A Corpus-driven Study Of Phraseological Sequences In English Academic Texts

This thesis is a systematic corpus-driven study of phraseological sequences (henceforth PSs) in English academic texts, with a view to characterizing their most prominent formal and functional features. Theoretically set in Firthian linguistics, this study is an elaboration of the Firthian contextual theory of meaning, the Sinclairian extended units of meaning, and the Hallidayan metafunctions of language. A new computing method is developed for extracting contiguous PSs by measuring their internal associations. NEW-JDEST corpus serves as the testbed. The new method involves three main elements:(1) Further developing the concept of pseudo-bigram transformation (Silva & Lopes 1999) according to the 1st-order Markov model. In this way we may extend the use of current statistics-based measures, which are proposed for extracting bigrams, to the computing of n-grams, where n≥2.(2) Constructing a new normalizing algorithm of probability-weighted average for refining the current statistics-based measures, enhancing the precision and recall of PSs extracted by these measures.(3) Identifying the size of each PS by integrating the methods of"frequency threshold"and"local maxima".This research evaluates the new computing method by comparing its effectiveness in identifying PSs with that of other known measures and software (e.g. Traditional entropy, Wordsmith Tools 4.0). Data of the sampling test suggest that the new method enhances the extraction precision to a greater extent (79.8%). The extracted data thus better reflect the semantic and structural characteristics of PSs.Within Halliday's framework of the metafunctions of language, this study generalizes a three-dimensional model of discourse functions for the extracted PSs: experiential, stance, and organizational. The three prime functions are not complementary but are used to describe a PS from three parallel perspectives.Experiential PSs form the largest category of phraseology in academic texts. PSs of this kind are the main conveyor of information, characterizing a high density of information in academic texts. Structurally, the majority of these PSs do not constitute full clauses but sequences of one or more (single or multiple) clause constituents. Semantically, the most common experiential PSs are assigned to five main categories: specifying processes and actions; identifying entities, notions and activities; specifying attributes; expressing time and space; and expressing vagueness.Lexico-grammatical marking of stance is distinguished between phrasal-level and clause-level PSs. The analysis of clause-level PSs is based on the differentiation of three kinds of structure: overt-subject, it-extraposition and null-subject. Within the systemic framework, stance it-extraposition is characterized by expressing four types of modality: epistemic, i.e. the expression of some degree of possibility and certainty; deontic, i.e. the expression of obligation and inclination; dynamic, i.e. the expression of potentiality and ability; and evaluative, i.e. the expression of attitude, opinion and evaluation. Frequent occurrences of it-extraposed PSs reflect that although researchers try not to use subjective and affective locutions overtly in academic texts where objectivity and scientificness are essentially required, they still manage to reveal their attitudes and stands through somewhat covert means.Organizational PSs are described at three levels of discourse functions: discourse acts, meta-discourse, and general logical marking. Drawing on corpus evidence, we investigate six main types of discourse acts: focusing, presenting views or facts, presenting results, reporting, exemplifying, and text deixis. We also discuss recurrent meta-discourse PSs for"outlining purposes"or"announcing present research"(Step 1 of Move 3 in the CARS model), for the purpose of characterizing their most common patterns and functions in Introduction and Abstract of research articles. NEW-JDEST data demonstrate that both discourse acts and meta-discourse have their own phraseological patternings in academic texts. Some PSs are so frequently used to perform concrete discourse acts or meta-discourse that in a sense, the most typical of them become signals of the concrete functions.All the data and discussion in this study suggest that co-selection is at the very heart of choosing language forms to realize experiential, stance, and organizational meanings in academic text production. Conventionality prevails into every aspect of language use and co-selection is going on at multiple levels. Specifically, NEW-JDSET corpus evidence points to four types of co-selection in academic texts: the co-selection of lexis and lexis, the co-selection of lexis and grammar, the co-selection of PSs and research topics, and the co-selection of PSs and discourse structures. These co-selective relationships have provided strong indications that lexis, grammar and meaning are essentially an integrated whole, a fact which has not received due attention in current language descriptions, as well as in linguistic theorizing.The present study may yield unique insights and implications in the following aspects: the construction of the new extraction method, the re-identification of phraseologies, the elaboration of co-selection at multiple levels, the remodeling of traditional linguistic description, the categorization of discourse functions of PSs, and the EAP teaching in China.