This study investigates long-term cultural change in Japan by analyzing emotional shifts in enka song lyrics from 1945 to 2020. Using a corpus of 2,500 songs, I tested two hypotheses: (1) that lyrics have become more emotionally negative and intense over time, and (2) that expressions of mixed emotions within songs have declined. In Study 1, I applied the NRC-VAD lexicon to measure valence, arousal, and dominance in common nouns. Results show flat valence scores but significant increases in arousal and dominance, suggesting that lyrics have become more emotionally intense but not more negative over time. In Study 2, a large language model assigned sentiment scores to over 43,000 lyric chunks. Contrary to expectations, average sentiment became more positive, and mixed emotions remained stable. These findings suggest that emotional ambivalence remains culturally accepted, even as emotional expression becomes more intense. Peaks and dips in sentiment scores align with periods of national unrest, suggesting that lyrics are sensitive to shifting social conditions. I conclude by highlighting parallels with research on language change, suggesting that both linguistic and cultural shifts may be driven by similar top-down and bottom-up processes.

The purpose of this study was to examine the differences in lexical bundle use between the different registers of speaking and writing in a learner corpus. While various studies have examined multi-word items by foreign language learners, there is currently a lack of research on lexical bundle use in spoken corpora, particularly monologues. A comparative register analysis was designed to compare the lexical bundle output of L1 Japanese students of English in two corpora of argumentative written essays and spoken monologues, with two native speaker corpora being compared for reference. 4-word lexical bundles were extracted and analyzed through a Generalized Linear Mixed Model (GLMM) and categorized into two functional and structural taxonomies. Findings suggest bundle frequency decreases as proficiency level increases. Compositionally, it was found that both registers are structurally and functionally similar to each other, though the Japanese spoken monologue sub-corpus was the least varied. Furthermore, the use of prompt-specific language was found to be more influenced by register than L1 background, with both groups producing a significant amount of language that was dependent on the prompt. The register descriptions found in this study can inform pedagogical approaches to writing and monologue tasks.

The purpose of this study is to explore the educational applications of the Corpus of Contemporary American English (COCA) to enhance English communicative competence in alignment with the 2022 revised English curriculum in Korea. This paper analyzes COCA's primary functions-List, Chart, Word, Collocates, Compare, and KWIC-and proposes pedagogical methods for their use in the classroom. The study reveals three major findings. Firstly, COCA's extensive, genre-balanced data provides authentic language contexts for learning vocabulary and grammar. Secondly, its analytical tools enable learners to understand lexico-grammatical patterns, collocations, and genre-specific language variations, thereby improving their writing and reading skills. Finally, while COCA is a powerful resource, its limitations include a focus on American English and access restrictions for free users. Web-based corpora like COCA can foster a new paradigm of data-driven, self-directed learning. This approach effectively supports the goals of the 2022 curriculum by equipping learners with practical English skills and critical thinking abilities, making it a valuable supplementary tool for educators.