Figure 1

Representing judicial decisions as bipartite networks. (A) For each decision in the corpus we consider the words used (blue), the cited legislation (red), and metadata such as the date of the ruling and the court (green). We then create two bipartite networks: (B) A network of decisions and words; (C) A network of decisions and cited legislation. We select relevant words through a process of n-gram extraction and removal of words with low informational content, whereas we obtain cited legislation directly from the data (Section 2)