User Tools

Site Tools


projects:year2:13.2

13.2 - Large-Scale Social Media Analytical Tools with Application to Detecting Emerging Events

Project - Summary

Objectives:

This project has four primary objectives. First, we developed improved methods for sentiment analysis that simultaneously generated aspects and the related user sentiments. Second, we developed improved topic evolution models, which better captures topics and their evolution over time. Third, we developed a high-speed, distributed clustering algorithm, as graph clustering is utilized by many social media analytic techniques. Fourth, we sought to enhance the ability to detect emerging events by detection and tracking of subevents as well as incorporate data from multiple data sources.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Xiaohua Tony Hu PI Not available Not available Drexel University
Ryan G. Benton PI Not available Not available UL Lafayette
Satya S. Katragadda Graduate Student satya@louisiana.edu (337) 482-0625 UL Lafayette
Yue Shang Graduate Student Not available Not available Drexel University
Xiaoli Song Graduate Student Not available Not available Drexel University
Dhanalakshmi Veeramachaneni Graduate Student Not available Not available UL Lafayette

Project - Impact and Uses/Benefits

Simultaneous Aspect and Sentiment Generation Sentiment analysis aims to detect subjective information from messages, giving people a quick overview about public opinions towards a certain entity, such as a hotel, a product or an event. Especially with the information explosion online, the role of sentiment analysis is getting more and more important. For example, a restaurant may have hundreds of online reviews, and it’s hard, if not impossible, for a reader to read over all the reviews to get a whole picture of the restaurant. Sentiment analysis then comes to rescue. It may tell the readers that most people express positive sentiment on the price of the restaurant but negative sentiment on the service; therefore, the readers can get a quick review about the restaurant and make a decision.

Enhance Topic Evolution We are particularly interested in discovering hierarchical structures in data, as hierarchies have been shown to be instrumental in human sense making. Hierarchies expose relationships in data that may be useful for efficient and meaningful real-world evaluation. A sibling relationship between topics may be useful in understanding different facets of general patterns.

Affinity Propagation Converting the Affinity Propagation algorithm into the MapReduce framework realizes the implementation of the two availability and responsibility procedures using Hadoop’s constraints. But once realized (as outlined above), the method can take advantage of Hadoop’s distributed computing power and draw on enormous amounts of data. The Affinity Propagation method by message-passing between data points holds exciting possibilities when paired with Hadoop MapReduce and could be a key to answering various clustering problems in computer science. AP effectiveness on large data collections has not been tested but with the proposed MapReduce-based Pruned AP, the method can scale with distributed computing resources through experiments on the Hadoop platform.

Emerging Event Detection The EDOS method should be capable of providing organizations and individuals notice of events of interest when they are occurring. It provides the potential for early reaction and allows the end user the ability to prepare and control their response (as needed). For instance, the system could provide advanced warning (or confirmation) of disasters/emergencies for Emergency Event Managers. In the case of advanced warning, this allows them to react in a more timely fashion; while in the confirmation mode, it allows them to gauge potential impact. For news organizations, the events may lead them to potentially interesting (or news worthy) stories. Or in the case of subevents, allow them to track and respond to reactions to an ongoing event (via subevent detection).

Project - Deep Dive

Project - Documents

FilenameFilesizeLast modified
13.2_year_2_ip_disclosure_letter.pdf1004.5 KiB2019/08/22 11:13
13.2_year2_final_project_report_combined.pdf5.1 MiB2019/08/22 10:12
projects/year2/13.2.txt · Last modified: 2021/06/02 15:49 by sally.johnson