sec_num": "5" |
| }, |
| { |
| "text": "This leaves the lack of natural classes. It is true that our current approach is still too lenient a characterization of the class of possible island constraints. For instance, one can easily write an IBSP grammar over a-strings that does not allow topicalization across a ditransitive verb. Similarly, the ability to account for some exceptions such as (4) also allows us to specify ludicrous exceptions, for instance that the head of an adjunct induces an island unless it is a palindrome. These are clearly undesirable options, but they are typical of computational work. Our primary goal was to analyze island constraints from a subregular perspective to more accurately pinpoint their overall complexity. This allows us to put an upper bound on what island constraint may look like, but this is still a very generous bound. The formal restrictions must be paired with a theory of linguistic substance to ac-curately circumscribe the class of possible island constraints (see e.g. Graf 2013 for one such account for the adjunct island constraint).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Linguistic Naturalness", |
| "sec_num": "5" |
| }, |
| { |
| "text": "We have argued that the most common cases of strong islands can be expressed as IBSP-1 (or IBSP-0) constraints on string representations that encode only containment. This formal characterization establishes new parallels to phonology and tightens the linguistic typology by excluding logically conceivable yet unattested island constraints. While a lot of empirical modeling work remains to be done, we are confident that this novel perspective on islands will prove very fertile.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "6" |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "The work reported in this paper was supported by the National Science Foundation under Grant No. BCS-1845344. We would like to thank the three anonymous reviewers for their detailed comments, which prompted a major rewrite of the paper.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgments", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "On the treatment of scrambling and adjunction in Minimalist grammars", |
| "authors": [ |
| { |
| "first": "Werner", |
| "middle": [], |
| "last": "Frey", |
| "suffix": "" |
| }, |
| { |
| "first": "Hans-Martin", |
| "middle": [], |
| "last": "G\u00e4rtner", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "Proceedings of the Conference on Formal Grammar", |
| "volume": "", |
| "issue": "", |
| "pages": "41--52", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Werner Frey and Hans-Martin G\u00e4rtner. 2002. On the treatment of scrambling and adjunction in Minimal- ist grammars. In Proceedings of the Conference on Formal Grammar, pages 41-52.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "The syntactic algebra of adjuncts", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Proceedings of CLS 49", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf. 2013. The syntactic algebra of adjuncts. In Proceedings of CLS 49. To appear.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "The power of locality domains in phonology", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Phonology", |
| "volume": "34", |
| "issue": "", |
| "pages": "385--405", |
| "other_ids": { |
| "DOI": [ |
| "10.1017/S0952675717000197" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf. 2017. The power of locality domains in phonology. Phonology, 34:385-405.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Locality domains and phonological c-command over strings", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "NELS 48: Proceedings of the Forty-Eighth Annual Meeting of the North East Linguistic Society", |
| "volume": "1", |
| "issue": "", |
| "pages": "257--270", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf. 2018a. Locality domains and phonolog- ical c-command over strings. In NELS 48: Proceed- ings of the Forty-Eighth Annual Meeting of the North East Linguistic Society, volume 1, pages 257-270, Amherst, MA. GLSA.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "Why movement comes for free once you have adjunction", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of CLS 53", |
| "volume": "", |
| "issue": "", |
| "pages": "117--136", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf. 2018b. Why movement comes for free once you have adjunction. In Proceedings of CLS 53, pages 117-136.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Sensing tree automata as a model of syntactic dependencies", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| }, |
| { |
| "first": "Aniello De", |
| "middle": [], |
| "last": "Santo", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the 16th Meeting on the Mathematics of Language", |
| "volume": "", |
| "issue": "", |
| "pages": "12--26", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf and Aniello De Santo. 2019. Sensing tree automata as a model of syntactic dependencies. In Proceedings of the 16th Meeting on the Mathematics of Language, pages 12-26, Toronto, Canada. Asso- ciation for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "C-command dependencies as TSL string constraints", |
| "authors": [ |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| }, |
| { |
| "first": "Nazila", |
| "middle": [], |
| "last": "Shafiei", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the Society for Computation in Linguistics (SCiL) 2019", |
| "volume": "", |
| "issue": "", |
| "pages": "205--215", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Thomas Graf and Nazila Shafiei. 2019. C-command dependencies as TSL string constraints. In Proceed- ings of the Society for Computation in Linguistics (SCiL) 2019, pages 205-215.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "The computational nature of phonological generalizations", |
| "authors": [ |
| { |
| "first": "Jeffrey", |
| "middle": [], |
| "last": "Heinz", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Phonological Typology, Phonetics and Phonology", |
| "volume": "", |
| "issue": "", |
| "pages": "126--195", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jeffrey Heinz. 2018. The computational nature of phonological generalizations. In Larry Hyman and Frank Plank, editors, Phonological Typology, Pho- netics and Phonology, chapter 5, pages 126-195. Mouton De Gruyter.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Counter-Free Automata", |
| "authors": [ |
| { |
| "first": "Robert", |
| "middle": [], |
| "last": "Mcnaughton", |
| "suffix": "" |
| }, |
| { |
| "first": "Seymour", |
| "middle": [], |
| "last": "Papert", |
| "suffix": "" |
| } |
| ], |
| "year": 1971, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Robert McNaughton and Seymour Papert. 1971. Counter-Free Automata. MIT Press, Cambridge, MA.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Syntactic semigroups", |
| "authors": [ |
| { |
| "first": "Jean-Eric", |
| "middle": [], |
| "last": "Pin", |
| "suffix": "" |
| } |
| ], |
| "year": 1997, |
| "venue": "Handbook of Language Theory", |
| "volume": "", |
| "issue": "", |
| "pages": "679--764", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jean-Eric Pin. 1997. Syntactic semigroups. In Hand- book of Language Theory, pages 679-764. Springer, Berlin.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Constraints on Variables in Syntax", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "John", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Ross", |
| "suffix": "" |
| } |
| ], |
| "year": 1967, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "John R. Ross. 1967. Constraints on Variables in Syn- tax. Ph.D. thesis, MIT.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Derivational Minimalism", |
| "authors": [ |
| { |
| "first": "Edward", |
| "middle": [ |
| "P" |
| ], |
| "last": "Stabler", |
| "suffix": "" |
| } |
| ], |
| "year": 1997, |
| "venue": "Logical Aspects of Computational Linguistics", |
| "volume": "1328", |
| "issue": "", |
| "pages": "68--95", |
| "other_ids": { |
| "DOI": [ |
| "10.1007/BFb0052152" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Edward P. Stabler. 1997. Derivational Minimalism. In Christian Retor\u00e9, editor, Logical Aspects of Compu- tational Linguistics, volume 1328 of Lecture Notes in Computer Science, pages 68-95. Springer, Berlin.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Computational perspectives on Minimalism", |
| "authors": [ |
| { |
| "first": "Edward", |
| "middle": [ |
| "P" |
| ], |
| "last": "Stabler", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "Oxford Handbook of Linguistic Minimalism", |
| "volume": "", |
| "issue": "", |
| "pages": "617--643", |
| "other_ids": { |
| "DOI": [ |
| "10.1093/oxfordhb/9780199549368.013.0027" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Edward P. Stabler. 2011. Computational perspectives on Minimalism. In Cedric Boeckx, editor, Oxford Handbook of Linguistic Minimalism, pages 617- 643. Oxford University Press, Oxford.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Extraction from adjuncts and the structure of events", |
| "authors": [ |
| { |
| "first": "Robert", |
| "middle": [], |
| "last": "Truswell", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Lingua", |
| "volume": "117", |
| "issue": "", |
| "pages": "1355--1377", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Robert Truswell. 2007. Extraction from adjuncts and the structure of events. Lingua, 117:1355-1377.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Case assignment in TSL syntax: A case study", |
| "authors": [ |
| { |
| "first": "Mai", |
| "middle": [ |
| "Ha" |
| ], |
| "last": "Vu", |
| "suffix": "" |
| }, |
| { |
| "first": "Nazila", |
| "middle": [], |
| "last": "Shafiei", |
| "suffix": "" |
| }, |
| { |
| "first": "Thomas", |
| "middle": [], |
| "last": "Graf", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the Society for Computation in Linguistics (SCiL) 2019", |
| "volume": "", |
| "issue": "", |
| "pages": "267--276", |
| "other_ids": { |
| "DOI": [ |
| "10.7275/sywz-xw23" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Mai Ha Vu, Nazila Shafiei, and Thomas Graf. 2019. Case assignment in TSL syntax: A case study. In Proceedings of the Society for Computation in Lin- guistics (SCiL) 2019, pages 267-276.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "Contributions to the Theory of Finite-State Based Grammars", |
| "authors": [ |
| { |
| "first": "Anssi", |
| "middle": [], |
| "last": "Yli-Jyr\u00e4", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anssi Yli-Jyr\u00e4. 2005. Contributions to the Theory of Finite-State Based Grammars. Ph.D. thesis, Univer- sity of Helsinki.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "text": "D + D + V Merge which :: N + D wh car :: N did :: T + wh + C \" :: V + nom + T buy :: D + D + V yesterday :: V \u21e0 Mary :: D nom which :: N + D wh car :: N Figure 1: X 0 -tree, MG derivation tree, and equivalent dependency tree for Which car did Mary buy yesterday", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF1": { |
| "text": "did :: C + wh + C \" ::V + nom + T leave :: D + V because :: T + V \u21e0 \" :: V + nom + T talked about :: D + D + VMary :: D nom which :: N + D wh topic :", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF2": { |
| "text": "did :: T + wh + C \" :: V + nom + T hear :: D + D + V you :: D nom \" :: N + D rumors :: C + N that :: T + C \" :: V + nom + T bought :: D + D + V John :: D nom what :: D wh", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF3": { |
| "text": "Violation of the complex NP constraint", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF4": { |
| "text": "did :: T + wh + C \" ::V + nom + T bother :: D + C + V that :: T + nom C \" :: V + nom + T kissed :: D + D + VMary :: D nom who :: D wh you :: D Violation of the subject condition", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF5": { |
| "text": "Who do you think [that [under no circumstances] will leave Mary]?", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| }, |
| "FIGREF6": { |
| "text": "18) a. * Which wine did [Ed brew beer and Greg drink ]? b. Which wine did [Ed brew and Greg drink ]?", |
| "type_str": "figure", |
| "num": null, |
| "uris": null |
| } |
| } |
| } |
| } |
