Tentative Project Schedule & Yearly Work Plan
There are nine major tasks in this project. The major tasks in the first year will focus on the design and implementation of the KDC social media collection tools and the prototype of San Diego OES outreach platform. The second year will focus on the analysis of two scenarios (public response to disaster alerts and referendum of controversial topics), and the development of agent-based models, predictive models, and influence maximization algorithms at Kent State University. The Year 3 and Year 4 will follow all eight major tasks, continue to enhance the KDC and M3D frameworks and to create more effective visualization and analytics tools. Each task will be led by one researcher.
Task 1: Tsou and Spitzberg
Task 1A (Tsou): Use the KDC framework and social media APIs to collect and analysis social media messages, and build the spatiotemporal models for human dynamics.
Year-1: Finalize the KDC framework and social media APIs (Twitter, YouTube, and Flickr) for collecting social media messages with two scenarios.
Year-2: Develop new Facebook Graph APIs and Facebook Query Language (FQL) to get more social media messages for social network analysis.
Year-3: Develop new tools to retrieve implicit spatial information from texts (place names), time zones, and friends’ networks in social media messages.
Year-4: Develop in-depth analytical tools and automatic data filter/transformation functions.
Task 1B (Spitzberg): Link the M3D communication theory with computational methods and linguistic research for studying information diffusion.
Year-1: Deliver a more articulated multilevel model of meme diffusion with operationalization recommended for each model component that is compatible with team skill set.
Year-2: Identify falsifiable hypotheses to team adapted to the contexts selected for demonstrating viability of project (i.e., disaster, health crisis, political referenda, etc.)
Year-3: Lead author two to three manuscripts for conference and publication submission; continue refining hypotheses and model to evolving team applications.
Year-4: Pursue manuscript revision and submission, and present a second-generation multilevel model of meme diffusion based on the team’s evolved progress.
Task 2: Tsou, Gawron, and Corliss
Task 2A (Tsou): Develop San Diego OES social media outreach and monitor platform to distribute and analyze the spread of social media messages related to disaster alerts, emergency responses, and public announcements via different social media channels (for scenario one).
Year-1: Coordinate with OES staff and understand their social media management needs.
Year-3: Revise the prototype with user feedback and volunteer suggestions.
Task 2B (Gawron) : Develop computational linguistic methods (named entity extractor, hand-classified corpus, social topic-specific ontology, classifier, etc.) and social network analysis algorithms for analyzing social media diffusion patterns related to selected referendum topics (for scenario two).
Year 1: Preliminary classifier and Source (influence/anti-influence) extraction system for chosen topic area, trained on documents written by some group organized around the topic.
Year 2: Source (influence/anti-influence) extraction system for chosen topic area trained on pro and anti documents, not necessarily produced by dedicated group members.
Year 3: Source extraction system augmented with compositional semantics rules, increasing precision with which influences and anti-influences are identified.
Year 4: Sentiment analysis system based on influence and anti-influence extraction, informed by influence network and community analysis.
Task 2C (Corliss): Develop and implement behavioral theory analysis methods and predictive models for studying social media messages and identify the applicability of behavioral theories for understanding meme diffusion (for both scenarios).
Year-1: Develop behavioral analysis coding list and begin to collect, process, and code data for the qualitative analysis.
Year-2: Use qualitative results to develop and code variables for the quantitative analysis.
Year-3: Develop and test the quantitative predictive models and use findings to inform and refine meme diffusion models.
Task 3: Jin, Lee, and Ye
Task 3A (Jin): Develop predictive models and influence maximization algorithms to understand and predict the spreads (speed, scale, range) of meme in the social network with spatial, temporal, and topic constraints.
Year-1: Investigate and analyze different meme and information diffusion model and develop predictive models to understand and predict the spreads (speed, scale, range) of meme in the social network with spatial, temporal, and topic constraints.
Year-2: Investigate and analyze the existing influence maximization algorithms and develop new approaches with spatial, temporal, and topic constraints to select a list of seeds (opinion leaders) for best meme spreading.
Year-3: Using the developed diffusion model together with influence maximization algorithm to help build a volunteer database to help government deal with emergency response.
Year-4: Monitor and improve the volunteer database to help increase the quick response and reach of emergency notifications.
Task 3B (Lee): Development of agent-based models to simulate how memes flow over different network configurations and the spatiotemporal processes of meme diffusion with GIS maps..
Year 1: Investigate and analyze different configurations for meme networks. Categorize network configurations based on tested meme contents.
Year 2: Development of an agent-based model to model meme diffusion processes. Build a software tool to run the simulation of meme diffusion processes over different networks.
Year 3: Apply proposed methods to analyzing the spatio-temporal trends in the diffusion patterns and processes of memes. Improve the algorithms for various social media contents.
Year 4: Calibrate the agent-based model with additional case studies and scenarios.
Task 3C (Ye): Conduct space-time analysis and modeling on the dynamics of information landscapes and meme diffusion. Analyze driving forces underlying memes flow over different network configurations.
Year-1: Investigate the spatiotemporal effect in the information diffusion process. Characterize the dynamic changes of information landscape through a series of indicators.
Year-2: Apply proposed data model and methods to analyzing the patterns and trends of memes across networks. Develop modeling approaches to examine the driving forces.
Year-3: Extend, modify and improve the algorithms of space-time analysis and modeling for memes in various contexts.
Year-4: Incorporate the spatiotemporal approach development into Task 3B and 4A. Develop more in-depth tools for web-based spatiotemporal visualization.
Task 4: Shi
Task 4A (Shi): Develop HPC solutions and web tools for social media data pre-processing and transformation task, spatial clustering and correlations of online communications, and visualization of dynamic spatiotemporal patterns, in accordance with the social media data structure and database system designed and developed by the KDC framework.
Year-1: Develop HPC solutions for social media data pre-processing and transformation with parallelized/optimized computer programs.
Year-2: Develop HPC solutions for data mining to identify the spatial clustering and correlations for two scenarios.
Year-3: Create parallelized/optimized web-based computer programs for analyzing the social network relationship and spatiotemporal patterns of social media diffusion.
Year-4: Provide effective web tools for visualization and online analytics for studying human dynamics and social media.