This study aims to identify characteristic lexical bundles in spoken discourse produced by speakers of the Gyeongsang and Jeolla dialects and to clarify the mechanisms that generate their distinctive features in contrast to other dialects. Noting that...
This study aims to identify characteristic lexical bundles in spoken discourse produced by speakers of the Gyeongsang and Jeolla dialects and to clarify the mechanisms that generate their distinctive features in contrast to other dialects. Noting that dialectal uniqueness is revealed not only in the forms of individual words but also in the combinational patterns of words that are identical in form to Standard Korean, this study seeks to move beyond the limitations of previous dialect research, which has focused primarily on the standard-nonmatching type, by systematically analyzing atypical combinational patterns among standard-matching lexical items.
Methodologically, the study adopts key concepts and techniques from corpus linguistics—lexical bundles, N-gram analysis, and keyness analysis. Morphologically annotated corpora were constructed using the AI-Hub Gyeongsang and Jeolla dialect corpora (approximately 1.3 million words each), and the Seoul–Gyeonggi dialect corpus (approximately 4.17 million words) was set as a reference corpus to extract statistically significant lexical bundles. To restrict the analysis to the standard-matching type, the data were preprocessed to identify and separate the standard-nonmatching type and dialect-specific forms. Variant forms of particles and endings were also normalized to representative forms to improve analytic accuracy.
Chapter 1 presents the necessity and objectives of the study. It confirms the importance of multiword units in language production and comprehension and emphasizes the need to focus on dialect lexical bundles of the standard-matching type, alongside those of the standard-nonmatching type. A review of prior research traces international trends in lexical-bundle studies and their reception in Korea, demonstrating that lexical-bundle research on dialects has been rarely conducted. The chapter also outlines the research procedures in detail, from building a morphologically annotated corpus to selecting and classifying lexical bundles.
Chapter 2 establishes the theoretical background. Beginning with Sinclair’s notion of the “lexical item,” it examines Biber’s concept of lexical bundles and their properties, confirming that N-gram analysis is a core extraction method. The chapter then reviews definitions of dialect and classification systems in dialectology, positioning the present work within contrastive dialect studies. It further argues that earlier dialect studies have largely addressed multiword units in terms of individual expressions or constructions of the standard-nonmatching type, leaving a gap in frequency-based, objective analyses of atypical combinational patterns among standard-matching items.
Chapter 3 describes the extraction procedure for dialect lexical bundles. First, a morphological analyzer was applied to the target corpora, and manual corrections were made to improve accuracy in light of the characteristics of dialect and spoken data. Variant forms of particles and endings were normalized into representative forms. Using the Urimalsaem API, items were tagged so that forms of the standard-nonmatching type and dialect-specific forms could be identified and separated, thereby ensuring that the subsequent analysis targeted only the standard-matching type. Candidate lists were extracted through frequency-based N-gram analysis, after which keyness was calculated using the Seoul–Gyeonggi corpus as a reference and the log-likelihood (LL) test. As a result, 78 characteristic lexical bundles were identified for the Gyeongsang dialect and 89 for the Jeolla dialect. An examination focusing on 2-grams shows that the Gyeongsang dialect is characterized by frequent use of plain-style sentence-final endings, omission of the quotative particle ‘-go’, and lexicalized “address term + vocative particle” patterns, whereas the Jeolla dialect exhibits a high frequency of interrogative endings in the hage and plain styles, preferential use of certain lexical items, and adverbial formations created through noun–particle combinations.
Chapter 4 inductively categorizes the mechanisms that give rise to dialectal distinctiveness into five types—omission, addition, substitution, preference, and specialization—and analyzes their realizations in each dialect. In the Gyeongsang dialect, findings include omission of the quotative particle go (e.g., -ra ha-), habitual addition of vocative particles (e.g., eonni-ya), lexical substitution involving auxiliary verbs such as ssahda or bound nouns such as jeok and dwi, a preference for short-form negation (e.g., -eoya an doeget-), and specialized constructions such as the past adnominal form -eotneun. In the Jeolla dialect, patterns include omission resulting in the use of gatgo without a main verb, additions such as jeil-lo and mwo-i-nya, lexical substitutions involving items such as jane, geu-nom, and sori, a preference for full forms over contractions (e.g., -ji an ha-), and specialized constructions such as -eo gatgo, haneun mari, and mwun nom-ui. These results demonstrate that inter-dialectal differences are systematically realized through the five mechanisms.
The contributions of this study are as follows. First, it expands dialect research methodologically by applying corpus-based, objective analytical approaches in a systematic manner. Second, by empirically demonstrating atypical combinational patterns among standard-matching lexical items, it shows that dialectal uniqueness can emerge at the level of syntactic combination as well. Third, it proposes an analytical framework by systematizing the mechanisms that generate differences across dialects. Fourth, it provides foundational resources for dialect language-resource development and for facilitating inter-dialect communication.
Nevertheless, this study has several limitations. First, due to frequent errors in morphological analysis inherent to spoken dialect corpora, the study could not conduct a full analysis targeting lexical bundles of the standard-nonmatching type or those composed of dialect-specific forms. Second, limitations in corpus size restricted the ability to secure a quantitatively sufficient inventory of lexical bundles and to generalize the findings. Future research should expand the comparative scope by including additional regional dialects and extend the analysis to lexical bundles of the standard-nonmatching type through the development of dialect-specialized morphological analyzers.