cs.CL提交于 2026-10-27已译

LLM-SC 在活态圣遗中的适配：Pashupatinath Temple 评论的领域抽取、语义网络与人工评估

Adapting LLM-SC to Living Sacred Heritage: Domain-Specific Extraction, Semantic Networks and Human Evaluation on Pashupatinath Temple Reviews

Hari Sitaula · Manish Sakhakarmy · Pujan Pant · Sagar Newpane

中文摘要

大型语言模型(LLM)流程可将游客评论转换为结构化数据，但通用方案可能遗漏活态圣遗地的仪式、 sacred spaces、文化限制及情感交织体验。研究将 LLM-Social Computing(LLM-SC) 框架适配至尼泊尔 Pashupatinath Temple：在 GPT-4o 抽取提示中指定 12 类实体（4 类原有 LLM-SC 类型与 8 类圣遗类型）、4 种关系和 9 个方面—情感维度，从经处理的 3,939 条评论中随机固定抽取 2,500 条英文 Tripadvisor 评论。导出数据含 11,286 条实体、6,232 条方面—情感和 5,590 条关系记录，分别覆盖 2,485、2,487 和 2,433 条评论；新增的 8 类圣遗实体占实体记录的 64.8%。相对于两名标注者裁决形成的 200 条评论参考集，归一化表面形式与类型的精确匹配获得暂定实体精确率 41.0%、召回率 21.8% 和 micro-F1 28.4%。方面存在 F1 为 53.6%，方面与极性端到端 F1 为 42.6%。人工裁决支持 194 个首列关系候选中的 153 个（每条评论一个候选，78.9%）；该候选级比例不能估计关系召回率或整个图谱的精确率。成果是一项植根文化语境、面向特定领域的适配，并包含分组件人工评估及明确的可复现性与泛化局限。

关键要点

01针对活态圣遗评论中的仪式、空间、文化限制与复合情感，提出领域特定实体与关系分类。
02在 GPT-4o 抽取流程中引入 8 类圣遗实体，其占全部实体记录的 64.8%。
03实体抽取暂定精确率为 41.0%、召回率为 21.8%，micro-F1 为 28.4%。
04方面存在 F1 为 53.6%，方面加极性的端到端 F1 为 42.6%。
05人工裁决支持 78.9% 的首列关系候选，但该结果不能估计关系召回率或图谱精确率。

解读

尚无解读。

原始英文摘要

arXiv:2610.35685v1 Announce Type: new Abstract: Large language model (LLM) pipelines can turn visitor reviews into structured data, but generic schemes may miss rituals, sacred spaces, cultural restrictions and emotionally mixed experiences at living sacred heritage sites. We adapted the LLM-Social Computing (LLM-SC) framework to Pashupatinath Temple, Nepal, by specifying a 12-type entity taxonomy (four original LLM-SC types plus eight sacred-heritage types), four relation types and nine aspect-sentiment dimensions in a GPT-4o extraction prompt, which was applied to a fixed random sample of 2,500 English-language Tripadvisor reviews from a processed corpus of 3,939. The exports contain 11,286 entity, 6,232 aspect-sentiment and 5,590 relation records, covering 2,485, 2,487 and 2,433 reviews, respectively; the eight added sacred-heritage types account for 64.8% of entity records. Against an adjudicated two-annotator reference for 200 reviews, exact normalized surface-and-type matching gave provisional entity scores of 41.0% precision, 21.8% recall and 28.4% micro-F1. Aspect-presence F1 was 53.6% and end-to-end aspect-plus-polarity F1 was 42.6%. Human adjudication supported 153 of 194 first-listed relation candidates (one per review; 78.9%), a candidate-level rate that does not estimate relation recall or whole-graph precision. The contribution is a culturally grounded, domain-specific adaptation with component-wise human evaluation and explicit reproducibility and generalization limitations.