LLMs extract intertextuality in classical Chinese histories with expert validation

Researchers have developed a novel agentic approach using large language models (LLMs) to extract and evaluate intertextuality in classical Chinese histories. This method goes beyond simple similarity scores by grounding reuse in exact text spans and categorizing it across five dimensions: form, aspect, source-marking, function, and stance. When applied to the Analects and the Book of Han, the LLM-based extractor achieved a precision of 56%-93%, with expert adjudication revealing a gradient in reliability based on whether reuse required inference of intent. Scaling this validated extractor to the full Twenty-Four Histories uncovered corpus-level structures and demonstrated a consistent interpretive composition of citation over eighteen centuries, despite a trend towards less literal quoting. AI

IMPACT This research demonstrates a novel application of LLMs for nuanced textual analysis, potentially advancing digital humanities and historical research methods.

RANK_REASON Academic paper detailing a new computational method for analyzing historical texts. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• Analects
• arXiv
• Book of Han
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• ScienceCast
• scite Smart Citations
• Twenty-Four Histories

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →