Social Network Data Science Research

Hi friends. I recently earned a PhD in statistical science from Duke. For much of the last four years, I’ve researched text data in social networks… with a passion. Early in life, I learned virtual conversations transcend space and time, as I maintained interaction over multiple days, with people who lived across town. I just remembered some amusing screen names on AIM, haha.

I believe conversations like these matter. Our messages, and the evolving network of friendships they contribute to, represent a part of our personalities and values. Collectively, they represent an aspect of planet earth’s society. Two examples from my life are finding a college best friend on Facebook before our first class together and my current partner on Coffee Meets Bagel prior to dinner.

My purpose in this realm is to summarize the overwhelming amounts of data to clarify topics which matter most and characterize communities which unite those with similar wills. I do this by writing code, implementing models and presenting it. For now I summarize and visualize a network of political bloggers and conversation topics.

Social Network Analysis:

Researching over 100,000 political blog posts, written on hundreds of websites, revealed a network visualization. Information from January 2012 is shown below. More details about how I made it are here: link_block_lda_results. Some methods I’ve applied from the literature are the mixed membership stochastic block model for networks, latent Dirichlet allocation topic model, and other similar versions of these.

network

These results were obtained using a probabilistic graphical model developed with David Banks at Duke.

Text Mining:

With simulation we provide an assignment for each blog post such that bloggers in the same group typed about the same set of topics and shared hyperlinks similarly. It’s implemented in Python and presented in R. Here is a summary of some topics learned:

topics_and_words

Hopefully this has provided some insight into the capabilities of social network research and text mining using publicly available data. This kind of work might be helpful or applicable in digital marketing and political campaigns. If you’re interested in talking or working with me on either of these topics, feel free to text or call 541-633-5550 or email dmoo2009@gmail.com. Peace be with us!

 

My well being partly depends on g1fts from generous supporters. Please send a friend whatever you can to help by P@yPal at dmoo2009@gmail.com. Thanks for any help!