Artificial Intelligence in the Public Sector Database

This Final_Deliverable.xlsx dataset contains the metadata of a sample of N=1,923 academic texts related to the use of Artificial Intelligence in the Public Sector. Its contents are split into the tabs below:

Datamap: provides a complete list of metadata variables including their codes, definitions and where they were pulled from (i.e. OpenAlex, BERTopic, or Manually Created).

Full_Results: contains full metadata of all N=1,923 academic texts.

Topic_Counts_Framework_Assign: contains variables that illustrate how we renamed and classified / integrated BERTopic produced topics into our functional framework categories (based identifying representative documents and qualitative review).

Intertopic_map: corresponds to Table 2 in the manuscript.

Cosine_Similarity_Matrix: shows pairwise cosine similarity scores for all 22 topics generated by BERTopic. A variation of this is included in Figure 6 within the manuscript.

Hyperparameter_Combos: records the 720 hyper parameter combinations we tested while fitting our BERTopic model, organized in descending order based on variable 'coherence_score'. Our final hyperparamter combination is in Record 2.

本最终交付物.xlsx数据集收录了N=1,923篇与公共部门人工智能应用相关的学术文本的元数据。其内容分为以下标签页：

数据映射：提供包括代码、定义及其来源（即OpenAlex、BERTopic或人工创建）在内的完整元数据变量列表。

全面结果：包含所有N=1,923篇学术文本的完整元数据。

主题计数框架分配：包含展示我们将BERTopic生成的主题重命名、分类/整合至功能框架类别中的变量（基于识别代表性文档和定性审查）。

跨主题映射：对应于手稿中的表2。

余弦相似度矩阵：显示由BERTopic生成的所有22个主题的成对余弦相似度得分。手稿中的图6包含了这一矩阵的变体。

超参数组合：记录了我们调整BERTopic模型时测试的720种超参数组合，按变量'coherence_score'降序排列。我们的最终超参数组合位于第2条记录。