Automatically identifying people shot by police from media sources

Despite several instances of societal attention and widespread protests, there is no database of police-involved fatal shootings. To this end, it is extremely important to develop a system that will monitor media reports of police use of force in nearly real time. In particular, my thesis leverages the recent developments in the field of text classification and event extraction to achieve this goal. In order to develop a database of police-involved fatal shootings, we propose a multiple layer structure. The first layer is a Boolean query to extract articles from the Solr database which stores articles scraped from the internet. We then show various comparisons on how text classification performs in this domain and show a comprehensive analysis of the errors such a system makes. Finally, we show our results on a number of event extraction systems and successfully conclude that using event extraction on top of text classification improves the task of victim name extraction.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2019

Author dc:creator

• Satapathy, Sidhartha

Contributors dc:contributor

• Hockenmaier, Julia

• Natural Language Processing
• Police Shooting
• Artificial Intelligence
• Deep Learning
• Machine Learning
• Text Classification
• Event Extraction

Statement dc:rights

• Copyright 2019 Sidhartha Satapathy

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/105243

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/105243