Statistical language modelling may not only be used to uncover the patterns which underlie the composition of utterances and texts, but also to build practical language processing technology. Contemporary language applications in automatic speech recognition, sentence interpretation and even machine translation exploit statistical models of language. Spoken dialogue systems, where a human user interacts with a machine via a speech interface in order to get information, make bookings, complaints, etc., are example of such systems which are now technologically feasible. The majority of statistical language modelling studies to date have concentrated on written text material (or read versions thereof). However, it is well-known that dialogue is significantly different from written text in its lexical content and sentence structure. Furthermore, there are expected to be significant logical, thematic and lexical connections between successive turns within a dialogue, but "turns" are not generally meaningful in written text. There is therefore a need for statistical language modeling studies to be performed on dialogue, particularly with a longer-term aim to using such models in human-machine dialogue interfaces. In this thesis, I describe the studies I have carried out on statistically modelling the dialogue material within the British National Corpus (BNC) - a very large corpus of modern British English compiled during the 1990s. This thesis presents a general introductory survey of the field of automatic speech recognition. This is followed by a general introduction to some standard techniques of statistical language modelling which will be employed later in the thesis. The structure of dialogue is discussed using some perspectives from linguistic theory, and reviews some previous approaches (not necessarily statistical) to modelling dialogue. Then a qualitative description is given of the BNC and the dialogue data within it, together with some descriptive statistics relating to it and results from constructing simple trigram language models for both dialogue and text data. The main part of the thesis describes experiments on the application of statistical language models based on word caches, word "trigger" pairs, and turn clustering to the dialogue data. Several different approaches are used for each type of model. An analysis of the strengths and weaknesses of these techniques is then presented. The results of the experiments lead to a better understanding of how statistical language modelling might be applied to dialogue for the benefit of future language technologies.

Although there has been a great deal of research into identifying the constraints responsible for the distribution of reflexive pronouns (like <i>himself, herself, themselves</i>), accounts of reflexivisation are often based entirely on syntactic factors. Furthermore, it is usually assumed that a reflexive and its antecedent (i.e. the noun phrase to which a reflexive pronoun 'refers') must be from the same sentence, and the anaphoric relationship between a reflexive and its antecedent is often established based on a complete structural analysis of the sentence containing them. In this thesis, we propose a treatment of reflexives in which an anaphoric relationship between pronoun and antecedent can be established based on information contained in partial structures associated with linguistic expressions. There is no need to obtain a complete structural analysis before performing anaphora resolution, and information obtained from anaphora resolution can be used to constrain possible analyses. In presenting an account for the distribution of locally bound reflexives in English and of long distance reflexives in Icelandic, we will see that the same general treatment will be applicable to local as well as non-local reflexive anaphoric phenomena. Our first goal is to establish the constraints that are relevant for the distribution of reflexive pronouns. Then a declarative unification-based linguistic framework will be introduced in which these constraints can be stated. Within a declarative framework, constraints can be stated independent from any processing strategy. The basic grammar structures of this framework will be partial specifications of trees, and the framework will require only a single grammar rule to combine these partial specifications. Finally, we will illustrate how various phenomena associated with reflexive pronouns can be accounted for in this framework. We will provide an account for the distribution of reflexives appearing in complement causes, picture-noun constructions, possessives, unbounded dependency constructions, prepositional phrases, and constructions where reflexives can have either 'sloppy' or 'strict' readings.

No description available.

The thesis investigates the phenomena of quantification in Mandarin Chinese. I provide an account of the word order of quantification in Mandarin Chinese by studying its syntax and syntax-external interface. I argue that the unique word order of Chinese quantification results from the obligatory requirement of two lexical items—dou ‘all’ (from mei ‘every’) and you ‘have’ (from the indefinites). I show that the syntax and semantics of the two items bring about flexibility of preverbal word order of universals and indefinites, which further leads to the absence of quantifier raising in Mandarin. On dou, I argue it is an associated adverb rather than a head of functional projection. Semantically, it is seen as a D operator and universal quantifier. It allows ‘intermediate distributivity’ and guarantees the maximality of the domain of the associated DP. Mei ‘every’ is argued to be a distributive dependant and set the value for Cover, a variable brought by dou. I show that you and indefinite is a parallel pair of mei…dou. I argue the licenser of preverbal indefinites you is a functional head occupying the Infl position and I argue against the verbal analysis. I also account for the adjacency between you and indefinite. I argue that the quantification mechanism in Mandarin provides important evidence for Huang’s argument that Chinese is more analytic than English and many other languages on the syntactic-analytic parameter (2005). I discover a new type of distributivity other than that marked by dou. It is formed with the plural subject under focus. I claim that it is a discourse-oriented distributivity and explore the intricate relation between the grouping of focus-introduced alternatives and the selection of the distributive reading under the Economy Principle.

The present study deals with the impairment of prepositions, a somewhat neglected topic in aphasia research. It is the first to investigate the availability of all types of prepositions (i.e., spatial, temporal, other meaningful, subcategorized, syntactic prepositions, and particles) in a variety of comprehension and production tasks in one anomic aphasic and four Broca’s aphasic patients and healthy speakers. While the availability of spatial, temporal, or subcategorized prepositions has been investigated, other preposition types have never been studied before. The data revealed that prepositions were impaired in the patients, and that the degree of impairment differed for different types of prepositions. Three of the main findings are: first, meaningless prepositions were not the most vulnerable subcategory of prepositions in the patients. In fact, four of the five aphasic patients performed best on (meaningless) syntactic prepositions. Second, patients made few omissions and many substitution errors which were mostly within-category (a preposition was substituted by another preposition). Third, there was no difference in the performance of Broca’s and anomic aphasic patients. These results differ from those of previous studies (e.g., Bennis et al., 1983; Friederici, 1982). They found that (i) meaningful prepositions remained relatively well preserved in Broca’s aphasia, while meaningless subcategorized and/or syntactic prepositions were very impaired, (ii) that Broca’s aphasic patients tended to omit rather than substitute prepositions, and (iii) that patients of contrasting clinical profiles performed differently. The preservation of syntactic prepositions together with the large number of within-category substitutions (which indicate sensitivity to the grammatical class of prepositions) were interpreted to suggest that the preposition deficit of the patients is not due to syntactic impairments. Rather, a post syntactic deficit in selection of the correct preposition at spell-out – a construct in modern linguistic theory that links syntax with phonology – is put forward.

The Cicipu language (Kainji, Benue-Congo) of northwest Nigeria has the kind of robust noun class system characteristic of Benue-Congo languages – GENDER agreement is found on a great many agreement targets inside and outside the noun phrase. For a number of these targets, gender agreement is in competition with a separate paradigm, that of PERSON agreement. The dissertation focuses on the distribution of this alternation with respect to subject prefixes, object enclitics, and pronouns, based on a corpus of 12,000 clauses of spoken language. The alternation proves to be complex to describe, involving a constellation of lexical, phonological, morphosyntactic, semantic and discourse-pragmatic factors. In particular, both animacy and topicality are CONDITIONS (Corbett 2006) on agreement. While inanimate or animal participants normally trigger gender agreement, if they are topics then they may trigger person agreement. Likewise while human nouns typically trigger person agreement, this is not always the case, and gender agreement is more likely if the referent is of incidental importance to the discourse. Furthermore it is argued that this alternation is sensitive to discourse topic (e.g. Dooley 2007) rather than sentence topic (e.g. Lambrecht 1994). Both gender and person subject prefixes are ambiguous agreement markers according to the typology of Bresnan and Mchombo (1987) and Siewierska (1999), since both can take part in grammatical or anaphoric agreement. Thus the Cicipu data supports Culy's (2000) contention that topicality is an independent dimension for the classification of agreement markers, rather than derivative of the grammatical vs. anaphoric agreement distinction, and leads us to re-evaluate the common assumption that dependent person markers (Siewierska 2004) cannot vary with respect to their discourse function. Since Cicipu is otherwise undescribed, a major part of the dissertation consists of a phonological and grammatical sketch.

This thesis examines the effects of contact on the grammars of the languages of two Saharan oases, Siwa and Tabelbala. These share similar linguistic ecologies in many respects, and can be regarded as among the most extreme representatives of a language contact situation ongoing for centuries across the oases of the northern Sahara. This work identifies and argues for contact effects across a wide range of core morphology and syntax, using these both to shed new light on regional history and to test claims about the limits on, and expected outcomes of, contact. While reaffirming the ubiquity of pattern copying, the results encourage an expanded understanding of the role of material borrowing in grammatical contact, showing that the borrowing of functional morphemes and of paradigmatic sets of words or phrases containing them can lead to grammatical change. More generally, it confirms the uniformitarian principle that diachronic change arises through the long-term application of processes observable in synchronic language contact situations. The similarity of the sociolinguistic situations provides a close approximation to a natural controlled experiment, allowing us to pinpoint cases where differences in the original structure of the recipient language appear to have influenced its receptivity to external influence in those aspects of structure.

No description available.

No description available.

No description available.