Vijay-Shanker, K. (1987). A Study of Tree Adjoining Grammar. PhD thesis, University of Pennsylvania, Philadelphia.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is heavy or light. ffl In each substitution and foot node a specific Theta role feature indicates the semantic relation which is realized when the corresponding operation is performed. 3. 1 How to build Shared bubbles We suppose here that we use the definition of LTAG derivation proposed by [Vijay Shanker, 1987] (no multiple adjunction on the same node) As explained in [Kallmeyer and Joshi, 1999] the 32 Kim Gerdes Patrice Lopez Tbingen from professor the former persuades examine Sandy theme John patient agent agent j agent i patient sincerely Figure 8: Shared bubble graph ....

....dependency representations for LTAG 33 (7) John persuades the doctor to examine Sandy. 8) John believes the doctor to examine Sandy. Semantic dependency relations which are not localized in the tree structure of elementary trees are obtained with the classical two steps unification processing [Vijay Shanker, 1987]. The cost of this mechanism is important since it supposes percolation and consequently values of variables which depend on the current derivation (which is possibly exponential in n being the length of the string to parse) But we argue that all phenomena that could not be localized in the ....

Vijay-Shanker, K. (1987). A Study of Tree Adjoining Grammar. PhD thesis, University of Pennsylvania, Philadelphia.

....seems to like the raising construction seems adjoins into the tree for like between the subject and the verb, as shown in Figure 2. Then the bridge construction Mary thinks adjoins onto the initial tree on the other 3 This is a simplified formulation of the Feature based TAGs defined in (Vijay Shanker, 1987), which are used as standard in the world of TAGs; see e.g. XTAG, 1998) 36 I I VP V eat DP e to V VP IP DP C CP DP C that I I t IP DP I I to V eat DP e VP CP C DP what C that a b[seem] eat] I t I V VP I b[seem] a[eat] seems seems what Figure 2: a[eat] IP C DP C that CP ....

Vijay-Shanker, K. 1987. A study of tree adjoining grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

....w 2 v 4 j n, jv 0 1 w 0 1 v 0 2 v 0 3 w 0 2 v 0 4 j n, such that for all i 0, hu 1 v i 1 w 1 v i 2 u 2 v i 3 w 2 v i 4 u 3 ; u 0 1 v 0i 1 w 0 1 v 0i 2 u 0 2 v 0i 3 w 0 2 v 0i 4 u 0 3 i 2 L. The proof is similar to that of the normal pumping lemma for TALs (Vijay Shanker, 1987). The intuition is that the pumping lemma for local sets is applied to the derivation trees, and since paired derivation trees are isomorphic, the pumping constant can be chosen so that the pumping lemma holds for both sides simultaneously. Claim. L = fha i 1 j 2 j b i c i 3 j 4 j d ....

Vijay-Shanker K. (1987). A study of tree adjoining grammars. PhD thesis, Department of Computer and Information Science, University of Pennsylvania.

....must be made between the (object level) trees that are derived in a derivation and the (meta level) trees that are used to fully encode what happens in derivations. These trees are referred to as derived and derivation trees, respectively. A scheme for encoding LTAG derivations was proposed by Vijay Shanker (1987) and later modified by Schabes and Shieber (1994) Derivation trees show, in a very direct way, how the elementary trees are combined in derivations. Nodes of the derivation trees are labeled by the names of elementary trees, and edge labels identify tree addresses (i.e. node locations) in ....

....of substitution and adjunction nodes. The tree addresses can be omitted from derivation trees if a fixed linear order is established on all of the adjunction and substitution nodes in each elementary tree and this ordering is used to order siblings in the derivation tree. Given this possibility, Vijay Shanker at al. 1987) have shown that the set of derivation trees associated with a LTAG forms a local set and can therefore be generated by a context free grammar (CFG) 5 . The productions of this meta grammar encode possible derivation steps of the grammar. In other words, each meta production encodes one way of ....

Vijay-Shanker, K. (1987). A Study of Tree Adjoining Grammars. PhD thesis, Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA.

....nodes in trees can be uniquely identified by stating the tree and their address within the tree. 42 A derivation consists in choosing an initial tree, and then adjoining and substituting trees into it or each other. We do not give a formal definition for the derivation here, but refer to (VijayShanker, 1987, p.16) The set of derived trees of a TAG G, T (G) consists in all trees derived from I and A by substitution and adjunction such that the frontier, read off as a string in (V N [ VT ) contains only terminal symbols. The string language derived by G, L(G) is the set of all strings in V T on ....

....the difference between a top down and a bottom up PDA only lies in the way the particular automaton is specified (given a CFG) not in how the formal automaton is defined qua formal system. This is different in the case of the embedded pushdown automaton (EPDA) an automaton introduced by Vijay Shanker (1987) that is equivalent to TAG. EPDA simulates a top down derivation, and we cannot use this automaton to simulate a bottom up derivation. The problem is related to the simulation of adjunction. We therefore need to define a new automaton, called the bottom up EPDA, or BEPDA. The BEPDA was first ....

[Article contains additional citation context not shown here]

Vijay-Shanker, K. (1987). A study of Tree Adjoining Grammars. PhD thesis, Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA.

....Kahane, 1998] This paper focuses on robust chart parsing with Lexicalized Tree Grammars. We do not consider here probabilistic approximation but only complete or partial structures obtained with valid rule applications. The classical parsing algorithms for LTAG, for instance CKY like algorithm [Vijay Shanker, 1987], Head Corner [van Noord, 1994] or Earley like algorithm [Schabes, 1994] focus on the parsing of complete grammatical utterances. We argue that an algorithm dedicated to the parsing of a Lexicalized Tree Grammar can take into account more eciently EDL and lexicalization in order to (1) obtain bene ....

....the topology of the trees to decrease the average complexity with a new level of granularity for a linearized tree called connected route. 2. 3 Finite States Automata representation of an elementary tree The existing representations of the parsing process of a elementary tree are dotted tree [Vijay Shanker, 1987] and dotted rules [Nederhof, 1997] In both case this representation is constrained by their locality. We propose an alternative representation that allows to express contraints of signi cant parts of the tree. The linearization of a tree can be represented with a Finite State Automaton (FSA) as ....

Vijay-Shanker, K. (1987). A Study of Tree Adjoining Grammar. PhD thesis, University of Pennsylvania, Philadelphia.

....Indexed Languages ae CSL ffl TALs are semilinear; ffl All closure properties of context free languages also hold for tree adjoining languages. In fact, TALs are a full abstract family of languages (full AFLs) ffl a variant of the push down automaton called embedded push down automaton (EPDA) [Vijay Shanker1987] characterizes exactly the set of tree adjoining languages, just as push down automaton characterizes CFLs. ffl there is a pumping lemma for tree adjoining languages. ffl tree adjoining languages can be parsed in polynomial time, in the worst case in O(n 6 ) time. Some well know properties of ....

....affirmative in [Schabes and Waters1995] The tree insertion grammar they define 17 There are several important papers about TAGs describing their linguistic, computational and formal properties. Some of these are: Joshi [Joshi1987] Joshi, Vijay Shanker and Weir [Joshi et al..1991] Vijay Shanker [Vijay Shanker1987], Weir [Weir1988] Schabes [Schabes1990, Schabes1991] Schabes and Joshi [Schabes and Joshi1988, Schabes and Joshi1989] Kroch [Kroch1987] Kroch and Joshi [Kroch and Joshi1985] Abeill e, Bishop, Cote and Schabes [Abeill e et al..1990] Abeill e [Abeill e1988] Schabes and Waters [Schabes and ....

[Article contains additional citation context not shown here]

K. Vijay-Shanker. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

.... Grammars (TAGs) are attractive because they can express some natural language phenomena (see Abeill e and Schabes [1] and many systems are based upon this framework (see for example [11] and [7] Formal properties of TAGs have been studied (see Vijay Shanker and Joshi [17] and Vijay Shanker [16]) and a recognizer for TAGs (see [19] based upon a Cocke Kasami Younger method ( 5] and [20] works in O(n 6 ) worst time. Unfortunately, with such algorithms, this complexity is always reached. More practical methods, which are usually based upon the Earley parsing algorithm [4] have also ....

Vijay-Shanker, K. 1987. A study of tree adjoining grammars. PhD. thesis, University of Pennsylvania.

....comfortable way of assessing the formal complexity of linguistic constructions is based on extensions of the well known pumping lemma for context free languages. The following version of a pumping property can be proved for the class of languages recognized by linear context free rewriting systems [VS87] [SMFK91] discussed in section 2.2) Definition 1 (k pumpability) Let L be a language. Then L is (universally) k pumpable if there are constants c 0 ; k such that for any w 2 L with jwj c 0 , there are strings u 0 ; u k and v 1 ; v k such that w = u 0 v 1 u 1 v 2 u 2 Delta ....

....2 Delta Delta Delta u k Gamma1 v k u k , for each i: 1 jv i j c 0 , and for any p 0, u 0 v p 1 u 1 v p 2 u 2 Delta Delta Delta u k Gamma1 v p k u k 2 L. Regular languages are 1 pumpable; Context free languages are 2 pumpable [HU79] tree adjoining or head languages are 4 pumpable [VS87]. For every language recognized by an LCFRS, there is a k such that it is k pumpable [SMFK91] So if a language can be shown not to be 2 pumpable, then it is not a CFL; if it is not 4 pumpable, it is not a TAL, and so forth. A fragment containing cross serial sentences such as (1.9) sec is ....

K. Vijay-Shanker. A Study of Tree Adjoining Grammars. PhD thesis, Univ. of Pennsylvania, 1987.

....Even for non cyclic grammars, the number of parse trees can be exponential in the size of the input. Moreover, it is problematic that a worst case polynomial size structure could be reached by some sharing compatible both with the syntactic and the isemanticj features. However, we know (see (Vijay Shanker, 1987)) that derivations in TAGs are context free and (Vijay Shanker and Weir, 1993) exhibits a CFG which represents all possible derivation sequences in a TAG. We will show that the analogous holds for LIGs and leads to an O(n 6 ) time parsing algorithm. Denition 7 Let L = V N ; V T ; V I ; PL ; ....

K. Vijay-Shanker. 1987. A study of tree adjoining grammars. PhD thesis, University of Pennsylvania.

....The Greek letters ## # and # are used to denote nodes in elementary trees. Greek letters # and # are used to denote nodes in trees from the set C. We assume that multiple adjunctions on a single node are allowed #Schabes and Shieber# 1994#. This is a modi#cation from the standard TAG derivation #Vijay#Shanker# 1987# where it was disallowed. C D S S B A b Figure 22# An auxiliary tree with a contraction set A layer of an elementary tree is represented textually in a style similar to a production rule# e.g. #X # # Y #Z . For instance# the tree in Figure 22 is represented by the production rules in ....

Vijay#Shanker# K. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis# Department of Computer and Information Science# University of Pennsylvania.

....modi er is heavy or light. In each substitution and foot node a speci c role feature indicates the semantic relation which is realized when the corresponding operation is performed. 3. 1 How to build Shared bubbles We suppose here that we use the de nition of LTAG derivation proposed by [Vijay Shanker, 1987] (no multiple adjunction on the same node) As explained in [Kallmeyer and Joshi, 1999] the 7 Tbingen from professor the former persuades examine Sandy theme John patient agent agent j agent i patient sincerely Figure 8: Shared bubble graph representation for the example 6. de ....

....(7) here object control verbs) 8 (7) John persuades the doctor to examine Sandy. 8) John believes the doctor to examine Sandy. Semantic dependency relations which are not localized in the tree structure of elementary trees are obtained with the classical two steps uni cation processing [Vijay Shanker, 1987]. The cost of this mechanism is important since it supposes percolation and consequently values of variables which depend on the current derivation (which is possibly exponential in n being the length of the string to parse) But we argue that all phenomena that could not be localized in the ....

Vijay-Shanker, K. (1987). A Study of Tree Adjoining Grammar. PhD thesis, University of Pennsylvania, Philadelphia. 13

....The Greek letters , ae and are used to denote nodes in elementary trees. Greek letters and are used to denote nodes in trees from the set C. We assume that multiple adjunctions on a single node are allowed (Schabes and Shieber, 1994) This is a modification from the standard TAG derivation (Vijay Shanker, 1987) where it was disallowed. C D S S B A b Figure 22: An auxiliary tree with a contraction set A layer of an elementary tree is represented textually in a style similar to a production rule, e.g. X Y ae Z . For instance, the tree in Figure 22 is represented by the production rules in (1) ....

Vijay-Shanker, K. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

....which could adjoin into the tree for the book to give the syntax book. Our grammar incorporates two additional principles. First, the grammar is LEXICALIZED (Schabes, 1990) each elementary structure in the grammar contains at least one lexical item. Second, our trees include FEATURES, following (Vijay Shanker, 1987). LTAG elementary trees abstract the combinatorial properties of words in a linguistically appealing way. All predicate argument structures are localized within a single elementary tree, even in long distance relationships, so elementary trees give a natural domain of locality over which to state ....

K. Vijay-Shanker. 1987. A Study of Tree Adjoining Grammars.

.... domain of locality combined with substitution and adjunction yield Lexicalized TAGs (LTAGs) TAGs were first introduced by Joshi, Levy and Takahashi (1975) and Joshi (1985) For more details on the original definition of TAGs, we refer the reader to Joshi (1985) Kroch and Joshi (1985) or Vijay Shanker (1987). It is known that Tree Adjoining Languages (TALs) are mildly context sensitive. TALs properly contain context free languages. TAGs with substitution and adjunction are naturally lexicalized because they use an extended domain of locality. 6 A Lexicalized Tree Adjoining Grammar is a tree based ....

....them as arguments (0 for subject, 1 for first object, etc. 9 Optional arguments are stated as such in the structure. 1 Introduction to Lexicalized Tree Adjoining Grammars 6 1.3 Feature structures in TAGs The Lexicalized TAG we are using is a feature structure based formalism. As defined by Vijay Shanker (1987) and Vijay Shanker and Joshi (1988) to each adjunction node in an elementary tree two feature structures are attached: a top and a bottom. 10 When the derivation is completed, the top and bottom features of all nodes are unified simultaneously. If the top and bottom features of a node do not ....

K. Vijay-Shanker. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

.... domain of locality combined with substitution and adjunction yield Lexicalized TAGs (LTAGs) TAGs were first introduced by Joshi, Levy and Takahashi (1975) and Joshi (1985) For more details on the original definition of TAGs, we refer the reader to Joshi (1985) Kroch and Joshi (1985) or Vijay Shanker (1987). It is known that Tree Adjoining Languages (TALs) are mildly context sensitive. TALs properly contain context free languages. TAGs with substitution and adjunction are naturally lexicalized because they use an extended domain of locality. 6 A Lexicalized Tree Adjoining Grammar is a tree based ....

....them as arguments (0 for subject, 1 for first object, etc. 9 Optional arguments are stated as such in the structure. 1 Introduction to Lexicalized Tree Adjoining Grammars 6 1.3 Feature structures in TAGs The Lexicalized TAG we are using is a feature structure based formalism. As defined by Vijay Shanker (1987) and Vijay Shanker and Joshi (1988) to each adjunction node in an elementary tree two feature structures are attached: a top and a bottom. 10 When the derivation is completed, the top and bottom features of all nodes are unified simultaneously. If the top and bottom features of a node do not ....

K. Vijay-Shanker. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

....result is a single derived tree. Of interest here is set local MCTAG, which because of the independent parallelism is more powerful than standard TAG. For example, the MCTAG in Figure 3 generates the language count 8, fa n 1 a n 2 : a n 8 j n 0g, which, by the Pumping Lemma for TALs (Vijay Shanker, 1987: 96) is beyond the generative capacity of standard TAGs. In addition, unlike for standard TAGs, the path sets of the derived trees may not be context free, and the paths are not independent. The derivation trees for MCTAGs are similar to those for TAGs: they are similarly context free, this ....

....Figure 12. It will be clear that, if separate occurrences of fi are adjoined into ff, the right projection of this S TAG generates the string language fa n b n c n d n a n b n c n d n j n 0g, which is not a TAL; this can be confirmed by application of the Pumping Lemma for TALs (Vijay Shanker, 1987: 96) which states that TALs must be of the form u 1 v i 1 w 1 v i 2 u 2 v i 3 w 2 v i 4 u 3 , with u i ; v i ; w i 2 V , with V the alphabet of terminal and non terminal symbols. ff: A 1 ffl S A 1 ffl . k times . A 1 ffl AE fi 1 : A 1 ANA a A 1 A NA a AE fi ....

Vijay-Shanker, K. 1987. A Study of Tree Adjoining Grammars. PhD thesis, University of Pennsylvania.

....along with closure under Boolean operations gives decidability of emptiness of intersection. This gives non definability of the context free tree languages of Rounds 1970. It also gives non definability of the set of trees generated by Tree Adjoining Grammars (TAGs, Joshi et al. 1975, Joshi 1987, Vijay Shanker 1987) in which nodes are allowed to be re labeled during adjunction. In both these cases, non decidability of emptiness of intersection comes from the ability to define, for any context free string language, a set of linear trees in which the labeling of the (only) branch of each tree in the set is a ....

Vijay-Shanker, K. 1987. A Study of Tree Adjoining Grammars. Doctoral dissertation, Department of Computer and Information Science, University of Pennsylvania.

....no need to distinguish between derivation trees and derived trees. This is not always the case. In particular, for a tree rewriting system like tag we need to distinguish between derived and derivation trees. In fact there are at least two ways one can encode tag derivation trees. The first (see [ Vijay Shanker, 1987 ] captures the fact that derivations in tag are context free, i.e. the trees that can be adjoined at a node can be determined a priori and are not dependent on the derivation history. We capture this context freeness by giving a cfg to represent the set of all possible derivation sequences in ....

K. Vijay-Shanker. A Study of Tree Adjoining Grammars. PhD thesis, University of Pennsylvania, Philadelphia, PA, 1987.

No context found.

Vijay-Shanker K : 1987. A study of Tree Adjoining Grammars. PhD. Univ. of Pennsylvania. Philadelphia.

No context found.

K. Vijay#Shanker. A Study of Tree Adjoining Grammars. PhD thesis# Department of Computer and Information Science# University of Pennsylvania# 1987.

No context found.

Vijay-Shanker, K. (1987). A study of Tree Adjoining Grammars. PhD Thesis, University of Pennsylvania.

No context found.

K. Vijay--Shanker. 1987. A Study of Tree Adjoining Grammars. Ph.D. Thesis, Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA, USA.

No context found.

K. Vijay-Shanker. A study of tree adjoining grammars. PhD thesis, University of Pennsylvania, Philadelphia, PA, December 1987.

No context found.

Control, 25:55--74. K. Vijay-Shanker. 1987. A Study of Tree Adjoining Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC