CROCorp: Corpus of Parliamentary Debates in Croatia

The repository contains a cleaned and pre-processed corpus of parliamentary debates from the Croatian Parliament (Sabor). The corpus is accompanied by the metadata on elected representatives and their political parties. It covers the period of 2003-2020 (five complete terms) and counts over 500 thousand speeches. <strong>If you use the dataset, please cite</strong>: Mochtak, Michal, Josip Glaurdić, and Christophe Lesschaeve (2022): CROCorp: Corpus of Parliamentary Debates in Croatia (v1.1.1), https://doi.org/10.5281/zenodo.6521372. v1.1.1 (<strong>latest version</strong>)<br> - added the concept DOI to codebooks (DOI was generated only after the repository was published) v1.1.0<br> - improved coding of dummy variable "moderator" (using less error-prone alghoritm for detecting the modertor role)<br> - fixed issue with agenda points which are conncatenated while preserving a unique web link<br> - recoded agenda points tags using better ML model (transformer architecture) v1.0.0<br> - originally posted on GESIS repository (migrated to ZENODO due to limitations concerning the concept DOI)本数据集仓库收录了经清洗与预处理的克罗地亚议会（Sabor）议事辩论语料库。该语料库附带当选议员及其所属政党的元数据，时间跨度覆盖2003年至2020年的五届完整议会任期，总计收录超50万条发言。 <strong>若使用本数据集，请按以下方式引用</strong>： Mochtak, Michal、Josip Glaurdić与Christophe Lesschaeve（2022）：《CROCorp：克罗地亚议会辩论语料库（v1.1.1版）》，https://doi.org/10.5281/zenodo.6521372。 v1.1.1（<strong>最新版本</strong>） - 向编码手册中新增DOI（数字对象标识符）字段（DOI仅在仓库发布后生成）v1.1.0 - 优化虚拟变量moderator（会议主持人）的编码规则（采用错误率更低的算法识别主持人角色） - 修复了议程条目拼接时保留唯一网页链接的相关问题 - 采用性能更优的机器学习模型（Transformer架构）重新编码议程条目标签v1.0.0 - 最初发布于GESIS仓库（因DOI相关功能限制迁移至ZENODO）