邵僅、林巧敏 / Chin Shao、Chiao-Min Lin 

本研究以清代黃冊檔案中之刑事案件紀錄為分析對象，結合生成式人工智慧（Generative AI）、Python與DocuSky等數位人文工具，探討犯罪類型、犯罪動機與案件的時空分布特徵。研究採用中央研究院近代史研究所典藏之1,745筆漢文黃冊檔案，透過文本預處理與語義分析、自動化標註、犯罪類型與動機分類，並進行詞頻統計及地理資訊視覺化處理。結果顯示，殺人與傷害等暴力犯罪為最主要的案件類型，「家庭糾紛」、「感情糾紛」與「偶發衝突」為常見犯罪動機，並具有高度的時代與空間集中性；犯罪紀錄高峰期集中於康熙中後期與乾隆末期，地理分布則以山東、江南與廣東等人口稠密地區為高頻區。本研究旨在以數位人文方法分析非結構化的歷史檔案資料，探究將數位方法應用於犯罪動機語義判讀與司法歷史研究的潛力，期許可提供應用史料的不同研究取徑。

This study analyzes criminal case records found in Qing Dynasty Yellow Registers, employing a digital humanities approach that integrates generative artificial intelligence (GAI), Python programming, and the DocuSky platform. Drawing upon 1,745 Chinese-language Yellow Registers entries preserved by the Institute of Modern History, Academia Sinica, the research applies text preprocessing and semantic analysis to enable automated annotation and classification of crime types and motives, complemented by word frequency statistics and geospatial visualization. Findings indicate that violent crimes such as homicide and assault constitute the predominant categories, with “domestic disputes,” “romantic conflicts,” and “spontaneous altercations” emerging as the most frequent motives—each displaying notable temporal and spatial concentration. Peaks in criminal activity are observed during the mid-to-late Kangxi reign and the late Qianlong period, with high case density in populous regions such as Shandong, Jiangnan, and Guangdong. This study employs digital humanities approaches to analyze unstructured historical archives, exploring the potential of applying computational methods to the semantic interpretation of criminal motives and the broader field of judicial history. The research aims to provide alternative methodological perspectives for engaging with historical sources.DOI: 10.6575/JILA.202512_(107).0002

107-02邵僅_林巧敏