CSI Screenplay Summarization Corpus

该数据集包含第 1-5 季 39 个“CSI：犯罪现场调查”剧集的黄金标准摘要标签。每集都包含完整的剧本和摘要的人工注释。注释包括： 场景级二进制标签，表示场景是否属于情节摘要 属于摘要的场景的基于方面的标签，即场景涉及摘要的哪个方面（例如，关于受害者的信息、犯罪现场、肇事者等）句子级二进制标签，表示属于数据集 10 集摘要的剧本句子This dataset contains gold-standard summary annotations for 39 episodes of *CSI: Crime Scene Investigation* across Seasons 1–5. Each episode includes its full script and human annotations for its episode summary. The annotations include three categories: 1. Scene-level binary labels indicating whether a given scene falls within the episode's plot summary; 2. Aspect-based labels for scenes that are part of the summary, specifying which aspect of the summary the scene relates to (e.g., information about victims, crime scenes, perpetrators, etc.); 3. Sentence-level binary labels for script sentences that correspond to the summaries of 10 episodes included in this dataset.

搜集汇总

背景与挑战

背景概述

该数据集是CSI Screenplay Summarization Corpus，包含《CSI：犯罪现场调查》第1-5季39集的剧本及人工标注摘要，标注涵盖场景级和句子级标签，用于分析情节摘要的叙事结构。由爱丁堡大学和墨尔本大学于2020年发布，旨在支持剧本摘要研究。