This paper describes a system for extracting typed dependency parses of English sentences from phrase structure parses. In order to capture inherent relations occurring in corpus texts that can be critical in real-world applications, many NP relations are included in the set of grammatical relations used. We provide a comparison of our system with Minipar and the Link parser. The typed dependency extraction facility described here is integrated in the Stanford Parser, available for download. 1.

Logical hidden Markov models (LOHMMs) upgrade traditional hidden Markov models to deal with sequences of structured symbols in the form of logical atoms, rather than flat characters. This note formally introduces LOHMMs and presents solutions to the three central inference problems for LOHMMs: evaluation, most likely hidden state sequence and parameter estimation. The resulting representation and algorithms are experimentally evaluated on problems from the domain of bioinformatics. 1.

istic methods for syntax, just as for a long time McCarthy and Hayes (1969) discouraged exploration of probabilistic methods in Artificial Intelligence. Among his arguments were that: (i) Probabilistic models wrongly mix in world knowledge (New York occurs more in text than Dayton, Ohio, but for no linguistic reason), (ii) Probabilistic models don't model grammaticality (neither Colorless green ideas sleep furiously nor Furiously sleep ideas green colorless have previously been uttered -- and hence must be estimated to have probability zero, Chomsky wrongly assumes -- but the former is grammatical while the latter is not, and (iii) Use of probabilities does not meet the goal of describing the mind-internal I-language as opposed to the observed-in-the-world E-language. This chapter is not meant to be a detailed critique of Chomsky's arguments -- Abney (1996) provides a survey and a rebuttal, and Pereira (2000) has further useful discussion -- but some of these concerns are still importa

Exemplar-based models of language propose that human language production and understanding operate with a store of concrete linguistic experiences rather than with abstract linguistic rules. While exemplarbased models are well acknowledged in areas like phonology and morphology, common wisdom has it that they are intrinsically flawed for syntax where infinite generative capacity is needed. This paper shows that this common wisdom is wrong. It starts out by reviewing an exemplar-based syntactic model, known as Data-Oriented Parsing, or DOP, which operates on a corpus of phrase-structure trees. While this model is productive, it is inadequate from the point of grammatical productivity. We therefore extend it to the more sophisticated linguistic representations proposed by Lexical-Functional Grammar theory, resulting in the model known as LFG-DOP, which does allow for meta-linguistic judgments of acceptability. We show how DOP deals with first language acquisition, suggesting a unified model for language learning and language use, and go into a number of syntactic phenomena that can be explained by DOP but that challenge rulebased models. We argue that if there is anything innate in language cognition it is not Universal Grammar but “Universal Representation”. 1.

The same categorical phenomena which are attributed to hard grammatical constraints in some languages continue to show up as statistical preferences in other languages, motivating a grammatical model that can account for soft constraints. The effects of a hierarchy of person (1st, 2nd 3rd) on grammar are categorical in some languages, most famously in languages with inverse systems, but also in languages with person restrictions on passivization. In Lummi, for example, the person of the subject argument cannot be lower than the person of a nonsubject argument. If this would happen in the active, passivization is obligatory; if it would happen in the passive, the active is obligatory (Jelinek and Demers 1983). These facts follow from the theory of harmonic alignment in OT: constraints favoring the harmonic association of prominent person (1st, 2nd) with prominent syntactic function (subject) are hypothesized to be present as subhierarchies of the grammars of all languages, but to vary ...

We present a relatively large coverage French grammar written with the formalism of Interaction Grammars. This formalism combines two key ideas: the grammar is viewed as a constraint system, which is expressed through the notion of tree description, and the resource sensitivity of natural languages is used as a syntactic composition principle by means of a system of polarities. We give an outline of the expressivity of the formalism by modelling significant linguistic phenomena and we show that the grammar architecture provides for re-usability and tractability, which is crucial for building large coverage resources: a modular source grammar is distinguished from the object grammar which results from the compilation of the first one, and the lexicon is independent of the grammar. Finally, we present the results of an evaluation of the grammar achieved with the LEOPAR parser with a test suite of sentences.

We describe a method that automatically induces LFG f-structures from treebank tree representations, given a set of f-structure annotation principles that define partial, modular c- to f-structure correspondences in a linguistically informed, principle-based way.

Lexical-Functional Grammar f-structures are abstract syntactic representations approximating basic predicate-argument structure. Treebanks annotated with f-structure information are required as training resources for stochastic versions of unification and constraint-based grammars and for the automatic extraction of such resources. In a number of papers (Frank, 2000; Sadler, van Genabith and Way, 2000) have developed methods for automatically annotating treebank resources with f-structure information. However, to date, these methods have only been applied to treebank fragments of the order of a few hundred trees. In the present paper we present a new method that scales and has been applied to a complete treebank, in our case the WSJ section of Penn-II (Marcus et al, 1994), with more than 1,000,000 words in about 50,000 sentences.

The development of large coverage, rich unification- (constraint-) based grammar resources is very time consuming, expensive and requires lots of linguistic expertise. In this paper we report initial results on a new methodology that attempts to partially automate the development of substantial parts of large coverage, rich unification-(constraint-) based grammar resources. The method is based on a treebank resource (in our case Penn-II) and an automatic f-structure annotation algorithm that annotates treebank trees with proto-fstructure information. Based on these, we present two parsing architectures: in our pipeline architecture we firstextract a PCFG from the treebank following the method of [Charniak, 1993; Charniak, 1996] , use the PCFG to parse new text, automatically annotate the resulting trees with our f-structure annotation algorithm and generate proto-f-structures. By contrast, in the integrated architecture we firstautomatically annotate the treebank trees with fstructure information and then extract an annotated PCFG (A-PCFG) from the treebank. We then use the A-PCFG to parse new text to generate proto-fstructures. Currently

We present two companion methods for automatically enriching phrase-structure oriented treebank resources with functional structures. Both methods define systematic patterns of correspondence between partial PS configurations and functional structures. These are applied to PS rules extracted from treebanks, or to flat term representations of treebank trees.