To facilitate the use of BCCWJ for writing style studies, we proposed a feature index system that characterizes the writing styles of the book samples in the corpus and then, following the proposed system, annotated all of the 10,551 samples included in its library subcorpus. For the samples with a simple text structure (84%), we chose five axes (specificity, objectivity, formality, softness, and spokenness) and assigned a five-dimensional index to each sample based on a subjective assessment. For the remaining samples (16%), that is, the samples with a complex text structure or some specific format, we employed a different set of feature annotations. This approach allowed a systematic analysis of the diverse writing styles of the samples included in the library subcorpus. Statistics such as the number of samples with a specific style feature, and correlations between the styles and NDC (Nippon Decimal Classification) categories were obtained.
〈共同研究プロジェクト紹介〉萌芽・発掘型 : テキストの多様性を捉える分類指標の策定 書籍サンプルの文体を分類する
review040106.pdf
(1.27MB)
概要
