Invited Talk: Opinion Mining and Sentiment Analysis via Divide and Conquer

Most of the existing approaches to sentiment analysis/opinion mining attempted to propose general algorithms to solve the entire problem. Although a great deal of progress has been made in this direction, we argue that it is unlikely to have a one-technique-fit-all solution to the problem because different types of sentences express sentiments/opinions in very different ways. A divide-and-conquer approach is needed, e.g., focused studies on different types of sentences. In this talk, I will discuss opinion mining of two types of sentences, namely comparative sentences and conditional sentences. These sentences have some unique characteristics and meanings which require special handling in opinion mining. In the talk, I will give examples, define sentiment analysis of these types of sentences, and present techniques for their analysis.

Speaker
Bing Liu is a professor of Computer Science at the University of Illinois at Chicago. He obtained his PhD in AI from the University of Edinburgh. Before joining UIC in 2002, he was with the National University of Singapore. He has published extensively in data mining, Web mining and sentiment analysis/opinion mining in leading conferences and journals, e.g., KDD, WWW, AAAI, IJCAI, ICML, SIGIR, and TKDE. He has given many invited talks and tutorials on sentiment analysis and Web data mining. He also authored a textbook titled “Web Data Mining”. On professional services, Liu has served or are serving as associate editors of several journals, as program chairs of ACM International Conference on Web Search and Data Mining (WSDM-2010), ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD-2008), SIAM Conference on Data Mining (SDM-2007), ACM Conference on Information and Knowledge Management (CIKM-2006), Pacific Asia Conference on Data Mining (PAKDD-2002), and as data mining area chairs of International Conference on Web Wide Web (WWW-2005 and WWW-2010). Additional information about him can be found at http://www.cs.uic.edu/~liub.

Industrial Presentation: The Emotional Aspect of an Unfeeling Machine Called JamiQ

Humans created the Computer. Loved it. Nurtured it. Fed it more Ram. Now Computers are learning to understand the thoughts and emotions of its creators. Will it ever happen? Has the evolution already begun? And how mining user-generated content makes it possible. We take apart one such computer and tell you the answer.

Speaker
After training as a Computer Engineer in Nanyang Technological University, JiaYi Lee co-founded JamiQ Pte Ltd, a start-up that specializes in monitoring the Social Media for trends and anomalies. At JamiQ, he develops technology that he believes can reduce Information Overload.

Research Presentations

Searching for Rising Stars in Bibliography Networks

Identifying the rising stars is an important but difficult human resource exercise in organizations. Rising stars are those who currently have relatively low profiles but may eventually emerge as prominent contributors to the organizations or communities. We propose a PubRank algorithm to identify rising stars in research communities by mining the social networks of researchers in terms of their co-authorship relationships. Experimental results show that PubRank algorithm can be used to effectively mine the bibliography networks to search for rising stars in the research communities.

Speaker
See-Kiong Ng currently heads the Data Mining Department at Institute for Infocomm Research. See-Kiong obtained his PhD in computer science from Carnegie Mellon University. He wrote the TrueAllele software when he was a graduate student at CMU. The program was subsequently used by a biotech company in Iceland to genotype the Icelandic population, thereby beginning his journey into the exciting field of genomics as a computer scientist. See-Kiong has since continued to work in the field of bioinformatics, though his recent research also includes attempts to apply what he has learned from bioinformatics to other application domains.  For example, after working on unraveling the underlying functional mechanisms of protein interaction networks, he hopes to apply some of the approaches developed to understand such other real-world networks as human social networks. In general, See-Kiong's research interests include bioinformatics, text mining, privacy-preserving data mining, and social network mining.

Semantic Social Network (SSNet): User Behavior Study in Online Communities

Semantic Social Network, or SSNet, is a project funded by A*Star, Singapore. The two main focuses of the project are (i) to model and discover the behaviors of users and objects interacting with one another in online communities, and (ii) to develop a social network engine to provide storage, query and analytics services to exploit the discovered behaviors. This talk will mainly cover the first main focus of the project. To study user behavior of person interacting with online objects, we studied the user contribution pattern in Wikipedia on approaches to measure the qualities of the articles as well as the controversies of the articles. For user behaviors of person interacting with person, we studied two social networks formed in Yahoo! Answers and Epinions respectively. For Y!A, we aim to rank answers to a given question considering the answers' quality derived from their contributors. For Epinions network, we try to find out the reason behind establishing a trust relationship between users.  For blogosphere study, we look at how the comments can be used to capture readers' understanding about blog posts and also bloggers' tagging behavior.

Speaker
Aixin Sun is an Assistant Professor with School of Computer Engineering (SCE), Nanyang Technological University (NTU), Singapore. He received his B.A.Sc (First class honours) and Ph.D. in 2001 and 2004 respectively, both in Computer Engineering from NTU. He was a Postdoc Fellow with School of Computer Science and Engineering (CSE) at The University of New South Wales (UNSW), Sydney, Australia. He has published more than 40 papers in major international conferences and journals including ACM SIGIR, ACM WSDM, ACM CIKM, ACM/IEEE JCDL, IEEE ICDM, IEEE TKDE, JASIST, and KAIS. Aixin is serving as a PC member of various data mining/information retrieval conferences and reviewer for various journals. He is a member of ACM and a member of IEEE.

Domain Adaptive Information Extraction with Applications in Mining Social Content

Information extraction is an important task in text mining and has many applications ranging from database population to structured search. State-of-the-art information extraction methods are usually based on supervised learning and rely heavily on the availability of training data. They cannot easily adapt to new domains when training data is hard to obtain. In this talk, I will introduce our recent work on domain adaptive information extraction. I will also talk about other text mining problems we have been working on with an emphasis on applications in mining user-generated social content.

Speaker
Jing Jiang is an assistant professor in the School of Information Systems at Singapore Management University. She received her PhD in Computer Science from the University of Illinois at Urbana-Champaign in 2008. Her research interests include information extraction, text mining, information retrieval and machine learning. She has published in a number of major conferences and journals in these areas, and served on the program committees of a number of leading conferences including ACL, SIGIR, CIKM and EMNLP.

Ratings Intelligence in Web 2.0

Ratings intelligence is essentially the knowledge found in ratings.  As the quality of user-generated content varies, many Web 2.0 sites have relied on user ratings to collectively identify the good quality content items.  In this talk, we will focus on mining interesting user behaviors embedded in ratings data. The behavior examples used include user bias and leniency. As user behaviors can depend on behaviors of other users and the characteristics of rated items, we develop behavior models that incorporate such dependencies.  We will also show the empirical results of our proposed behavior models.

Speaker
Ee-Peng Lim is a professor at the School of Information Systems of the Singapore Management University (SMU). He received Ph.D. from the University of Minnesota, Minneapolis in 1994. His research interests include social network/web mining, information integration, and digital libraries. He is currently an Associate Editor of the ACM Transactions on Information Systems (TOIS), Journal of Web Engineering (JWE), IEEE Intelligent Systems, and International Journal of Digital Libraries (IJDL). He is a member of the ACM Publications Board. He c hair ed the Steering Committee of the International Conference on Asian Digital Libraries (ICADL) from 2006 to 2008, and remains to be a member of the committee till now. He is also a member of the Steering Committee of the Pacific Asia Conference on Knowledge Discovery and Data Mining (PAKDD).