Open Data, Networks and the Law
summary
Court Listener supplies about 3 million opinions and ~30 million citations; a SCOTUS subnetwork of 30,000 SCOTUS cases and 250,000 citations drives the analysis. Python notebooks use networkx and igraph (igraph is C-backed and much faster). PageRank reliably ranks older precedents (PageRank favors older cases), closeness highlights procedural connectors, and a big outlier (Detroit Timber Lumber) proves a citation-parsing error from reporter syllabi (Detroit Lumber data error). Work proceeds on time-aware growth models, goodness-of-fit for network growth, and integrating topic models with citation structure for recommendations.
AI-written summary · about
chapters (11)
- 0:00Law and legal data
- 3:33Free Law Project data
- 7:49Network tools for legal data
- 9:13Measuring case importance
- 13:11Degree centrality in citations
- 15:43Closest centrality in cases
- 18:14PageRank for legal cases
- 23:32Other centrality measures
- 27:04Community detection in law
- 30:13Summary and future research
- 32:41Q&A on predictions
related talks
41mExploring Patent Data with Python
40mDiscovering Emerging Open Source Communities Through...
36mAutomatic Citation generation with Natural Language Processing
02mDBTB INT Daniel Lewis r
21mCase Law and ML on Spark
19mBig data in the service of reliable news
41mDemystifying Social Network Analysis (SNA)
48mStreet Fighting Trend Research