Files in this item

FilesDescriptionFormat

application/pdf

application/pdfRui_Li.pdf (4MB)
(no description provided)PDF

Description

Title:Towards a general platform for effectively analyzing social media
Author(s):Li, Rui
Director of Research:Chang, Kevin C-C.
Doctoral Committee Chair(s):Chang, Kevin C-C.
Doctoral Committee Member(s):Han, Jiawei; Zhai, ChengXiang; Getoor, Lise
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Social Media
Crawling
User Profiling
Abstract:Social media (e.g., Twitter) now become a popular information channel for general users to create and consume information. With many unique advantages, social media provide tremendous opportunities of analyzing what people are talking about in different domains, such as business intelligence (e.g., finding tablets mentioned by users in Christmas) and emergency management (e.g., finding places where users tweet about tornado) and political analysis (e.g., finding occupations of users who support President Obama). My PhD thesis aims to design a general platform for supporting social media analysis to enable those opportunities. First, I propose a general platform, called BigSocial, which abstracts three essential functionalities, data monitoring, data argumentation and data analysis, for supporting different social media analytic tasks. Then, I study several research problems to realize BigSocial. In the data monitoring layer, I focus on how to efficiently collect relevant data for any given analytic task (e.g., emergency management) from social media (e.g., Twitter), and present the first automatic monitor that continuously collects most relevant tweets for a given task under cost budgets. In the data augmentation layer, I focus on how to accurately and completely profile users’ missing attributes in social media to enable advanced analysis. I begin with exploring how to profile users’ single value attributes (e.g., home location), and develop a probabilistic approach, which accurately profiles Twitter users’ home locations and improves the state of the art methods by 13%. Further, I look into how to profile users’ multiple value attributes (e.g., location, occupation), and design a probabilistic approach, which can discover their multiple locations completely. These studies pave the way for conducting advanced analysis in social media.
Issue Date:2014-05-30
URI:http://hdl.handle.net/2142/49526
Rights Information:Copyright 2014 Rui Li
Date Available in IDEALS:2014-05-30
Date Deposited:2014-05


This item appears in the following Collection(s)

Item Statistics