Browse Prior Art Database

Learning- Based Predictive Preservation

IP.com Disclosure Number: IPCOM000247367D
Publication Date: 2016-Aug-29
Document File: 3 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method for predicting the criteria for preservation of potential evidence and making it possible to quickly preserve data based on machine learning from past legal incidents and current legal matter characteristics.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Learning- Based Predictive Preservation

Litigation and regulation require that evidence be managed in a proactive and systematic manner, including the ability to produce such information on demand for litigation. Determining what data to preserve is a complex process. Lack of timely preservation may lead to destruction of evidence.

The novel contribution is a method for predicting the criteria for preservation of potential evidence and making it possible to quickly preserve data based on learning from past legal incidents and current legal matter characteristics.

The knowledge base created while identifying evidence in past matters can be highly applicable and reusable in the context of future matters. The core idea is to have a cognitive machine learning process evaluate the existing matter corpus and learn terms and phrases used in legal context. The system then applies the learning to define possible search terms relevant to a current legal incident, and then create the request that can be executed to preserve evidence.

This automated process can dramatically decrease the time to create such requests (and preserve data), reducing risk (i.e., of the destruction of evidence). The machine learning and predictive process helps in reducing errors by defining terms, which is

humanly difficult when the enterprise is highly litigated and the volume of legal incidents

is high.

Figure 1: High-level process flow

An example implementation is for a legal incident related to harassment. The keyword is "harassment'. When a legal matter is created, the matter name, description, comments, or an attached documentation on the details of the matter most likely has

1


Page 02 of 3

the word "harassment". If the cognitive engine is made to learn that "harassment" is a relevant keyword, then the system can automatically generate the preservation request and have the evidence preserved within the enterprise, without the legal team having to

go through the manual processes o...