This dataset contains 4176 non-empty official public EU legal judgments that were finalized between 2008 and 2018, categorized in one or more subject matters, that fall within the case-law sector and the Court of Justice. In the dataset, we can find 133 distinct subject matters. In order to build the set of citation-based embeddings, we adopted a custom strategy to extract citations from the dataset, since they were not available as structured data. In particular, we reached EUR-Lex to identify common rules adopted for citations in the legal judgments of this specific dataset. Following their indications, we pre-processed the set of judgments J by: i) lowercasing the text, ii) removing punctuation except for the forward slash and the parenthesis (commonly used in citations), and iii) removing stop words except for the word of (commonly used in citations). Subsequently, we designed custom regular expressions to extract citations towards Directives, Decisions, and Regulations, following the numbering rules for articles and sub-levels.