-
The #MainBhiChowkidar Campaign: 30% Verified, 36% non-verified Handles added Chowkidar to their Name
Below is the first occurrence of the #MainBhiChowkidar by @narendramodi The tweet was dated March 15th 0900hrs IST. This hashtag was part of the campaign; we found this hashtag trending in India by 1100hrs. We were interested in studying the different facets of this campaign, in particular, name changes in the user name of the accounts, verified and others. Screen name / handle is restricted to 15 alphanumeric characters (letters A-Z, numbers 0-9) with the exception of underscores and name of the account is restricted to 50 alphanumeric characters, including special characters and emojis. First occurrence of MainBhiChowkidar Of the 1,268 verified handles (we reported 1,252 in our blog post…
-
MIME: A Social Network for Meme Enthusiasts. A course project turned into 10,000+ Lines of Code.
Memes have become a predominant means of expression in the Internet culture, they are not only used as a means of entertainment but as an important tool for advertising, spreading propaganda and even political ideologies. However, the creation of memes is limited to a niche community of content creators who are skilled in photo/video editing software. MIME is a platform which empowers people who lack these skills to become a part of the meme conversation. We built a one-stop solution for all meme needs – a maker space where people can create memes, a personalized meme feed, and a space to connect with other meme enthusiasts. Further, we hypothesize that…
-
Hardening Deep Neural Networks via Adversarial Model Cascades
Deep neural networks (DNNs) are vulnerable to malicious inputs crafted by an adversary to produce erroneous outputs. Works on securing neural networks against adversarial examples achieve high empirical robustness on simple datasets such as MNIST. However, these techniques are inadequate when empirically tested on complex data sets such as CIFAR10 and SVHN. Further, existing techniques are designed to target specific attacks and fail to generalize across attacks. We propose Adversarial Model Cascades (AMC) as a way to tackle the above inadequacies. Our approach trains a cascade of models sequentially where each model is optimized to be robust towards a mixture of multiple attacks. Ultimately, it yields a single model which…
-
Female Political Handles: Followed by more, Post more, Re-tweet more, and Follow less
We manually annotated the 1,252 verified handles, for various characteristics of the handles, like party affiliation, state / city, and gender. Among the 1,252 handles, we found 865 (69.1%) to be men, 133 (10.6%) to be women, and the rest 254 (20.3%) handles to be associated with parties. This blog is dedicated to analysing men and women handles. Analysis was done with the data as of March 1, 2019. We found female political handles to have more followers than males. On average, females had 368.5K (min: 422, max: 12.2M) followers, while male handles has 347.8K (min: 47, max: 46.2M) followers. With @narendramodi having the maximum followers (46.2M) in male and…
-
Verified Political Handles: Party Distribution, Political Status, Gender Diversity, and Network Interactions
Our data collection covers 2,585 Twitter handles which we find have some affiliation to an Indian political party. This list was manually curated and we have been collecting data for these handles from late 2018 through the Twitter API. Among these 2,585 handles, we found 1,252 political handles to be Verified Twitter Handles. It is interesting to see this large number of verified accounts in this elections. When we analysed the Twitter data for 2014 Elections, we found only 71 Verified Indian political handles. We manually annotated these handles for their state representation, party affiliation, Lok Sabha / Rajya Sabha affiliation, and gender (in case of a handle representing an…