Legal Case Retrieval Using SVM Classifier

The legal case retrieval task involves reading a new case and extracting supporting cases for its decision from the entire case law corpus. These supporting cases are referred to as noticed cases. The proposed approach uses information retrieval models BM25, TF-IDF, Doc2Vec, Partial String Matching, and Topic Clustering through Latent Dirichlet Analysis to obtain similarity scores between query and candidate case documents. A machine learning algorithm, Support Vector Machine, uses these similarity scores as features to classify a new case as relevant or non-relevant. Our proposed approach shows promising results with high values for precision and recall.