WordNet is an on-line lexical reference system whose design is inspired by current

this dissertation. I single out for special thanks first a few of the Ventura Hall crowd, including Mfirvet Eng, Nancy Wiegand, Susan Stucky (the other Mennonite formal linguist), and Kathie Carpenter, Suzanne Kemmer and Michael Barlow, with whom I have happily shared every step of the Stanford graduate pilgrimage. Next, I warmly thank Gina Wein for her competent administrative support and for her friendship. Finally, I gratefully acknowledge the strong shaping influences of the members of the Stanford linguistics faculty, who teach and also model a vibrant and professional approach to linguistic research. Representative of these scholars are the three members of my reading committee, whose work and counsel have had a profound effect on my work; I thank Joan Bresnan, Ivan Sag, and my principal advisor, Thomas Wasow, whose patience, cheerful persistence, unstinting support, solid critique, creative ideas, and common sense made the writing of this thesis possible and enjoyable. Every student should have such an advisor

This paper introduces an extension of collocational analysis that takes into account grammatical structure and is specifically geared to investigating the interaction of lexemes and the grammatical constructions associated with them. The method is framed in a construction-based approach to language, i.e. it assumes that grammar consists of signs (form-meaning pairs), and is thus not fundamentally different from the lexicon. The method is applied to linguistic expressions at various levels of abstraction (words, semi-fixed phrases, argument structures, tense, aspect and mood). The method has two main applications: first, to increase the adequacy of grammatical description by providing an objective way of identifying the meaning of a grammatical construction and determining the degree to which particular slots in it prefer or are restricted to a particular set of lexemes; second, to provide data for linguistic theory-building

This paper focuses on this second proposal, called ON-LINE TYPE CONSTRUCTION.

This dissertation shows that there are syntactic constraints involved in the aspectual interpretation of the VP that involve the internal structure of the verbal complements. Although a simple compositional semantics would predict certain interpretations, the syntactic computational system has its own mechanisms, and the relevant configurations do not always obtain. Consequently, certain readings end up being blocked. Chapter 1 defines the relevant output conditions for aspectual interpretation based on the semantic work of Verkuyl (1995) and Krifka (1989), and proposes a syntactic configuration within the Minimalist Program for Linguistic Theory (Chomsky 1993) at which aspect is calculated. Chapter 2 shows that two classic cases of apparent grammaticalization of aspect, Finnish and Slavic can be reduced to independently needed elements in the grammar: A-quantifiers and movement to AgrO for Case checking. Chapter 3 examines Accusative Clitic Doubling in Spanish and provides a new analysis which accounts for the aspectual interpretations of the construction. Chapter 4 proposes a theory of Determiner Transparency in which definite determiners in relative clauses, certain kinds of adjectives and demonstratives do not select for their head NP and as a result, durative reading of the VP predicate are obtained if the head is non-quantized. Chapter 5 extends the analysis to "type " expressions, measure phrases and possessives (genitives). Chapter 6 provides an aspectual analysis of the copula verbs 'ser' and 'estar' in Brazilian Portuguese, and argues against their being the grammaticalization of the stage-level and individual-level distinction.

Time adverbials introduced by until impose restrictions on the aspectual class of the main clause they combine with: they only combine with durative sentences. In negative sentences, the situation is more complex. The question arises whether negative sentences are durative, or whether there is a separate use of until as a negative polarity item. In this paper, I discuss the three treatments of not... until that are characterized in the literature as the scope analysis, the ambiguity thesis and the lexical composition approach. I work out the interpretation of the three approaches in an event-based semantics, and argue that they are truth-conditionally equivalent in sentences containing an explicit negation. Furthermore, they generate the same pragmatic implicatures. A separate negative polarity use of until is motivated by sentences containing NPI-licensers different from explicit negation, though. The observation that the scope analysis, the ambiguity thesis and the lexical composition approach are semantically and pragmatically equivalent in sentences containing an explicit negation helps us describe the similarities and differences between the expression of exclusion of a range of values on the rime axis in a variety of languages.

In this paper, we address the issue of integrating semantic lexicons into NLG systems and argue that the problem of lexical choice in generation can be approached only by such an integration. We take the approach of Generatire Lexicon Theory (GLT) (Pustejovsky, 1991, 1994c) which provides a system involving four levels of representation connected by a set of generatire devices accounting for a compositional interpretation of words in context. We are interested in showing that we can reduce the set of collocations listed in the lexicon by introducing the notion of "semantic collo.cations" which can be predicted within GLT framework. We argue that the lack of semantic welldefined calculi in previous approaches, whether linguistic or conceptual, renders them unable to account for seman- tic collocations.

In this paper, we address the issue of syntagmatic expressions from a computational lexical semantic perspective. From a representational viewpoint, we argue for a hybrid approach combining linguistic and conceptual paradigms, in order to account for the continuum we find in natural languages from free combining words to frozen expressions. In particu-lar, we focus on the place of lexical and semantic restricted co-occurrences. From a t)rocessing view-point, we show how to generate/analyze syntag-matic expressions by using an efficient constraint-based processor, well fitted for a knowledge-driven approach. 1

It is well-established that (non-linguistic) categorization is driven by a functional demand of prediction. We suggest that prediction likewise may well play a role in motivating the learning of semantic generalizations about argument structure constructions. We report corpora statistics that indicate that the argument frame or construction has roughly equivalent cue validity as a predictor of overall sentence meaning as the morphological form of the verb, and has greater category validity. That is, the construction is at least as reliable and more available. Moreover, given the fact that many verbs have quite low cue validity in isolation, attention to the contribution of the construction is essential.

Whether particular arguments are overtly realized in languages like English is not random. A number of researchers have put forward sweeping generalizations in order to capture certain general tendencies. In this paper, however, it is argued that these analyses underestimate the role of constructions, detailed lexical semantics and discourse factors. Given sufficient attention to these factors, the general tendencies, as well as productive classes of systematic exceptions, follow without additional grammatical stipulation. 1.