전문성에 따른 공간·환경 디자인 어휘 및 품사 양상 연구 -우수디자인(GD) 상품선정과 대한민국디자인전람회 수상작을 중심으로-

(Background and Purpose) Korean, as an agglutinative language, has led much domestic text-mining research to skew toward noun-centric analysis. Prior studies in spatial design show the same noun bias and often do not differentiate by design expertise or report groups separately, making it difficult to estimate expertise effects. To address these limitations, this study combines noun, verb, and adjective indices—the open class parts of speech(POS)—to quantify, by expertise level, differences in lexical choice, POS usage ratios, and lexical diversity, and proposes a reference for auditing narrative strategy bias in spatial design proposal discourse. (Method) We analyze 140 award citations (2019-2024) in the Space·Environmental Design category from two competitions run by the same institution: Good Design (GD) Product Selection and the Korea Design Exhibition. After preprocessing with the Kiwi morphological analyzer, we lemmatized tokens, excluded proper and dependent nouns and auxiliary predicates, and applied a domain lexicon to normalize compound nouns, light verb constructions, and adjectival derivations to their base part of speech lemmas, mitigating noun heavy tagging. We then derived group top terms via mean document level TF IDF, identified group biased terms using log odds with an informative Dirichlet prior, compared document mean POS proportions, and estimated POS specific lexical diversity with MATTR and MTLD. (Results) The results are as follows. First, TF-IDF-based common top terms included the nouns ‘공간·문화·속’, the verbs‘되다·위하다·있다·통하다·하다’, and the adjectives‘가능하다·같다·높다·다양하다·새롭다·쉽다’. Second, in the bias test, Good Design showed group-biased terms such as‘자연·계획하다·적용하다’, while the Korea Design Exhibition showed‘지역·제안하다’; no biased adjectives were detected. Third, in document-mean POS ratios, Good Design had a significantly higher noun share and the Korea Design Exhibition a higher verb share; no group difference was observed for adjectives. Fourth, for POS-specific lexical diversity, Good Design was slightly higher only in nouns’ local diversity (MATTR), with a very small effect size; no group differences appeared in document-level sustained diversity (MTLD) or in verb/adjective diversity. (Conclusions) The conclusions are as follows. First, experts favor attribute-defining nouns and execution/implementation verbs, whereas students favor context/scope nouns and proposal verbs—patterns consistent with the submission purposes of the competitions. Second, experts show a relatively higher noun share and students a relatively higher verb share; adjectives do not differ across groups, which may reflect lexical convergence or limited power due to sample size. Third, experts show marginally higher local diversity for nouns, but overall document-level diversity is broadly similar regardless of expertise. This study quantifies POS level group differences beyond nouns. Limitations include the modest corpus (140 documents) and the exclusion of other POS, which precluded inter POS narrative analysis. Future work with larger, all POS corpora will further clarify spatial design discourse.