Reduced Gun Violence Frame Corpus data set for the Text2Story 2024 article: "Evaluating the Ability of Computationally Extracted Narrative Maps to Encode Media Framing"

Title: Simplified Gun Violence Frame Corpus (GVFC) Subset Description:This data set is a simplified subset of the Gun Violence Frame Corpus (GVFC) from Liu et al. (2019). The original GVFC consists of 1300 news articles in English from multiple U.S. based sources extracted during the year 2018, focusing on media frames commonly used when reporting the issue of Gun Violence. The original data set has 9 types of frames, including both issue-specific and generic frames. Due to high computational costs in our analysis methods, we decreased the data set size from 1300 articles to 131 articles using stratified sampling, maintaining the original distribution of the frame labels. We also manually searched for the original sources of each article based on its headline and added the missing temporal information and news source to the data set, as it was required by our algorithms. To further reduce the complexity of the framing model and account for the smaller data set size, we grouped the original nine frames into three higher-level frames: 1. Frame 1: Political Issues - Combining the first, second, and third frames, which focus on political issues mostly related to gun control.2. Frame 2: Public Services - Combining the fourth and fifth frames, which focus on mental healthcare issues, as well as school and public safety.3. Frame 3: Cultural and Societal Issues - Combining the last four frames, which are oriented towards cultural or societal issues, including discussions around race and ethnicity, public opinion, and economic consequences. The resulting simplified data set contains 131 news articles, each labeled with one of the three higher-level frames, along with the necessary temporal information and news source for the narrative extraction process. If you use this data set, please make sure to cite the original GVFC paper and our workshop paper please. References: Liu et al. (2019) "Detecting Frames in News Headlines and Its Application to Analyzing News Framing Trends Surrounding US Gun Violence", 23rd Conference on Computational Natural Language Learning (CoNLL 2019). Concha Macías, Sebastián and Keith Norambuena, Brian (2024). "Evaluating the Ability of Computationally Extracted Narrative Maps to Encode Media Framing", Text2Story 2024 Workshop, ECIR 2024.标题：简化版枪支暴力框架语料库（Gun Violence Frame Corpus, GVFC）子集 描述：本数据集为Liu等人（2019）提出的枪支暴力框架语料库（Gun Violence Frame Corpus, GVFC）的简化子集。原始GVFC包含2018年期间采集自美国多家媒体来源的1300篇英文新闻文章，聚焦报道枪支暴力议题时常用的媒体框架。原始数据集共包含9类框架，涵盖议题专属框架与通用框架两类。由于本研究采用的分析方法计算成本较高，我们通过分层抽样将数据集规模从1300篇缩减至131篇，同时保留了框架标签的原始分布比例。此外，我们根据每篇文章的标题手动检索其原始来源，并补充了算法所需的缺失时间信息与新闻来源字段。 为进一步降低框架建模的复杂度并适配较小的数据集规模，我们将原始9类框架整合为3个高阶框架： 1. 框架1：政治议题——合并原第1、2、3类框架，聚焦与枪支管控高度相关的政治议题。 2. 框架2：公共服务——合并原第4、5类框架，聚焦精神卫生保健、学校与公共安全相关议题。 3. 框架3：文化与社会议题——合并原最后4类框架，围绕文化或社会议题展开，涵盖种族与族裔讨论、公众舆论及经济后果等内容。 最终生成的简化数据集包含131篇新闻文章，每篇均标注有上述3类高阶框架中的一类，并附带叙事提取流程所需的完整时间信息与新闻来源信息。 若使用本数据集，请务必引用原始GVFC相关论文与本团队的研讨会论文。 参考文献： Liu等人（2019）《检测新闻标题中的框架及其在分析美国枪支暴力相关媒体框架趋势中的应用》，第23届计算自然语言学习会议（Conference on Computational Natural Language Learning, CoNLL 2019）。 Concha Macías, Sebastián 与 Keith Norambuena, Brian（2024）。《评估计算提取的叙事图谱编码媒体框架的能力》，Text2Story 2024研讨会，欧洲信息检索会议（European Conference on Information Retrieval, ECIR 2024）。