Case Outcome Detection: Challenges and Methods
Friday, March 25, 2016 - 1:25pm - 2:25pm
The Case Outcome Detection task is the problem of automatically determining the outcome of a lawsuit (dismissal, jury verdict, etc.) based on electronic docket entries created and kept by the respective court. A solution to this problem has important commercial value because it allows service providers to collect valuable statistics of outcome distributions by multiple dimensions (i.e., judge, law firm, attorney, company) We will introduce the data sources for the task and describe how we model the docket progression as a linear-chain conditional random field (CRF). We will show that the CRF approach considerably increases outcome accuracy compared to prior state of the art.