Research

This dissertation investigates the topic of verbal ellipsis in English. Two main issues are addressed in this work: (i) the identity condition that restricts the application of ellipsis and (ii) the different locality restrictions that apply to elliptical constructions. The identity condition is examined from the point of view of competence, while the locality condition is given a natural answer from the processing domain. Furthermore, a parsing algorithm based on minimalist grammars is defined. Chapter 1 introduces the topic. Chapter 2 and Chapter 3 deal with the syntactic identity condition. Chapter 2 reviews some proposals in the literature, namely, Lasnik (1995b), Kitagawa (1991) and Fiengo and May (1994). All these analyses examine controversial examples where, apparently, partial syntactic identity between antecedent and gap is found. Chapter 3 presents a new analysis which assumes late lexical insertion, in the spirit of derivational morphology (Marantz 1993), and offers a unified account of all the cases of partial identity introduced in the previous chapter. It is argued that syntactic identity must be respected, and that the crucial notion for ellipsis is identity of syntactic categoriesa condition that is met before lexical items are inserted. Also, the different readings that obtain under ellipsis (i.e., sloppy and strict readings) are explained as emerging at different points in the derivation: before and after lexical insertion, respectively. Chapter 4 reviews one proposal in the parsing literature (Lappin and McCord 1990) as well as the problems it faces. Chapter 5 offers a processing account of the locality restrictions on gapping (as opposed to VPE and Pseudogapping)), those are analyzed as a result of (i) tense absence/presence (Fodor 1985), (ii) low initial attachment of coordinates, and (iii) Spell-out operations which render syntactic structure unavailable (Uriagereka 1999). A two-fold ellipsis resolution process is presented herewhere some work is done on-line, but some at the LF level. Chapter 6 defines an algorithm based on minimalist grammar operations, precisely on the preference of Merge-over-Move-over-Spell-out (as defined by Weinberg 1999); thus, showing that minimalist grammar models can be translated into computational models. Chapter 7 presents the conclusions.

The general aim of this thesis is to provide support for the claim that movement can be driven by theta-features, advanced by Bokovič (1994), Hornstein (1999, 2001), Manzini and Russo (2000), and O'Neil (1997) among others, through a study of Japanese Psych Verb constructions that exhibit interesting peculiarities. In some psych verb constructions, theta-roles are projected in an order that diverges from the canonical order found in other dyadic constructions. The theme role of Object Experiencer (OE) verbs is realized in the subject position of the sentence, while the experiencer role is linked to the object position. On the other hand, Subject Experiencer (SE) verbs map the theme role to the object position while the experiencer role is realized in the subject position. Given that in general experiencers are mapped to the subject/external argument position, OE verb constructions raise some critical issues for the Principles and Parameters theory (Chomsky, 1981), in particular for the theories of argument structure. The first goal of the thesis is to provide a solution to this linking puzzle as well as other peculiarities of OE verbs in Minimalist terms. In particular, I claim that the subject of an OE verb sentence is derived by thematically driven movement. By allowing such movement, the inverse linking pattern, backward binding phenomenon, and scope patterns of OE verbs can be accounted for straightforwardly. The second goal is to investigate the structures of SE verbs and OE verbs and how they are related one another. I propose that an OE verb is a mono-clausal causative, composed of an SE verb base and a causative morpheme -sase, and that SE verbs are bare VPs without vP projection. This amounts to saying that SE verbs do not project the external argument. It is shown that SE verbs do not allow passivization, supporting the claim that SE verbs do not project the external argument.

A dissertation on the syntax and comprehension of wh-dependencies.

Word Sense Disambiguation (WSD) is the process of resolving the meaning of a word unambiguously in a given natural language context. Within the scope of this thesis, it is the process of marking text with explicit sense labels. What constitutes a sense is a subject of great debate. An appealing perspective, aims to define senses in terms of their multilingual correspondences, an idea explored by several researchers, Dyvik (1998), Ide (1999), Resnik & Yarowsky (1999), and Chugur, Gonzalo & Verdejo (2002) but to date it has not been given any practical demonstration. This thesis is an empirical validation of these ideas of characterizing word meaning using cross-linguistic correspondences. The idea is that word meaning or word sense is quantifiable as much as it is uniquely translated in some language or set of languages. Consequently, we address the problem of WSD from a multilingual perspective; we expand the notion of context to encompass multilingual evidence. We devise a new approach to resolve word sense ambiguity in natural language, using a source of information that was never exploited on a large scale for WSD before. The core of the work presented builds on exploiting word correspondences across languages for sense distinction. In essence, it is a practical and functional implementation of a basic idea common to research interest in defining word meanings in cross-linguistic terms. We devise an algorithm, SALAAM for Sense Assignment Leveraging Alignment And Multilinguality, that empirically investigates the feasibility and the validity of utilizing translations for WSD. SALAAM is an unsupervised approach for word sense tagging of large amounts of text given a parallel corpus — texts in translation — and a sense inventory for one of the languages in the corpus. Using SALAAM, we obtain large amounts of sense annotated data in both languages of the parallel corpus, simultaneously. The quality of the tagging is rigorously evaluated for both languages of the corpora. The automatic unsupervised tagged data produced by SALAAM is further utilized to bootstrap a supervised learning WSD system, in essence, combining supervised and unsupervised approaches in an intelligent way to alleviate the resources acquisition bottleneck for supervised methods. Essentially, SALAAM is extended as an unsupervised approach for WSD within a learning framework; in many of the cases of the words disambiguated, SALAAM coupled with the machine learning system rivals the performance of a canonical supervised WSD system that relies on human tagged data for training. Realizing the fundamental role of similarity for SALAAM, we investigate different dimensions of semantic similarity as it applies to verbs since they are relatively more complex than nouns, which are the focus of the previous evaluations. We design a human judgment experiment to obtain human ratings on verbs’ semantic similarity. The obtained human ratings are cast as a reference point for comparing different automated similarity measures that crucially rely on various sources of information. Finally, a cognitively salient model integrating human judgments in SALAAM is proposed as a means of improving its performance on sense disambiguation for verbs in particular and other word types in general.

A dissertation on event nouns in Japanese.

An ERP study of processing complexity, cross-linguistically.

A dissertation on the representation of temporal structure in auditory perception.

Downward Entailment is a semantic property common to many linguistic expressions across natural languages (Ladusaw, 1979). This dissertation takes downward entailment as a yardstick in assessing children’s semantic competence. First, downward entailment is used as a case study for several alternative models of language acquisition, including those recently proposed by Tomasello (2000) and by Pullum and Scholz (2002). According to these researchers, children are initially conservative, and tend to (re)produce linguistic expressions that they have experienced in the input. Even at later stages, when children form generalizations, children’s linguistic generalizations are directly tied to the input, based on domain general learning mechanisms. These models are contrasted with one based on the principles and parameters of Universal Grammar. In an experimental study using the Truth Value Judgment task (Crain and Thornton, 1998), these alternative models are put to a test by investigating a phenomenon that displays a mismatch between the data available to the child and the semantic competence the child acquires, namely the interaction between downward entailment and c-command. In particular, we report the results of an experiment investigating children’s interpretation of the disjunction operator or in sentences in which that operator is c-commanded by negation, such as Winnie the Pooh will not let Eeyore eat the cookie or the cake, and in sentences in which disjunction is only preceded by negation, as in The Karate Man will give the Pooh Bear he could not lift the honey or the donut. Second, children’s knowledge of downward entailment is investigated in order to assess children’s knowledge of quantification. Beginning with Inhelder and Piaget (1964), children have been reported to have problems in interpreting sentences containing the universal quantifier every. These findings have recently been interpreted as showing that children and adults assign different semantic representations to sentences with the universal quantifier every (Philip, 1995; Drozd and van Loosbroek, 1998). A common assumption of these linguistic accounts is that children’s non-adult interpretation of sentences containing every fails to distinguish between the restrictor and the nuclear scope of the quantifier every. A Truth Value Judgment task was designed to evaluate this assumption. The findings, together with the results of previous research, show that children’s knowledge of quantification runs deeper than is anticipated either by recent linguistic accounts of children’s non-adult responses to universally quantified sentences or by input driven models of language development. Children’s adult-like knowledge of downward entailment and of the negative polarity item any stands in contrast with their non-adult interpretation of the positive polarity item some in negative sentences, e.g., The detective didn’t find some guys (see Musolino, 1998). To address this contrast, an experiment was conducted drawing upon the observation that negative statements are generally used to point out discrepancies between the facts and the listener’s expectations, and that this felicity condition was not satisfied in previous studies. The experimental findings show that children’s interpretation of indefinites in negative sentences is fully adult-like when the felicity conditions associated with negative statements are satisfied. The same picture emerges from the findings of a final experiment investigating children’s interpretation of sentences containing multiple scope bearing elements, as in Every farmer didn’t clean some animal. In sum, the experimental findings suggest that even in the domain of semantic competence, there is no reason to assume that child language differsfrom adult language in ways that would exceed the boundary conditions imposed by Universal Grammar, as maintained by the Continuity Assumption (Crain and Thornton, 1998; Pinker, 1984).

A single segment representation with dynamic features (Oneseg) explains differences between the phonologies of spoken words and signs better than current multiple segments phonological representations of signs (Multiseg). A segment is defined as the largest phonological unit where combinations of features are contrastive, but permutations and repetitions aren’t. Hayes (1993) distinguishes between static features (place, handshape) which don't reference motion, and dynamic features (direction, repetition) which do. Dynamic features are the only way that a single segment representation can sequence motion. Oneseg correctly predicts that number of repetitions is not contrastive in signs, because repetition is the result of a dynamic feature [repeat]. Multiseg incorrectly predicts that number of repetitions should be contrastive. About 50% of all spoken words repeat irregularly (unintended, hiphop); less than 1% repeat rhythmically (tutu, murmur). Non-compound signs never repeat irregularly; about 50% repeat rhythmically. Oneseg correctly predicts repetition in signs based on the probability of combinations including the feature [repeat]; Multiseg correctly predicts repetition in words based on combinations, permutations and repetition of segments. Oneseg correctly predicts that signs never have more than two underlying places. Multiseg predicts signs with any number of places. Some signs with two places allow places to occur in either order; some are ordered by constraints. Oneseg represents both without underlying sequence or redundancy, but Multiseg’s obligatory segmental sequence overgenerates or is redundant. Chapter 5 shows that inflected verbs and classifier predicates aren’t problems for Oneseg because they are predictably iconic. Predictable iconicity is the same across all sign languages, is produced by non-signers, and doesn’t always obey the phonological rules of the language. Lexically iconic elements have the reverse characteristics. Lexically iconic, but not predictably iconic, elements are part of the phonological representation. Chapter 6 proposes possible additional features and hierarchy for Oneseg and shows that the representations produced can be economically sparse by omitting redundant material. I examine the historical assimilation processes in compounds and show that Oneseg explains them.

This thesis is concerned with chains. Chains have been conventionally defined as follows: "(a1,..., an) is a chain only if, for 1 <= i < n, ai and ai+1 are nondistinct, and ai c-commands ai+1" (cf. Chomsky 1986a, Rizzi 1990, Brody 1995 and many others). As Hornstein (1998) points out, though chains were originally not real grammatical objects but mere notation to track the history of movement, they were promoted to a legitimate syntactic tool in no time. This thesis questions chains: Are chains really necessary? What would we lose if we did not have chains as a primitive of the theory of Universal Grammar (=UG)? This thesis especially investigates what we would gain if UG were free from the notion of the chains. The claim of the thesis is that there are good empirical reasons for eliminating chains.