This directory contains the code that we used in "Detecting Gang-Involved Escalation on Social Media Using Context" (Chang et al., 2018). Since we cannot release the content of the tweets, please note that most of the input content will be missing. Consequently, the released code is not meant to create an end-to-end program that takes in the corpus and outputs the results reported in the paper; rather, it is meant to make clear our process for the experiments. Also we do not release the word embedding file and lexical resources, since they might leak personal information, so please expect that many of the pickle/txt/resource files in the script are missing.
We include here our code for (1) our Convolutional Neural Net system, (2) our process for constructing context features from word embedings. Feel free to email the authors for any related questions.
Serina Chang, sc3003@columbia.edu
Ruiqi Zhong, rz2383@columbia.edu
Ethan Adams, ea2678@columbia.edu
Fei-Tzin Lee, fl2301@columbia.edu
Siddharth Varia, sv2504@columbia.edu
Kathy McKeown, kathy@cs.columbia.edu