Abstract: The formalism of synchronous tree-adjoining grammars, a variant of standard tree-adjoining grammars (TAG), was intended to allow the use of TAGs for language transduction in addition to language specification. In previous work, the definition of the transduction relation defined by a synchronous TAG was given by appeal to an iterative rewriting process. The rewriting definition of derivation is problematic in that it greatly extends the expressivity of the formalism and makes the design of parsing algorithms difficult if not impossible. We introduce a simple, natural definition of synchronous tree-adjoining derivation, based on isomorphisms between standard tree-adjoining derivations, that avoids the expressivity and implementability problems of the original rewriting definition. The decrease in expressivity, which would otherwise make the method unusable, is offset by the incorporation of an alternative definition of standard tree-adjoining derivation, previously proposed for completely separate reasons, thereby making it practical to entertain using the natural definition of synchronous derivation. Nonetheless, some remaining problematic cases call for yel more flexibility in the definition; the isomorphism requirement may have to be relaxed. It remains for future research to rune the exact requirements on the allowable mappings.

Abstract: The frequent and conventional use of nonliteral language has been a major stumbling block for natural language processing systems since the early machine translation efforts. Metaphor, metonymy, and indirect speech acts are among the most troublesome phenomena. Recent computational efforts addressing these problems have taken an approach that emphasizes the use of systematic knowledge about nonliteral language conventions. We are currently engaged in an effort to supply this knowledge in the case of conventional metaphor. We are constructing MetaBank: an empirically derived and theoretically motivated knowledge-base of English metaphorical conventions. This article describes our three-part approach to the construction of MetaBank: the collection of on-line textual resources and databases of linguistic generalizations, the development of a methodology for analyzing these resources, and the construction of a knowledge-base based on the preceding analyses.

Abstract: Partial descriptions are sets of constraints which do not necessarily determine a unique object. Two related issues arise when objects arc specified by partial descriptions. The first asks whether a given partial description is satisfied by any object. The second seeks to produce, from a satisfiable description, an appropriate representative of those objects which satisfy it (where the notion of appropriate is defined by the circumstances in which the partial descriptions are employed). We explore these issues in the context of two recent proposals utilizing partial descriptions of trees in the area of tree-adjoining grammars. One involves the use of descriptions of trees in reinterpreting TAGs as a constraint-based formalism. In the other these descriptions are the basis of a compact representation of lexicalized TAG grammars. In this context, what counts as an appropriate representative of the set of trees which satisfy a description is a tree in that set which is minimal in the sense that every other tree in the set can be derived from it by adjunction. A partial description for which there is a unique such minimal tree will be called a quasi-tree. Our work provides precise foundations for the use of descriptions of trees in these proposals. We formalize the notions of partial descriptions of trees and of quasi-trees, and using this, provide a mechanism which derives, from an arbitrary description, an equivalent set of quasi-trees. The description is satisfiable iff that set of quasi-trees is nonempty. We then provide a second mechanism which constructs the unique minimal trees satisfying these quasi-trees. Thus these mechanisms decide our issues in a strong way–they resolve any description into an equivalent set of descriptions which have unambiguous representative trees, and produce, from that set of descriptions, those representative trees. Such mechanisms are presupposed by the LTAG proposal, but are provided in a complete form for the first time here.

Abstract: This paper describes a bidirectional head-corner parser for (unification-based versions of) lexicalized tree-adjoining grammars.

Abstract: The ability to reuse existing plans to solve new planning problems can enable a domain-independent planner to improve its average case efficiency by exploiting the problem distribution and avoiding repetition of planning effort. The pay-off from plan reuse, however, crucially depends on finding effective solutions to two important underlying control problems: (i) controlling the retrieval of an appropriate plan and mapping to be reused in a new situation, and (ii) controlling the modification (refitting) of the retrieved plan so as to minimize perturbation to the applicable parts of the plan. This paper is concerned with the development of efficient domain-independent solutions to these two problems. For the retrieval, it provides a domain independent similarity metric that utilizes the plan causal dependency structure to estimate the utility of reusing a given plan in a new problem situation. For the refitting, it presents a minimum-conflict heuristic, again based on the causal dependency structure of the plan, to conservatively control the modification. The paper also discusses the implementation and evaluation of these strategies within the PRIAR plan modification framework.

Abstract: In many head-final languages such as German, Hindi, Japanese, and Korean, but also in some other languages such as Russian, arguments of a verb can occur in any order. Furthermore, arguments can occur outside of their clause (“long-distance scrambling”). Long-distance scrambling presents a challenge both to linguistic theory and to formal frameworks for linguistic description because it is very unconstrained: in a given sentence, there is no bound on the number of elements that can be scrambled nor on the distance over which each element can scramble. We discuss two formal frameworks related to tree-adjoining grammar. First, we show how scrambling facts from Korean can be handled by nonlocal multicomponent TAG (MC-TAG). Then, we argue that overt vWt-movement in German makes this analysis unattractive, and suggest a new version of MC-TAG, called V-TAG, which can handle both Korean and German word order variation. Interestingly, this new version has more attractive computational properties than nonlocal MC-TAG. We conclude that this formalism is an attractive basis for the development of psycholinguistic processing models and practical parsers alike.

Abstract: Exploiting an incremental and parallel processing scheme is useful to improve the performance of natural language generation systems. TAG-GEN is a TAG-based syntactic generator that realizes both principles. It is shown how the demands of incremental and parallel generation influence the definition, the design, and the processing of syntactic rules on the basis of tree-adjoining grammars.

Abstract: This paper discusses variants of nondeterministic one-way S-automata and context-free S-grammars where S is a storage type. The framework that these systems provide can be used to give alternative formulations of embedded pushdown automata and linear indexed grammars. The embedded pushdown automata is obtained by means of a linear version of a class of storage types called iterated pushdowns. Linear indexed grammar is obtained by using the pushdown storage type and restricting the way in which the grammar uses its storage.

Abstract: This paper presents an algorithm (a parser) for analyzing sentences according to grammatical constraints expressed in the framework of lexicalized tree-adjoining grammar. For the current grammars of English, the algorithm behaves much better and requires much less time than its worst-case complexity. The main objective of this work is to design a practical parser whose average-case complexity is much superior to its worst case. Most of the previous methods always required the worst-case complexity. The algorithm can be used in two modes. As a recognizer it outputs whether the input sentence is grammatically correct or not. As a parser it outputs a detailed analysis of the grammatically correct sentences. As sentences are read from left to right, information about possible continuations of the sentence is computed. In this sense, the algorithm is called a predictive left to right parser. This feature reduces the average time required to process a given sentence. In the worst case, the parser requires an amount of time proportional to G2n6 for a sentence of n words and for a lexicalized tree-adjoining grammar of size G. The worst-case complexity is only reached with pathological (not naturally occurring) grammars and inputs.

Abstract: This paper presents a knowledge-based learning method and reports on case studies in different domains. The method integrates abduction and explanation-based learning. Abduction provides an improved method for constructing explanations. The improvement enlarges the set of examples that can be explained so that one can learn from additional examples using traditional explanation-based macro learning. Abduction also provides a form of knowledge level learning. Descriptions of case studies show how to set up abduction engines for tasks in particular domains. The case studies involve over a hundred examples taken from diverse domains requiring logical, physical, and psychological knowledge and reasoning. The case studies are relevant to a wide range of practical tasks including natural language understanding and plan recognition; qualitative physical reasoning and postdiction; diagnosis and signal interpretation; and decision making under uncertainty. The descriptions of the case studies include an example, its explanation, and discussions of what is learned by macro-learning and by abductive inference. The paper discusses how to provide and represent the domain knowledge and meta-knowledge needed for abduction and search control. The main conclusion is that abductive inference is important for learning. Abduction and macro-learning are complementary and synergistic.

Abstract: A main purpose of natural language systems (NL-systems) is the simulation of human speech behavior, which divides into generation and analysis, in all kinds of situations. Humans process natural language in an incremental style, which means that sentences are generated or analyzed in a piecemeal fashion. Observations show that there are bidirectional dependencies between these processes. During parsing, hypotheses for the continuation of an utterance may be generated. During generation, the output may be parsed in order to detect ambiguities. This paper presents a two-level incremental parsing algorithm for LD/TLP-TAGs on the basis of Earley's context-free parser. LD/TLP-TAG is a new extension for TAG that increases the descriptive power of the formalism without changing its generative power. Elementary rules are split into local dominance description (LD) and linear precedence rules (TLP) allowing for generalizations of ordering constraints. The presented algorithm is used within an anticipation feedback loop for the incremental syntactic generator TAG-GEN.

Abstract: We show how multicomponent TAGs, although in their “local” version equivalent to “simple” TAGs, enable one to associate the derived structures with a richer set of derivations. This is especially useful for representing some nonlocal dependencies such as wh-extraction out of NP or extraposition, as soon as one interprets the derivation structures as representing the predicate-argument dependencies. However, there is a formal alternative in the formalism of synchronous TAGs that achieves a similar result, namely, that associates standard syntactic derivations with nonisomorphic semantic derivations, and relies on the latter to represent the dependency relations. We show that the latter treatment accounts for a different range of extraposition facts in French, where the relationship between the head noun and its extraposed dependent is a semantic one. Nous montrons comment les TAGs ensemblistes “locales” (MCTAGs), qui sont equivalentes aux TAGs simples, permettent ?associer aux structures syntaxiques derivees un ensemble plus large de structures derivationnelles. Cela est particulierement utile pour la representation des dependances non locales qu'on observe dans les extractions hors du groupe nominal et les extrapositions, a partirdu moment ouľon interprete les arbres de derivation comme representant les relations de dependance. II existe toutefois une autre possibilitye dans le cadre des TAGs synchrones qui associe aux derivations syntaxiques standard des derivations semantiques non isomorphes et interprete les relations de dependance en se basant sur ces dernieres. Nous montrons que cette deuxieme possibilitye est mieux adaptee pour des cas ?extraposition en francais ou la relation de dependance entre le nom těte et ľelement extrapose est ?ordre semantique.

Abstract: We address the problem of reasoning in cases when knowledge bases containing background knowledge are understood not as sets of formulas (rules and facts) but as collections of partially ordered theories. In our system, the usual, two-part logical structures, consisting of a metalevel and an object level, are augmented by a third level–a referential level. The referential level is a partially ordered collection of theories; it encodes background knowledge. As usual, current situations are described on the object level, and the metalevel is a place for rules that can eliminate some of the models permitted by the object level and the referential level. As a logic of reasoning the system generalizes the standard model of a rational agent: deducing actions and deriving new information about the world from a logical theory–its knowledge base. It is a natural logical system in which priorities on the possible readings of predicates, not special rules of inference, are the main source of nonmonotonicity. We introduce a theory forming operator PT(x) to exploit the priorities, and we investigate its basic logical properties. Then we show how such a system can be augmented by metarules. Although this paper concentrates on basic logical properties of the new theory, this formalism has already been applied to model a number of natural language phenomena such as the notion of text coherence, Gricean maxims, vagueness, and a few others. The paper also briefly compares it with the model of background knowledge of CYC, as proposed by Lenat and Guha.

Abstract: Pattern-weight pairs (pws) are a new form of search and planning knowledge. Pws are predicates over states coupled with a least upper bound on the distance from any state satisfying that predicate to any goal state. The relationship of pws to more traditional forms of search knowledge is explored with emphasis on macros and subgoals. It is shown how pws may be used to overcome some of the difficulties associated with macro-tables and lead to shorter solution paths without replanning. An algorithm is given for converting a macro-table to a more powerful pw set. Superiority over the Squeeze algorithm is demonstrated. It is also shown how pws provide a mechanism for achieving dynamic subgoaling through the combination of knowledge from multiple alternative subgoal sequences. The flexibility and execution time reasoning provided by pws may have significant use in reactive domains. The main cost associated with pws is the cost of applying them at execution time. An associative retrieval algorithm is given that expedites this matching-evaluation process.

Abstract: A number of grammatical formalisms have been proposed to describe the syntax of natural languages, and the universal recognition problems for some of those classes of grammars have been studied. A universal recognition problem for a class Q of grammars is the one to decide, taking a grammar G ∈ G and a string ui as an input, whether G can generate w or not. In this paper, the computational complexities of the universal recognition problems for parallel multiple context-free grammars, multiple context-free grammars, and their subclasses are discussed.

Abstract: This paper presents an evaluation of a heuristic for partial-order planning, known as temporal coherence. The temporal coherence heuristic was proposed by Drummond and Currie as a method to improve the efficiency of partial-order planning without losing the ability to find a solution (i.e., completeness). It works by using a set of domain constraints to prune away plans that do not “make sense,” or are temporally incoherent. Our analysis shows that, while intuitively appealing, temporal coherence can only be applied to a very specific implementation of a partial-order planner and still maintain completeness. Furthermore, the heuristic does not always improve planning efficiency; in some cases, its application can actually degrade the efficiency of planning dramatically. To understand when the heuristic will work well, we conducted complexity analysis and empirical tests. Our results show that temporal coherence works well when strong domain constraints exist that significantly reduce the search space, when the number of subgoals is small, when the plan size is not too large, and when it is inexpensive to check each domain constraint.

Abstract: Dimensional analysis, traditionally used in physics and engineering to identify quantitative relationships, has recently been applied to qualitative reasoning of physical systems. We illustrate some problems of this approach. In the light of this, we reexamine the fundamentals of dimensional analysis in order to more precisely characterize its scope and limitations as a tool in qualitative reasoning. We also explore its relationship to state equation representations of physical systems. In particular, we describe its value in providing a set of constraints to reduce the ambiguity that bedevils qualitative reasoning schemes. We argue that dimensional analysis should not be seen as a substitute for knowledge about the physics but rather a supplement to other sources of knowledge.

Abstract: In this paper, we introduce and define a new class of automata (pushdown automata with n stacks, abbreviated as n-SA). The ultimate aim is to give a new characterization of LCRFL, the class of languages accepted by a linear context-free rewriting system (LCFRS). In particular, we introduce 2-SA as a new automaton model for tree-adjoining grammars (TAG). In the simplest cases (0-SA and 1-SA), the languages that are accepted by the automata are the regular and context-free languages respectively. A more complex case is the case of a 2-SA which accepts TALs. The n-SA creates an infinite hierarchy of languages and it seems that this hierarchy corresponds to others in the class LCFRL. The 2-SA corresponds closely to the EPDA (embedded pushdown automaton, an automaton model equivalent to TAGs). Unlike the EPDA, which allows push operations “below the top stack,” an n-SA allows push and pop operations only on the top of their (multiple) stacks. So n-SA trade simpler operations against an also simpler but expanded storage structure.

Abstract: This paper explores the application of a rich model of pragmatic context to the problem of identifying and correcting real-word spelling errors. Results suggest that such a model can be useful for generating and ranking a list of possible corrections according to their contextual relevance. In the domain of expert consultation discourse, a mode! of pragmatic context must represent not only the user's domain plans, but also the problem-solving processes that explore alternative plans, refining and instantiating the intended plan, and the connections between those problem-solving moves and their resulting discourse manifestations. In the model presented, metaplans are used to represent these problem-solving and discourse levels, while heuristics that take into account the user's problem-solving strategies and world knowledge serve to rank the relative likelihood of different possible next queries. An implementation of this model has been used to suggest pragmatically coherent interpretations that can be matched against a partial parse of the input in order to generate possible corrections for real-word spelling errors.

Abstract: Coupled-context-free grammars are a natural generalisation of context-free grammars obtained by combining nonterminals to corresponding parentheses which can only be substituted simultaneously. Refering to the generative capacity of the grammars we obtain an infinite hierarchy of languages that comprises the context-free languages as the first and all the languages generated by TAGs as the second element. Here, we present a generalization of the context-free LL (k)-notion onto coupled-context-free grammars, which leads to a characterization of subclasses of coupled-context-free grammars-and in this way of TAGs as well-which can be parsed in linear time. The parsing procedure described works incrementally so that it can be used for on-line parsing of natural language. Examples show that important elements of the tree-adjoining languages can be generated by LL(k)-coupled-context-free grammars

Abstract: It is possible within tree-adjoining grammar to reproduce many of the syntactic analyses originally formulated by linguists in transformational terms. To the extent that these analyses are well motivated empirically, this fact makes TAG interesting for use in developing computational learning and processing models (Joshi 1990; Frank 1992; Rambow 1994), since the use of other nontransformational formalisms sometimes forces choices of linguistic description different from those ordinarily made by descriptive syntacticians. Thus, using TAG, one can take advantage in the construction of parsers and learners of the computational tractability of a mathematically restrictive formalism without having to reinvent empirical syntax in order to do so. At the same time, TAG analyses are not identical in every detail to their transformational counterparts; and it is interesting to compare them where they diverge. The differences arise because of a fundamental difference in the way that syntactic recursion is treated in the two frameworks. In TAG, recursive structures are generated by composing elementary syntactic objects, with the result that recursion is factored apart from the representation of local syntactic dependencies. By contrast, in transformational grammar, as in many other frameworks, recursive structure and local dependencies are represented together in a single, full representation of a complex sentence. Because of this difference in the treatment of recursion, it often turns out, when TAG is used to emulate a transformational analysis, that the TAG version has advantages, of both elegance and empirical coverage, over the original. This paper is a demonstration in a new empirical domain, that of nominal constructions, of the advantages of TAG-based syntax. By presenting a linguistically detailed account of these constructions and showing the advantages of using TAG to analyze them, we strengthen the case for the use of mathematically constrained and computationally tractable representational systems in competence-based as well as in computational linguistics.

Abstract: This paper extends the “parsing as deduction” approach to tree-adjoining grammars by showing how a TAG recognition problem can be reduced to a Datalog deduction problem, and presents an SLD selection rule that makes the proof search correspond to a top-down parse using the original grammar. Just as in the DCG extension of context-free grammars, this approach permits nodes tn be labeled with first-order terms (rather than only atomic symbols). Finally the paper discusses implementation matters, and describes how the control rule can be efficiently implemented in Prolog.

Abstract: Natural-language generation has gained importance since AI systems have become more sophisticated and canned text can no longer serve the elaborate communicative tasks in such a system. In this paper, suggestions are presented to apply uniformly the formalism of tree-adjoining grammar (TAG) for all tasks of natural-language generation. Up to now plan-based systems have been preferred for most of the processing during generation. Since the individual tasks of natural-language generation are very different and complex; for example, the determination of the propositional structure of a text is described using the TAG formalism–extended by constraints and feature descriptions (UCTAC). Nevertheless, this paper proposes basic ideas for the generalization of applying a grammar formalism to natural-language generation in order to build an overall integrated system for natural-language generation. Finally, advantages of such a uniform model are discussed.

Abstract: By means of a factorization of linear precedence (LP) information in tree-adjoining grammar (TAG), the possibility to capture syntactic generalizations can be enhanced, especially with respect to languages which exhibit relative free word-order. In TAG(LD/LP) LP-information is factorized in a way reminiscent of the ID/LP format of CFGs. However, in contrast to its context-free counterpart, in TAG(LD/LP) entanglement of constituents is allowed by admitting not only the permutation of sister nodes, but also of non-sister nodes in a structural description. In this paper a generalization of the predictive left-to-right parser for TAG is presented which is able to handle a restricted variant of TAG(LD/LP), which allows the permutation of sister nodes and non-sister nodes within elementary structures, the basic building blocks of TAG(LD/LP).

Abstract: This paper describes a natural language generation system developed for a spoken language translation system. Our system employs Feature-structure-based Tree Adjoining Grammar (FTAG) for generation knowledge representation. Each elementary tree of our grammar is paired with a semantic feature structure which is consistent with semantics defined in Head-driven Phrase Structure Grammar. Feature structures attached to nodes of elementary trees are unrestricted. Thus our formalism allows HPSG style phrase structure description as well as TAG style description. The advantage of our generation knowledge representation is the ability of incorporating HPSG style “core” grammar and TAG style case-based grammar. The system generates a syntactic tree by combining elementary trees so as to satisfy an input semantic structure. The generation algorithm is an application of a semantic head-driven generation. To carry out an adjoining operation. an elementary tree with an adjunction node is dynamically split at the adjunction node during generation.

Abstract: The work presented here attempts to bring out some fundamental concepts that underlie some known parsing algorithms, usually called chart or dynamic programming parsers, in the hope of guiding the design of similar algorithms for other formalisms that could be considered for describing the “surface” syntax of languages. The key idea is that chart parsing is essentially equivalent to a simple construction of the intersection of the language (represented by its grammar) with a regular set containing only the input sentence to be parsed (represented by a finite state machine). The resulting grammar for that intersection is precisely what is usually called a shared forest: it represents all parses of a syntactically ambiguous sentence. Since most techniques for processing ill-formed input can be modeled by considering a nonsingleton regular set of input sentences, we can expect to generalize these ill-formed input processing techniques to all parsers describable with our approach.

Abstract: Default logic has been introduced for handling reasoning with incomplete knowledge. It has been widely studied, and various definitions have been proposed for it. Most of the variants have been defined by means of fixed points of some operator. We propose here another approach, which is based on a study of the way in which general rules with exceptions, used in a default reasoning process, can contradict one another. We then isolate sets of noncontradicting rules, as large as possible in order to exploit as much information as possible, and construct, for each of these sets of rules, the set of conclusions that can be deduced from it. We show that our framework encompasses most of the existing variants of default logic, allowing those variants to be compared from a knowledge representation point of view. Our approach also enables us to provide an operational definition of extensions in some interesting cases. Proof-theoretical and semantical aspects are investigated.