Natural Language Processing Tools For Gang Violence Prevention

Acquire and prepare data to perform an initial but thorough qualitative study, and to develop and evaluate initial NLP technologies for the purpose of gang violence prevention.

Goal 1. To identify, acquire, and clean datasets of Twitter posts that will allow us to identify those posts that escalate the propensity of violence in Chicago and Baltimore neighborhoods.

Goal 2. To develop a basic set of natural language (NLP) processing tools adequate for the data set., our initial experiments have shown that, because of the nonstandard nature of the language used, off-the-shelf tools do not perform very well. We aim to create tools that can assign a part of speech (POS) as well as syntactic structure to the language in the Tweets with reasonable accuracy.

Goal 3. To investigate how qualitative methods can inform the development of a suite of computational algorithms to predict high stress incidents on Twitter.

Goal 4. To develop and evaluate a sentiment system which detects expressed sentiment and its target. This system will build on the results of Goal 2 (basic NLP) and Goal 3 (qualitative analysis). This aim will show that the goal of building systems that allow for more complex in-context automatic interpretation of our Tweet corpus is possible.

Our Projects have been supported by funding from: