Paragon Science offers a broad range software, research, and data analysis services for both government and commercial applications. Our general research and development areas include:
- computational physics
- applied mathematics
- modeling of complex systems
- pattern formation in nonlinear systems
- numerical analysis
- data mining
- anomaly detection
Unraveling Ebola One Tweet at a Time with Dynamic Graph Analysis
A sample of 2.5M tweets mentioning "Ebola" was collected during November 5-12, 2014. The titles of the 6227 web pages referenced by the tweets were used to cluster the web pages into roughly 100 topics. Paragon Science's patented dynamic anomaly detection software identified the top five most-anomalous topics. This research demonstrates how these techniques allow us to focus attention quickly on viral, emerging topics. A video showing an animation of those anomalous topics and the key related web pages for every hour of that week in November 2014 is available at https://www.youtube.com/watch?v=AEQ02hv4Xjw.
Paragon Science has been granted U.S. Patent #8738652 ("Systems and Methods for Dynamic Anomaly Detection"), issued on May 27, 2014.
Dr. Steve Kramer gave a joint presentation, "Got Chaos? Extracting Critical Business Intelligence from Email Using Advanced NLP and Dynamic Graph Analysis," with Matthew Russell, CTO of Digital Reasoning, at Data Day Texas 2014.
Paragon Science has collaborated with WCG on their MDigitalLife initiative to identify key Twitter influencers and viral topics in their curated community of clinical oncologists.
Paragon Science has been engaged by WhaleShark Media (now RetailMeNot, Inc.) to perform a preliminary data-mining project to extract valuable intelligence from its rich data sets, with the goals of enhancing the end-user experience and driving the success of its clients.
Analysis of the Obama Twitter Network for March 2012
Paragon Science, Inc. has developed a technology that examines dynamic networks and identifies the entities that show the highest levels of abnormal change. We analyzed the social networks induced by March 2012 Twitter data related to President Barack Obama and have identified a set of highly viral web links as key indicators of trends and user interest. In this video, small blue spheres represent Twitter users whose tweets reference websites, which are displayed as cubes. Each line represents a link between a user and a referenced URL. At the end of each day, the viral URL that exhibits the greatest change in user interest for that date is labeled and colored according to the relative change (blue=least, red=most). After the days of March 2012 are shown, the users' nodes are colored according to their respective communities, and network is rotated to show the community structures. The end of the video highlights the top three viral websites in this dynamic social network. URL #1 (http://www.theatlantic.com/international/archive/2012/03/obama-to-iran-and-israel-as-president-of-the-united-states-i-dont-bluff/253875/) is generally positive or neutral regarding President Obama, while the other two URLs are critical. It is interesting to note how the community of users who reference URL #1 is strongly separated from the other groups. In future videos, we will analyze users' influence metrics and perform sentiment analysis.
For more information, send email to firstname.lastname@example.org or call 512.569.9760.