Showing papers on "Tree-adjoining grammar published in 1985"

Natural language parsing: Tree adjoining grammars: How much context-sensitivity is required to provide reasonable structural descriptions?

Abstract: Since the late 1970s there has been vigorous activity in constructing highly constrained grammatical systems by eliminating the transformational component either totally or partially. There is increasing recognition of the fact that the entire range of dependencies that transformational grammars in their various incarnations have tried to account for can be captured satisfactorily by classes of rules that are nontransformational and at the same time highly constrained in terms of the classes of grammars and languages they define. Two types of dependencies are especially important: subcategorization and filler-gap dependencies. Moreover, these dependencies can be unbounded. One of the motivations for transformations was to account for unbounded dependencies. The so-called nontransformational grammars account for the unbounded dependencies in different ways. In a tree adjoining grammar (TAG) unboundedness is achieved by factoring the dependencies and recursion in a novel and linguistically interesting manner. All dependencies are defined on a finite set of basic structures (trees), which are bounded. Unboundedness is then a corollary of a particular composition operation called adjoining . There are thus no unbounded dependencies in a sense. This factoring of recursion and dependencies is in contrast to transformational grammars (TG), where recursion is defined in the base and the transformations essentially carry out the checking of the dependencies. The phrase linking grammars (PLGs) (Peters and Ritchie, 1982) and the lexical functional grammars (LFGs) (Kaplan and Bresnan, 1983) share this aspect of TGs; that is, recursion builds up a set a structures, some of which are then filtered out by transformations in a TG, by the constraints on linking in a PLG, and by the constraints introduced via the functional structures in an LFG.

Some computational properties of tree adjoining grammers

Abstract: Tree Adjoining Grammar (TAG) is a formalism for natural language grammars. Some of the basic notions of TAG's were introduced in [Joshi, Levy, and Takahashi 1975] and by [Joshi, 1983]. A detailed investigation of the linguistic relevance of TAG's has been carried out in [Kroch and Joshi, 1985]. In this paper, we will describe some new results for TAG's, especially in the following areas: (1) parsing complexity of TAG's, (2) some closure results for TAG's, and (3) the relationship to Head grammars.

The complexity of equivalence problems for commutative grammars

Abstract: In this paper we investigate the computational complexity of the inequivalence problems for commutative grammars. We show that the inequivalence problems for type 0 and context-sensitive commutative grammars are undecidable whereas decidability in nondeterministic exponential-time holds for the classes of regular and context-free commutative grammars. For the latter the inequivalence problems are Σp2-hard.

Boundary NLC Graph Grammars Basic Definitions, Normal Forms and Complexity ; CU-CS-293-85

Abstract: Node label controlled (NLC) grammars are graph grammars (operating on node labeled undirected graphs) which rewrite single nodes only and establish connections between the embedded graph and the neighbors of the rewritten node on the basis of the labels of the involved nodes only. They define (possibly infinite) languages of undirected node labeled graphs (or, if we just omit the labels, languages of unlabeled graphs). Here we consider a restriction of NLC grammars, so-called boundary NLC (BNLC) grammars , distinguished by the property that whenever in a graph already generated two nodes may be rewritten, then these nodes are not adjacent. The graph languages generated by this type of grammars are called BNLC languages . Although we show that this restriction leads to a smaller class of languages, still enough generative power remains to define interesting graph languages. For example, trees, complete bipartite graphs, maximal outerplanar graphs, k -trees, graphs of bandwidth ⩽ k , graphs of cyclic bandwidth ⩽ k , graphs of binary tree bandwidth ⩽ k , graphs of cutwidth ⩽ k (always for a fixed positive integer k ) turn out all to be BNLC languages. We prove a number of normal forms for BNLC grammars and then we indicate their usefulness by various applications. In particular, we show that for connected graphs of bounded degree the membership problem for BNLC languages is solvable in deterministic polynomial time.

On permutative grammars generating context-free languages

Abstract: A grammar is said to be permutative if it has permutation productions of the formAB ρBA in addition to context-free productions. Szilard languages and label languages are studied as examples of languages generable by permutative grammars. Particularly, sufficient conditions for a permutative grammar to generate a context-free language are studied.

Terminal weighted grammars and picture description

Abstract: Motivated by the idea of describing parquet deformations using grammars and also of describing an infinite number of terminals starting with only a finite set, this paper defines a terminal weighted grammar, where the terminal generated at any step of a derivation is defined as a function of time It is seen that terminal weighted regular grammars generate exactly the class of recursively enumerable sets Terminal weighted matrix grammars are used to describe parquet deformations The extension of terminal weights to array grammars is also discussed

Context-Free Grammars and Random Number Generation

Abstract: In Monte Carlo calculations, one often needs to generate a random quantity X that satisfies certain (cumulative) distribution function F(x), i.e. Pr{X ≤ x} = F(x). Numerous methods have been proposed for this purpose (see Ahrens and Dieter [1], Knuth [4]). An interesting question is: for a given F(x), how difficult is it to generate this distribution?

On some constructions of grammars for linear languages

Abstract: Two constructions of generalized grammars are investigated and families of languages are studied for which these constructions provide grammars. One of these families coincides with the family of linear languages, another includes the family of linear deterministic languages and the family of contextual languages.

Unrestricted gapping grammars

Abstract: Since the introduction of metamorphosis grammars (MGs) 'Colmerauer. 1978). with their associated type O-like grammar rules, there has been a desire to allow more general rule formats in logic grammars Gaps, which refer to strings of unspecified symbols, were added to the MG rule, resulting in extraposition grammars (XGs) (Pereira, 1981) and gapping grammars (GGs) (Dahl and Abramson, 1984). Unrestricted gapping grammars, which provide an even more general rule format, possess rules of the form "a - > B" where a and /3 many contain any number of terminal nonterminal or gap symbols in any order. FIGG. a Flexible Implementation of Gapping Grammars, is an implementation of a large subset of unrestricted GGs which allows either bottom-up or top-down parsing of sentences. This system provides more built in control facilities than previous logic grammar implementations, which allows the user to restrict the applicability of the rules, and to create grammar rules that will be executed more efficiently

String grammars with disconnecting

Abstract: The complexity of languages generated by context-free grammars with disconnecting is investigated. It is shown that even linear and deterministic context-free languages can generate languages of multisets, the membership problem of which is NP-complete. In contrast to that, this problem is in DSPACE(log n) for regular sets.

Fairness in Context-Free Grammars under Canonical Derivations

Abstract: In [PFMZ 82] the notion of Fair derivations in context free grammars was introduced and studied The main result there is a characterization of fairly terminating grammars as non-variable-doubling In this paper we show that the same characterization is valid under canonical derivations in which the next variable to be expanded is deterministically chosen, leaving nondeterminism only to the decision as to which rule to apply Two families of canonical derivations are introduced and studied: 1) Spinal derivations and 2) Layered derivations

On the Active and Full Use of Memory in Right-Boundary Grammars and Push-Down Automata ; CU-CS-311-85

Abstract: Abstract A coordinated pair system ( cp system for short) consists of a pair of grammars, the first of which is right-linear ( rl ) and the second is right-boundary ( rb ). A right-boundary grammar is like a right-linear grammar except that one does not distinguish between terminal and nonterminal symbols—still, the rewriting is applied to the last symbol of a string only (and erasing productions are allowed). A rewriting in a cp system consists of a pair of rewritings: one in the first and one in the second grammar—such a rewriting is possible if the pair of productions involved is in the finite set of rewrites given with the system. Is is easily seen that cp systems correspond very closely to (are another formulation of) push-down automata: the right-linear component models the input and the finite state control while the rb component models the push-down store. An rb grammar G transforms (rewrites) strings which are stored in a one-way (potentially infinite) tape. If one observes during a derivation δ the use of a fixed n th cell of the tape and one notes the symbol stored there, each time that (the contents of) the cell is rewritten, then one gets the n-active record of δ; the set of all n -active records for all successful derivations δ forms the n -active language of G , denoted ACT n ( G ). It is proved that, for each rb grammar G and each n ϵ N + , ACT n ( G ) is regular and moreover, for each M ⊆ N + , ∪ nϵM ACT n ( G ) is regular. Another way to register the use of memory during a derivation δ is to record the contents of (a fixed) n th cell during all consecutive steps of δ—in this way one gets the n-full record of δ. The set of all n -full records for all successful derivations δ forms the n -full record language of G , denoted FR n ( G ). It is proved that, as in the case of active records, ∪ nϵM FR n ( G ) does not have to be regular even if M = N + (actually, one can get arbitrarily complex languages in this way). Then we provide a representation theorem allowing one to represent a cp system by an rb grammar and using this theorem we transfer the above results on the use of memory to cp systems.

On coordinated selective towards a unified theory and machines substitutions: of grammars

Abstract: The notion of a coordinated table selective substitution system (acts system) is introduced. It provides a unifying framework for both grammars and machines (automata) and hence a really broad framework for formal language theory. An extensive number of examples is given which illustrate how a quite considerable number of grammars and automata considered in the literature may be 'naturally' interpreted as special instances (subclasses of the class) of cts systems.

A note on one-pass evaluation of attribute grammars

Abstract: An attribute grammar is considered to be one-pass if all the attribute instances of any derivation tree can be evaluated by a process that traverses the tree from left to right visiting each subtree at most once. It is shown that this general class of one-pass grammars properly includesL-attributed grammars; in fact,L-attributed grammars can be viewed as a practical subset of the general class. General one-pass grammars can be evaluated using recursive procedures in the same way as forL-attributed grammars, provided that call-by-name parameters are available.