The problems of access -- retrieving linguistic structure from some mental grammar -- and disambiguation -- choosing among these structures to correctly parse ambiguous linguistic input -- are fundamental to language understanding. The literature abounds with psychological results on lexical access, the access of idioms, syntactic rule access, parsing preferences, syntactic disambiguation, and the processing of garden-path sentences. Unfortunately, it has been difficult to combine models which account for these results to build a general, uniform model of access and disambiguation at the lexical, idiomatic, and syntactic levels. For example psycholinguistic theories of lexical access and idiom access and parsing theories of syntactic rule access have almost no commonality in methodology or coverage of psycholinguistic data. This paper presents a single probabilistic algorithm which models both the access and disambiguation of linguistic knowledge. The algorithm is based on a parallel parser which ranks constructions for access, and interpretations for disambiguation, by their conditional probability. Low-ranked constructions and interpretations are pruned through beam-search; this pruning accounts, among other things, for the garden-path effect. I show that this motivated probabilistic treatment accounts for a wide variety of psycholinguistic results, arguing for a more uniform representation of linguistic knowledge and for the use of probabilisticallyenriched grammars and interpreters as models of human knowledge of and processing of language.

Abstract – This paper presents a fundamental algorithm for parsing natural language sentences into dependency trees. Unlike phrase-structure (constituency) parsers, this algorithm operates one word at a time, attaching each word as soon as it can be attached, corresponding to properties claimed for the parser in the human brain. Like phrasestructure parsing, its worst-case complexity is O(n 3), but in human language, the worst case occurs only for small n. 1 Overview. This paper develops, from first principles, several variations on a fundamental algorithm for parsing natural language into dependency trees. This is an exposition of an algorithm that has been known, in some form, since the 1960s but is not presented systematically in the extant literature. Unlike phrase-structure (constituency) parsers, this algorithm operates one word at a time, attaching each word as soon as it can be attached. There is good evidence that the parsing process used by the human mind has these properties [1]. 2 Dependency grammar. 2.1 The key concept. There are two ways to describe sentence structure in natural language: by breaking up the sentence into constituents (phrases), which are then broken into smaller constituents (Fig. 1), or by drawing links connecting

Many theories have been proposed to explain difficulty with center embedded constructions, most attributing the problem to some kind of limited capacity short-term memory. However, these theories have developed for the most part independently of more traditional memory research, which has focused on uncovering general principles such as chunking and interference. This article attempts to gain some unification with this research by suggesting that an interesting range of core sentence processing phenomena can be explained as interference effects in a sharply limited syntactic working memory. These include difficult and acceptable embeddings, as well as certain limitations on ambiguity resolution, length effects in garden path structures, and the requirement for locality in syntactic structure. The theory takes the form of an architecture for parsing which can index no more than two constituents under the same syntactic relation. A limitation of two or three items shows up in a variety o...

This article addresses the question of how the processor decides on its initial strategy for syntactic ambiguity resolution. At a point of ambiguity, more than one analysis is possible. An effective strategy might be to adopt the analysis that has most frequently turned out to be correct in the past. Assuming that the world stays the same in most respects, the analysis that has most frequently been correct in the past should provide a good estimate of which analysis is most likely to be correct again. Hence, by adopting this analysis, the processor should make fewer errors than if it chose any other analysis

The strict competence hypothesis has sparked a small dialogue among several researchers attempting to understand its ramifications for human sentence processing and incremental interpretation in particular. In this paper, we review the dialogue, reconstructing the arguments in an attempt to make them more uniform and crisper, and provide our own analyses of certain of the issues that arise. We argue that strict competence, because it requires a synchronous computation mechanism, may actually lead to more complex, rather than simpler, models of incremental interpretation. Asynchronous computation, which is arguably both psychologically more plausible and conceptually more basic, allows for incremental interpretation to fall out naturally, without additional machinery for interpreting partial constituents. We show that this is true regardless of whether the presumed interpretation mechanism is top-down or bottomup, contra previous conclusions in the literature, and propose a particular i...

Linguistic ambiguity is the greatest obstacle to achieving practical computational systems for natural language understanding. By contrast, people experience surprisingly little difficulty in interpreting ambiguous linguistic input. This dissertation explores distributed computational techniques for mimicking the human ability to resolve syntactic ambiguities efficiently and effectively. The competitive attachment theory of parsing formulates the processing of an ambiguity as a competition for activation within a hybrid connectionist network. Determining the grammaticality of an input relies on a new approach to distributed communication that integrates numeric and symbolic constraints on passing features through the parsing network. The method establishes syntactic relations both incrementally and efficiently, and underlies the ability of the model to establish long-distance syntactic relations using only local communication within a network. The competitive distribution of numeric ev...

Abstract not found

A prevalent trend in modeling human sentence processing has been to account for both initial attachment preferences and reanalysis behaviors with minimal extensions to a presumed set of initial parsing operations. Here, an entirely different formulation of the initial attachment and revision processes is suggested. Rather than assuming that all parsing is (as much as possible) initial attachment, the opposite approach is advocated: that all parsing---even initial attachment---is restructuring. The realization of parsing as restructuring arises from a set of independently motivated computational assumptions within the competitive attachment architecture, a hybrid connectionist model of the human sentence processor. Central to the model is a unique parallel attachment operation that simultaneously attaches the current input phrase, while reattaching previously structured phrases. Within this model, reanalysis is not a separate process or module, but rather a side effect of the primary means of forming syntactic structures. The ease of performing possible reanalyses is therefore determined by the same conditions, such as recency and lexical preferences, that affect initial attachments. Furthermore, independently motivated constraints on the network structure determine the allowable syntactic configurations that may undergo restructuring within the competitive attachment operation. The model thus also provides a computational explanation of gardenpath sentences, in which automatic reanalysis is impossible.

This paper investigates the strength of the local attachment preference in syntactic ambiguity resolution, based on a study of a novel ambiguity for which the predictions of local attachment contrast with the predictions of a wide range of other ambiguity resolution principles. In sentences of the form ' 'Because Rose praised the recipe I made... " we show that the ambiguous clause "I made " is preferentially attached as a relative clause under some circumstances, as predicted by local attachment, and preferentially attached as a matrix clause under other circumstances. The implications for accounts of locality in parsing are discussed. THE LOCALITY PUZZLE This paper is a progress report on our work which investigates the strength and the generality of the local attachment preference (see Gibson, Pearlmutter, Canseco-Gonzalez, & Hickok 1996; Phillips 1995, 1996). We use this term to refer in a theory neutral way to whatever underlies the inter-This is a revised version of a talk presented at the Ninth Annual CUNY Conference on Sentence Processing, New York. We are grateflil to the audience for their comments and suggestions. We would also like to thank Neal Pearimutter, Carson ScMitze, San Tunstall, Andrea Zukowski, and an anonymous reviewer for valuable assistance and suggestions with this paper. Needless to say, all the remaining errors are our own. The first author's

Abstract not found