RELATE
Team Members Heading link
- Fatimah Ihmud
- John Mcdonald
- Aleksa Stojsic
- Dominik Szarek
Project Description Heading link
Many start-ups and established companies are trying to automate very discrete legal and accounting tasks in an effort to address the dramatic inflation in accountant and lawyer billing rates. For example, Don’t Pay is automating subscription cancellation and civil fine disputation. Blue J Legal is using convolutional neural nets to develop predictive law software that predicts the outcome of discrete legal issues. Relativity has been very successful in automating e-discovery (i.e., using computers to quickly scan and sort millions of pages of documents and automatically prepare responses to document discovery requests issued by opposing counsel or an investigating agency). RELATE is a live, web-based tool that automates just one narrow and specific, but very time-consuming, legal task. The specific task we chose was to determine the extent to which two (or more) persons are ‘related’ through various attribution provisions of the U.S. Internal Revenue Code. The ‘relatedness’ of two persons can dramatically impact the tax consequences of a transaction executed between those persons. Specifically, RELATE reads the user’s ownership data from an .xlsx file; stores the data in a series of unique graph data structures of our design; and traverses the graph(s) using proprietary algorithms we developed. In this way, RELATE converts a ‘legal’ problem into a ‘graph traversal’ problem that can more quickly and accurately be solved by computers.