Sentiment of keywords from annual reports of Slovak commercial bank

The dataset represents extracted keyword bigrams from documents published by commercial banks for the years 2007-2022. Lemmatization was applied to the extracted keywords and subsequently the sentiment (positive, negative, neutral) of these keywords was determined. The documents are divided into the categories of Pillar 3 Disclosure, which are documents that are required to be disclosed under Pillar 3. The second category is Pillar 3 Related, which are documents that are not directly required to be disclosed, but are related to mandatory disclosure. In the case of subcategories, we designated financial reports published at the level of a year and half a year with the summary category financial reports. The remaining subcategories match the name (type) of the document. The NAN value in the cells represents that the observed keyword did not appear in the observed document in that year.本数据集提取自2007年至2022年间商业银行发布的文档中的关键词二元组（keyword bigrams）。研究人员首先对提取得到的关键词执行词形还原（lemmatization）处理，随后确定了这些关键词的情感极性（positive、negative、neutral，即正、负、中性）。上述文档分为两大类别：其一为第三支柱披露文档（Pillar 3 Disclosure），即依据第三支柱要求需强制披露的文档；其二为第三支柱相关文档（Pillar 3 Related），即虽未被强制披露要求直接涵盖，但与强制披露内容相关的文档。在子类别划分上，我们将按年度及半年度发布的财务报告归类为汇总类财务报告（summary category financial reports），其余子类别直接沿用文档的官方名称（类型）。单元格中的NAN值表示该观测关键词在对应年份的观测文档中未出现。