Replication Files for Baumgärtner, Zahner (2025)

Replication Files for ------------------------------------------------------------------------------------------------------------------------------ ------- Whatever it takes to understand a central banker - Embedding their words using neural networks. ----------- ------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------- Contents and Replication Instructions This repository contains the replication code and dataset for the study. The current version reflects the code and data as of March 13, 2025. We maintain an updated GitHub repository at the following link, where we implement improvements, fix potential issues, and address code optimizations: https://github.com/martinbaumgaertner/whatever_it_takes This replication package includes: (i) detailed ReadMe file with clear, step-by-step instructions for replication. (ii) a mapping of scripts to outputs, specifying which scripts generate each figure and table in the paper. (iii) All necessary datasets and source code to fully replicate the analysis, including embedding generation, model training, and final application. --------------------------------------------------- Code Annotations and System Requirements The code includes extensive comments for ease of understanding, along with specific guidance and alternative approaches for users without access to high-performance computing (HPC) resources. The replication process has been successfully executed on modern hardware (e.g., a recent MacBook Pro) within approximately a few dozen hours. However, users with less powerful systems may encounter performance limitations. To address this, we have explicitly annotated hardware-related constraints in the code and provided shortcuts, such as the use of pre-trained models, to facilitate replication. Some larger shortcuts (e.g., those used in Section 3.3 for measuring “rhetorical stability”) are not included in this replication package due to file size constraints (several GBs). However, these files are available upon request—please feel free to contact us if needed.《不惜一切代价理解央行官员——利用神经网络对其言论进行嵌入表征》复现文件 ------------------------------------------------------------------------------------------------------------------------------ ------- 不惜一切代价理解央行官员——利用神经网络对其言论进行嵌入表征 ----------- ------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------- 内容与复现操作指南 本仓库收录本研究的复现代码与配套数据集，当前版本对应2025年3月13日的代码与数据快照。本项目的更新版GitHub仓库可通过以下链接获取，我们将在该仓库中发布功能迭代、问题修复及代码优化后的版本： https://github.com/martinbaumgaertner/whatever_it_takes 本复现包包含以下内容：(i) 详尽的ReadMe文档，内含清晰的分步复现操作指南；(ii) 脚本与输出映射表，明确标注论文中各图表对应的生成脚本；(iii) 完整复现分析所需的全部数据集与源代码，涵盖嵌入表征生成、模型训练及最终应用全流程。 --------------------------------------------------- 代码注释与系统要求 本代码附带大量注释以提升可读性，同时为无高性能计算（High-Performance Computing, HPC）资源的用户提供了专属操作指引与替代实现方案。本复现流程已在现代硬件环境（如新款MacBook Pro）中成功运行，总耗时约数十小时。不过，硬件配置较低的用户可能会遇到性能瓶颈。针对该问题，我们已在代码中明确标注了硬件相关的约束条件，并提供了简化复现的快捷方案，例如调用预训练模型。受限于文件体积（部分文件可达数GB），本复现包未包含部分较大规模的快捷方案（如论文3.3节中用于测量"修辞稳定性"的相关方案）。如有需求，可随时联系我们获取相关文件。