Improving operational efficiency in Law Enforcement using NLP

The Client

The client is the primary law enforcement agency of the state UP (India), with a population of over 200 million citizens.  

The Problem

The client is looking to optimize the process of reporting, investigating and resolving incidents reported by members of the public. To this end, they were looking to automate the task of a law enforcement officer who previously had to review each incident report manually to ensure its accuracy. 

Key Challenges

The team had to learn Natural Language Processing (a subset of AI)
Loss of data was not an option
The data had to be processed and treated for missing values before any analysis could be done or models applied
The data was stored in a secure environment and the algorithms had to be run in the same environment

How ElementalAI Helped 

The team designed and implemented a model that identified criminal codes applicable on a report by using topic extraction – a technique that groups key text and phrases based on their frequency of occurrence.  

This problem was a multi-class classification problem where each report could be associated with two or more section codes. Recommending a list of these penal codes reduced the time it took a police officer to analyze a report by 50%. 

This model was trained using a sample of 200,000 reports and then deployed in a secure environment.  

Technologies Used 

Several technologies were also critical to the project. These included: 

  • Databases – MongoDB 
  • Web Application Development – Python, Javascript
  • Translation API – Yandex