Shieber, S.: Using restriction to extend parsing algorithms for complex feature-based formalisms. In: Proceedings of the 23rd Meeting of the Association for Computational Linguistics, Chicago, IL (1985) 145 -- 152  Home/Search   Document Details and Download   Summary   Related Articles   Check

....considerably. For the current version of the grammar, it hardly makes a di#erences whether we use the left corner variant or the head corner variant. 12 Goal Weakening. Another important technique which is applied for e# ciency reasons is goal weakening (or abstraction [11] or restriction [24]) Although top down filtering in a bottom up parser is crucial for e#ciency, it can also have an undesirable e#ect: the same parse goal might have to be re computed simply because the top down expectations are slightly di#erent. For instance, a rule might have been applied which expects an ....

Stuart M. Shieber. Using restriction to extend parsing algorithms for complex-feature-based formalisms. In 23th Annual Meeting of the Association for Computational Linguistics, Chicago, 1985.

....rule schemata and the lexical categories. Applying standard parsing algorithms to such grammars is unsatisfactory for a number of reasons. Earley parsing is intractable in general, as the rule set. is simply too general. For some grammars, naive top down prediction may even fail to termi nate. [Shieber, 1985] therefore proposes a modified version of the Earley parser, using restricted topdown prediction. While this modification leads to termination of the prediction step, in practice it easily leads to a trivial top down prediction step, thus leading to inferior performance. Bottom up parsing is far ....

Stuart M. Shieber. Using restric- tion to extend parsing algorithms for complexfeature -based formalisms. In 2$th Annual Meeting of the Association for Computational inguistics, Chicago, 1985.

....do not distinguish between thern and refer to both by the term functional grammar . Application of the functional grammar [ormalisrn to machine translation is discussed in [19J. Attempts have being made to impk;rnent parsers using these grammars, most notably in the PATR II project at Stanford [22, 24], However, these efforts have not been integrated with external semantic knowledge bases, and have not been applied in the context of KBMT systems, There are two main advantages of using the functional grammar formalism in practical ma. chine translation systems: A system implemented strictly ....

....therefore need not be trained (with some justifiable resistance) to write grammars in arcane system specific formalisms, The general problem in parsing with functional gramrnars is implementation inefficiency for any practical application. Although much work bas been done to enhance efficiency [24, 22], the functiomtl grtrnmar formalisms are considered far less efficient than fro realisms like ATNs[28] or (especially) coutexbfree phrase struclure grammars. We resolve this efficiency problem by precompiling a grammars written in a the funr. tiorm grammar (together with a sepa. rate dumain ....

Shieber, S. M., "Using Restriction to Extend Parsing Algorithms for Complex-Feature-Based Formalisms," 23rd Annual Meeting of the Association for Computational Linguistics, Chicago, July 1985, pp. 145-152.

....we will show how we can combine this technique with a clever indexing mechanism that is used for parsing and generation) Use of a restrictor The prediction rule is used for predicting new instantiations of grammar rules using the selected element of a non unit lemma. As known from the work of [Shieber, 1985] prediction can lead to arbitrary numbers of consequents through repeated application when used with a grammar with an infinite structured nonterminal domain. For example, if a grammar handles subcategorization with list value features (such as in HPSG or the rule (rl) of the grammar given in ....

....the constant function yielding the trivial model, no top down information is used. In Shiebers original work, a grammar oriented version of a restriction function is presented, where only those features are predicted which have been chosen by the user as most useful for prediction. Following [Shieber, 1985] restriction is defined on the basis of a given restrictor R, which is a finite set of paths through a feature structure. The restriction of a feature structure F relative to R is the most specific feature structure F F, such that every path in F has either an atomic value or is an element of ....

[Article contains additional citation context not shown here]

S. M. Shieber. Using restriction to extend parsing algorithms for complex-feature-based formalisms. In 23th Annual Meeting of the Association for Computational Linguistics, Chicago, 1985.

....meaningfully to hpsg signs, two conditions must be met. Firstly, parse tree construction cannot be duplicated in the feature structures (by means of the hpsg DTRS feature) but is left to the parser (i.e. recorded in the chart) this is standardly achieved by feature structure restriction (see Shieber, 1985, and Kiefer et al. 1999) Secondly, semantic composition (performed via the hpsg attribute CONT) must be delayed until a complete analysis is derived; in general, it seems desirable to postpone processing of constraints that build up new structure but do not restrict the search space. 13 ....

Shieber, S. M. (1985). Using restriction to extend parsing algorithms for complex feature-based formalisms. In Proceedings of the 23rd Meeting of the Association for Computational Linguistics (pp. 145 - 152). Chicago, IL.

....through head arcs. However, this scheme does not seem to solve the problem of nonminimal derivations completely. 8 Another prominent issue concerning top down prossesing with unification grammars is the Prediction nontermination. There have been several solutions proposed for this problem (e.g. Shieber, 1985; Haas, 1989; Samuelsson, 1993) 1992) van Noord, 1997) and it may be applicable in restricting the nonminimal derivations in implementation. However, this scheme requires more control aspect of the parsing algorithm in the logical specification (i.e. proving the absense of more general ....

Shieber, S. (1985). Using restriction to extend parsing algorithms for complex-feature-based formalisms.

....can 9 A prediction is created after the filtering function ae is applied by the predictor. be characterized as a version of this abstract algorithm in which the most extreme prediction component is used, one in which all possible information is included in the predictions. Top down Filtering Shieber (1985) shows how Earley s algorithm can be extended to unification based grammars, and the extended algorithm in effect gives a greater power in performing top down filtering. He proposes restriction, a function which selects a set of features by which topdown prediction is propagated. By defining the ....

Shieber, S. (1985). Using restriction to extend parsing algorithms for complex-feature-based formalisms.

....if a path 0 pre xes and 0 6= then the type of the 0 value is , the most generic type in all the types. And as shown in part (b) the operation F= produces a feature structure which is the same as F except that the value of F is unde ned. This operation is called restriction (Shieber, 1985). 2.2 Grammars We denote an hpsg based grammar by a triple hRS ; LEX ; Rooti. RS is a nite set of rule schemata, each of which is a feature structure. LEX is a lexicon such that LEX F where is a nite set of all the words in that lexicon. If hw; Li 2 LEX , we say that L is a lexical ....

....may obtain in nitely many distinct variations of feature structures in the state generation. But we can always eliminate such problematic feature values from the grammar and obtain a grammar that produces nite lexical entry automata. This feature elimination is the same as the idea of restriction (Shieber, 1985). In addition, the size of lexical entry automata can be reduced in the same way so that the automaton generation procedure can be completed in a reasonable amount of time. For ltering impossible parse trees, such a transformed grammar is sucient and we actually applied such techniques in the ....

Shieber, S. M. (1985). Using restriction to extend parsing algorithms for complex feature-based formalisms. In Proceedings of the 23rd Meeting of the Association for Computational Linguistics (pp. 145 - 152). Chicago, IL.

.... or the linguistic formalism have been considered (tri gram models for the stochastic modelling of the syntactic rules (Maruyama, 1990) optimized first order dependence models (Wright et al. 1992) Spoken Language Constraint Networks corresponding to extensions of Constraint Dependency Grammars (Shieber, 1985; Kita et al. 1990) more complex syntactic formalisms, such as unification based grammars) 3. Lattice Parsing 3.1. Adaptation of the Parser Input For both tight and sequential coupling, lattice parsing (i.e. the parsing of a whole set of sequences compactly represented in the form of a ....

SHIEBER S. M. (1985). Using restriction to extend parsing algorithms for complex feature-based formalisms. In Proc. of ACL'85, p. 145--152.

....prediction deduces a new item on the basis of an instantiated rule, e.g, using the selected element of the start item 1 (see figure 3, Appendix A) the new tasks 1 and 2 are created and added to the agenda. 8 No item set will be created because I sPmM0 already exists. As known from the work of [ Shieber, 1985 ] prediction can lead to arbitrary numbers of consequents through repeated application when used with a grammar with an infinite structured nonterminal domain. In order to avoid such problems, prediction should be performed with an abstraction of the selected element s constraints (which is ....

....able to perform prediction with as many constraints as possible from the selected element. In parsing literature abstraction has been introduced under the term restriction . For more and detailed information on the definition and use of an abstraction restriction function during parsing see e.g. [Shieber, 1985] , Haas, 1989] and [Samuelsson, 1994] 17 else make new passive item: N i = hRl; ffl; head(Rl) Ef)i fi; create item set I index(Ni) if it does not exist; add task to agenda(N i,prio(N i) Agenda) od fi. Thus, if a lexical entry can be unified with the selected element of the active ....

S. M. Shieber. Using restriction to extend parsing algorithms for complexfeature -based formalisms. In 23th Annual Meeting of the Association for Computational Linguistics, Chicago, 1985.

....with a one pass beam decoder are implemented. ffl Receiving the decoder s word ending hypotheses frame by frame, a mapping of hypotheses belonging to the same word copy is performed during parsing. 1 The ideas which led to this work are based, besides others, on works like Shieber 85 [19], Gorz 88 [9] Ney 93 [16] Paseler 88 [17] Fujisaki et al. 91 [8] For a more detailled description see Weber 94 [20] Some of the work concerning the coupling with the beam decoder is a joint work with Andreas Hauenstein 1 and has already been published in Hauenstein Weber 94 [10, 11] 2 ....

....afterwards, using all of the features and propagating them through the series. All rules and daughters have integer codes assigned, so the table lookup as well as the redundancy check in the chart become cheap. Our treatment of rules can be seen as a variant of Shieber 85 s Restriction mechanism [19], based on types. As a result of this technique, there are no more unifications involved in predictor or scanner operations in Earley s terms. Only during the completer step unifications are actually carried out. The same technique cannot be used when parsing bottom up. The trigger for a SEEK UP ....

[Article contains additional citation context not shown here]

Stuart M. Shieber. Using restriction to extend parsing algorithms for complexfeature -based formalisms. In Proc. of the ACL 1985, volume 23, pages 145--152, 1985.

....meaningfully to hpsg signs, two conditions must be met. Firstly, parse tree construction must not be duplicated in the feature structures (by means of the hpsg DTRS feature) but be left to the parser (i.e. recorded in the chart) this is achieved in a standard way by feature structure restriction (Shieber, 1985) applied to all passive edges. Secondly, the processing of constraints that do not restrict the search space but build up new (often semantic) structure should be postponed, since they are likely to interfere with subsumption. For example, analyses that di er only with respect to PP attachment ....

Shieber, S. M. (1985). Using restriction to extend parsing algorithms for complex feature-based formalisms.

.... for more general or equivalent items rather than identical items calculate consequences: search for consistent rather than identical items Efficient Parsing ffl Head Corner (or Left Corner) Parser [63, 46, 56, 40, 79, 67, 73, 74, 72, 78, 62, 80, 45, 82, 50, 85, 87] ffl With Goal weakening [70] and Selective Memoization [99, 39] ffl Generalized for Word graphs ffl Suitable for Robust Parsing ffl Implemented in Prolog Why Prolog ffl Prolog provides a built in unification operation ffl Prolog provides a built in backtrack search procedure; memoization can be applied selectively ffl ....

Stuart M. Shieber. Using restriction to extend parsing algorithms for complex-feature-based formalisms. In 23th Annual Meeting of the Association for Computational Linguistics, Chicago, 1985.

....new automata. We do not have to worry about the tabular evaluation of these automata: we can always choose the TAB interpretation with ffl = 1 and in most cases find a better ffl. 5.1. 4 Shieber s restrictions We can re interpret and extend the mechanism of restrictions proposed by Shieber [15] with the Call Return model (and tabulation) Shieber s restrictions have been proposed to parse feature based formalisms in the context of computational linguistics, but we can discuss them for logic programming. 8 The idea is to use predictive bottom up evaluation (as Earley s or Magic Sets) ....

.... A can be unified using oe with r k:i ( X k:i ) and pushes call(B) where B = Phi(A k:i 1 oe) The resulting automaton may not be a SPDA: the altered transition [C] may violate the axioms [M CX] and [I COM] However, the automaton is an SPDA for all the restrictions considered by Shieber [15], which are monotonic (A B ) PhiA PhiB) and idempotent ( Phi 2 A = PhiA) More generally, a restriction Phi may be static for some terms A k:i ( PhiA k:i )oe = Phi(A k:i oe) and dynamic for the others. Use of restrictions in [15] was designed assuming that R k:i hold the same ....

[Article contains additional citation context not shown here]

Stuart M. Shieber. Using restriction to extend parsing algorithms for complex-- feature--based formalisms. In Proc. of the 23rd Annual Meeting of the Association for Computational Linguistics, pages 145--152, 1985.

....copy of the chart in which edges are labeled only by the nonterminal symbols contained in their context free backbone, which gives us more efficient indexing of the full grammar rules. Other optimizations include using one word look ahead before adding new predictions, and using restrictors (Shieber, 1985) to increase the generality of the predictions. For grammar with start symbol Sigma, phrase structure rules P , lexicon L, context independent categories CI, and context dependent categories CD; and for word string w = w 1 : w n : if Sigma 2 CD, predict( Sigma, 0) add empty categories(0) ....

....we do not believe that the bottom up and left corner algorithms we tested suffered from any unnecessary overheads from being implemented as special cases of our general algorithm, as we removed calls to subroutines that were unnecessary for those special cases. as Shieber s notion of restriction (Shieber, 1985) (which we also use) In methods such as Shieber s, predictions are weakened in ways that can result in an overall gain in efficiency, but predictions nevertheless must be dynamically generated for every phrase that is built bottom up. In our method, no predictions need to be generated for the ....

Shieber, S. 1985. Using Restriction to Extend Parsing Algorithms for Complex-FeatureBased Formalisms, in 23rd Annual Meeting of the Association for Computational Linguistics, Chicago, Illinois (July), pp. 145--152.

....The discussion of Earley s algorithm here will be necessarily brief. For details of the algorithm, the reader is referred to Earley (1970) For a logical reconstruction of this parsing algorithm, see Pereira and Warren (1983) For augmentation of Earley s algorithm with feature structures, see Shieber (1985). 4 To follow Earley (1970) the predictor would have to first predict every rule of the form N terminal and then scanning would be performed just for the prediction N duck. The scanner assumed here avoids this inefficiency by scanning preterminals instead of terminals. 168 Linguistics and ....

....with a particular semantic reading (represented as a feature structure) or perhaps a semantic reading with some additional syntactic properties. To be able to specify such a complex start category, we must assume that Earley s algorithm is augmented to allow feature structures as described in Shieber (1985). 2.1.1 Shieber s Approach Earley s algorithm was first adapted for generation by Shieber (1988) Shieber s approach is based on Pereira and Warren s (1983) idea of Earley deduction. In this approach, each edge is represented as a state [i; N Vm : V n; j] This state represents an edge ....

[Article contains additional citation context not shown here]

Shieber, S. 1985. Using Restriction to Extend Parsing Algorithms for Complex -Feature-Based Formalisms. In ACL Proceedings, 23rd Annual Meeting, 145--152.

....recognised word(Turino) Each parser is a unificationbased left corner chart parser which processes input bottom up with top down filtering (left corner dependencies) and left to right. The parsing algorithm largely corresponds to the enhancement of Earley s algorithm [8] presented by Shieber [9] or to the Earley algorithm described in [10] The parsers are implemented in C and include a standard ANSI C API (Application Program Interface) The API mainly provides commands for loading external grammar files, activating and deactivating sub grammars (if the grammar is organised into sub ....

S.M. Shieber,: Using Restriction to Extend Parsing Algorithms for Complex-feature-based Formalims. Proceedings of the 22nd Annual Meeting of the Association for Computational Linguistics, Chicago 1985, pp. 145-152.

....we will show how we can combine this technique with a clever indexing mechanism that is used for parsing and generation) Use of a restrictor The prediction rule is used for predicting new instantiations of grammar rules using the selected element of a non unit lemma. As known from the work of [ Shieber, 1985 ] prediction can lead to arbitrary numbers of consequents through repeated application when used with a grammar with an infinite structured nonterminal domain. For example, if a grammar handles subcategorization with list value features (such as in hpsg or the rule (r 1 ) of the grammar given in ....

....the constant function yielding the trivial model, no top down information is used. In Shieber s original work, a grammar oriented version of a restriction function is presented, where only those features are predicted which have been chosen by the user as most useful for prediction. Following [ Shieber, 1985 ] restriction is defined on the basis of a given restrictor R, which is a finite set of paths through a feature structure. The restriction of a feature structure F relative to R is the most specific feature structure F 0 v F , such that every path in F 0 has either an atomic value or is an ....

[Article contains additional citation context not shown here]

S. M. Shieber. Using restriction to extend parsing algorithms for complex-feature-based formalisms. In 23th Annual Meeting of the Association for Computational Linguistics, Chicago, 1985.

....Semantic descriptions are expressed there by additional constraints. Though theoretically very attractive, codescription has its price: ffl difficult to modularize ffl computational overhead when parsers use the complete descriptions Problems of these kinds which were already noted by Shieber 1985 motivated the research described here. The goal is to develop more flexible ways of using codescriptive grammars than having them applied by a parser with full informational power. The underlying observation is that constraints in such grammars can play different roles: ffl Genuine constraints ....

Shieber, S. M. 1985. Using Restriction to Extend Parsing Algorithms for ComplexFeature -Based Formalisms. In Proceedings of the 23rd Annual Meeting of the Association for Computational Linguistics, ACL-85, 145--152.

....unifies their correction proofs, and suggests improvements. When using stack elements in richer domains (e.g. featurestructures or constraints) we can improve computational performance by appropriately choosing the weak extractor ffl. This goes beyond techniques such as Shieber s restrictions[Shi85], since the use of ffl allows to carry less information in the return (bottom up) steps. The notion of stack presented can still be extended. In particular, we are currently working on analyzing tabular algorithms for TAGs or LIGs [VSW91] by considering automata on stack products (considered as ....

Stuart M. Shieber. Using restriction to extend parsing algorithms for complex--feature--based formalisms. In Proc. of the 23rd Annual Meeting of the Association for Computationnal Linguistic, pages 145--152, 1985.

No context found.

Shieber, S.: Using restriction to extend parsing algorithms for complex feature-based formalisms. In: Proceedings of the 23rd Meeting of the Association for Computational Linguistics, Chicago, IL (1985) 145 -- 152

No context found.

Shieber, S.M.: Using Restriction to Extend Parsing Algorithms for Complex Feature Based Formalisms, in: Proc. ACL 25, 1985

No context found.

Stuart M. Shieber. 1985. Using restriction to extend parsing algorithms for complex-feature-based formalisms.

No context found.

Stuart C. Shieber. 1985. Using restriction to extend parsing algorithms for complex feature based formalisms. In ACL85.

No context found.

Stuart C. Shieber. 1985. Using restriction to extend parsing algorithms for complex feature based formalisms. In Proceedings 23rd Annual Meeting of the Association for Computational Linguistics.

First 50 documents  Next 50

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