Machine Learning applied to judicial decisions
summary
A platform called Dr. Luzia processes Brazilian tax execution cases by scraping open court data, using regexes and jurist-labeled examples to build a training set, starting with bag-of-words + SVM baselines and moving to embeddings + deconvolution + CNN+LSTM models. It targets five repetitive procedural steps that cover 60–80% of work, cuts a 4-person five-day workload to under two minutes, yields 99.5–99.6% accuracy on ~2,000 labeled publications, and produces petitions automatically for attorneys.
AI-written summary · about
chapters (5)
related talks
01h 24mThe use of AI by the Judiciary in LatAm Countries Hallucination or Opportunity?
33mDeep research at scale
40mJudgement Day: Benchmarking "Black Box" LLMs With Open Legal Datasets
44mExtracting Structured Data from Legal Documents
46mPath of Data & AI Challenges
22mImplementing a Text Classifier based on Lucene & LIBSVM
21mCase Law and ML on Spark
08mLeveraging data insights to increase efficiency and innovation