LookingGlass Multimedia: Radio and Programmable Antenna for Social Media Monitoring
Abstract
Project Summary LookingGlass Multimedia: Radio and Programmable Antenna for Social Media Monitoring 1.0 Scope. Social movements are dynamic systems, evolving through different phases such as emergence, mergers, splits, alliances, and extinction. We need better tools and algorithms that answer questions, such as, WHAT to pay attention to and WHY, and HOW to put the information into use. In order to give actionable alerts, we need to understand how does the network change over time? In particular, was there a build up towards an event and was there an uptick post event? Users would like to know HOW big a network ripple an event has caused (e.g. a UAV strike in Kenya) and WHERE are the pockets of negative and positive reactions? WHAT topics, words, and phrases are persuasive or popular in WHICH camp? And WHAT are the causes of grouplevel evolutions (e.g., merging, splitting)? Community detection is a fundamental task in complex network analysis nowadays. There are many community detection algorithms.??However, these algorithms generally consider networks to be static, while most real-world networks change over time.??In evolutionary clustering, a good clustering result should fit the current data well, while simultaneously not deviating too dramatically from the recent history. It should be less sensitive to short-term noises while at the same time adaptive to long term cluster trends. We propose to develop smooth incremental evolutionary co-clustering algorithms for identifying groups, their dynamics and drivers. Specifically, we propose to (1) detect groups experiencing highest rates-of-change, (2) characterize the types of change (i.e. a set of groups are coming-together or moving-apart), and (3) identify their key drivers (i.e. events, issues). This project builds on findings and tools developed in Arizona State University’s DoD Minerva funded project, “Finding Allies for the War of Words” (ONR Award #: N00014-09-1-0815), where we developed visual interfaces that enable analysts to: (1) program full ideological spectrums of groups and organizations on various socio-cultural, political and behavioral antennas, (2) bootstrap an accurate receiver, named LookingGlass, which allows the analysts to tune into any channel, corresponding to a socio-political orientation, and browse spatially and temporally mapped shifting positions and activity of its online followers. In this proposal, we plan to pursue fundamental research in how to expand the reach of LookingGlass into new social media platforms and multimedia data types by (1) discovering evolving and threatening latent groups alongside their political values and core attitudes, (2) validating and re-programming LookingGlass antennas on-the-fly with newly discovered groups, (3) identifying “extended membership” of threatening groups in multiple social networks, (4) download their popular videos, images, and symbols from multiple social media sources (such as FaceBook, Flickr, Instagram, PhotoBucket, Vimeo, YouTube) using their public API s. We also propose to incorporate multimedia analytic capabilities into LookingGlass by integrating existing technologies for (i) facial recognition, (ii) logo/symbol matching, and (iii) speaker recognition in audio, video and images.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 08, 2016
- Source ID
- N000141512722
Entities
People
- Hasan Davulcu
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy