Implicit Affinity Networks--new research from the BYU Computer Science Department
Matt Smith, a PhD candidate researching with Dr. Christophe Giraud-Carrier in the BYU Computer Science Department, presented a research paper at the Workshop on Information Technologies and Systems (WITS) in December. The Workshop, in its seventeenth year, is well-known and prestigious within a growing body of cross-disciplinary scholarship which incorporates computer science, business, and the social sciences to interrogate various social issues.
WITS had a 32% acceptance rate, publishing only 36 papers in its proceedings. Smith's paper, Implicit Affinity Networks, was one of the few to be presented during the workshop as part of the social networking session. Smith was also one of only five presenters to be nominated for best paper award.
Smith, Dr. Giraud-Carrier, and the students researching in the Data Mining Laboratory pioneered the concept of implicit affinities, a method of making connections between individual entities that are not explicitly linked. Smith's research involved a study of the blogosphere and focused on linking the blogs through the content in each posting. Using latent Dirichlet allocation (LDA), Smith extracted topics from over 19,000 blog entries authored by about 2,000 bloggers. For example, one topic was created from blog postings using the phrases "Rudy Giuliani," "John McCain," "White House," "Mitt Romney," "Homeland Security," "Al Qaeda," and "Hillary Clinton." Another category was made up of postings which mentioned "fourth quarter," "stock symbol," "cash flow," and "net income." Using this information, Smith was able to create links between different bloggers who had no previous connections, solely based on the content of their blogs.
Smith's research and the concept of implicit affinities allow social scientists to explore and compute social capital in a revolutionary way. With the influence and omnipresence of the internet, a multitude of social networking devices have been introduced, including, among others, LinkedIn, MySpace, and Facebook. However, these programs rely on explicit affinities—each user must manually connect himself or herself with other users to find "friends" and associates. The innovative concept of implicit affinities, however, automatically creates connections based on common interests, a phenomenon known as "bonding" in the social sciences.
This new method of calculating social capital within a community gives a sense of trends within a community and allows one to see how tightly knit a given community is. Implicit affinities allows for bonding as well as bridging—creating connections between people who have different, yet complementary, interests and skills.
Smith's research has the potential to influence a variety of fields. The concept of implicit affinities opens the door for viral marketing, allowing companies to craft their marketing campaigns to target specific groups of people. It also has the potential to affect political organizations and campaign strategies, and it may even hold influence in the medical domain. For example, Smith hopes that implicit affinities will allow people with medical diseases and disorders to join up with others to create support and information-sharing groups.
Matt Smith was raised in Provo. He and his wife Camille have a one-year old baby boy. Upon graduation, Smith hopes to spend a few years in industry, either at a well-known organization such as Google or at a smaller start-up. Eventually, he would like to follow in the footsteps of his father, a mechanical engineering professor at BYU, and enter the world of academia.
For more information on the Data Mining Laboratory's research, please visit http://dml.cs.byu.edu