This repository contains code, additional results, and anonymized annotation data, for our recently accepted CSCW'22 paper. If you use this, please cite our paper (currently only the Arxiv version...)
@article{pathakMethodAnalyzeMultiple2021,
title = {A {{Method}} to {{Analyze Multiple Social Identities}} in {{Twitter Bios}}},
author = {Pathak, Arjunil and Madani, Navid and Joseph, Kenneth},
year = {2021},
month = jul,
language = {en}
}
There are three directories here, described below.
In addition, we include two other files:
- Our final document describing the annotation task in
Annotation Task Description.docx - Results for our Demographic and Twitter Status analysis for more identities, in
image_more_demographics.pdf
This directory contains the R code we used to perform the reliability and validity analyses, along with data on the pre-existing lists of identities, behaviors, and modifiers we used for that analysis
This directory contains only datasets that we do not believe can be used to easily re-identify individuals within the datasets we used, but still provide details on our qualitative analyses. The main file of interest is labels_identity.xlsx, which contains results from our 3 rounds of coding discussed in the paper. The other files are just intermediary files used for generating that annotation task.
This directory contains code to run our method for extracting personal identifiers (see 01_gen_identifiers.ipynb and the associated utility code in identifiers_from_bio.py) and the code to run our cluster analysis 02_cluster_analysis.ipynb).