Listening To the Crowd: Automated Alignment Aggregation and Analysis of Social Media Responses to Public Events

Project: Research project

Project Details

Description

Listening To the Crowd: Automated Alignment Aggregation and Analysis of Social Media Responses to Public Events Listening To the Crowd: Automated Alignment, Aggregation and Analysis of Social Media Responses to Public Events Abstract: Social media channels such as Google+ and Twitter have emerged as popular platforms for crowds to react and respond to public and/or breaking events. While this promises tremendous opportunities for computational journalism, opinion mining, computational advertisement and event analytics/playback, these promises come entwined with significant technical challenges. In particular, given an event and an associated large scale collection of social media posts, we need approaches to effectively figure out which parts of the event the posts refer to, and then use that alignment to develop further aggregate in-depth analyses. This in turn raises questions about (i) how to segment the event into smaller yet meaningful parts along its timeline, (ii) how to figure out whether a tweet is a general one about the entire event or specific one aimed at a particular segment of the event, and (iii) how to recognize the sentiment of the social media posts and propagate them to the appropriate segments. In this research, we propose to develop a toolbox for supporting this type of analysis, starting with two powerful techniques: (i) ET-LDA, a statistical model, which provides an effective method for aligning an event and its associated social media posts through joint modeling of topical influences from the events and their associated tweets and (ii) SOCSENT, an efficient tool for aggregating the social media posts to analyze the public reception of the individual segments and topics of the event. We share promising results from the preliminary evaluation of a prototype implementation over several public events, including the first Obama-Romney presidential debate. We then discuss our plans to support additional higher-level social media event analytics using our tools. Preliminary work on these tools has appeared at AAAI 2012, ICWSM 2012, ICWSM 2013 and IJCAI 2013.
StatusActive
Effective start/end date8/16/138/15/24

Funding

  • INDUSTRY: Domestic Company: $55,000.00

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.