CFP: International Workshop on Topic Feature Discovery and Opinion Mining (TFDOM'10)

International Workshop on Topic Feature Discovery and Opinion Mining (TFDOM'10)
Joint with the 10th IEEE International Conference on Data Mining (ICDM'10)
December 13, 2010, Sydney, Australia

Dates

• Full paper submission deadline: *** July 23, 2010 ***
• Notification of acceptance: September 20, 2010
• Camera-ready of accepted papers: October 11, 2010
• Workshop: December 13, 2010

Textual data in the world can be roughly categorized into two main types: facts and opinions. Much effort has been devoted to fact-based information processing in the past decades, and many useful techniques have been developed for information retrieval or text mining. In recent years opinion-based information processing has also been receiving increasingly more attention from researchers. Understanding people's opinions about some subject matters or issues is important for organizational decision making in general. For instance, organizations are keen on retrieving and analyzing customers' opinions about products and services so as to develop more effective business strategies for product design and customer centric marketing. Nevertheless, identifying opinion sources, extracting prominent topic features, summarizing relevant opinions, and effectively predicting the polarity of an opinion are all very challenging tasks. These open research problems are the primary focuses of this Topic Feature Discovery and Opinion Mining Workshop.

Topic feature discovery aims to identify on-topic information sources and extract relevant features for a given topic (e.g., a person, an event, or a government policy). The results of many empirical experiments suggested that the effectiveness of traditional text mining methods might be hindered when they were applied to topic feature discovery from opinionated sources. This might be caused by the nature of different problems being tackled, and/or by the inappropriate effectiveness measures borrowed from classical data mining research. For instance, the widely used measures such as support and confidence, turn out to be unsuitable for the leveraging stage. By way of illustration, given a specified topic, usually a highly frequent pattern (normally short in length) is general in semantics and a specific pattern is long in length and low in frequency. The objective of research on topic feature discovery is to design and develop effective and efficient methods to extract subset of features from textual document to describe the specific topics or opinion holders.

Opinion mining, also known as sentiment analysis, aims to summarize and classify opinionated expressions. When compared with traditional fact-based text analysis, research on opinion mining tries to address the new problems related to the identification and analysis of opinions about some topics or facts. More specifically, opinion mining techniques have been applied to predicting the polarity (or inclination) of an opinionated expression related to a topic (i.e., an opinion holder). They have also been applied to consolidating and summarizing the possibly contradictory opinions from a large number of electronic documents such as blogs, online news, consumer comments that contain opinionated expressions. The fundamental problems in opinion mining research include the retrieval of opinionated expressions, identification of opinion holders or the specific features of the opinion holders, classification of the polarities of sentiments related to some opinion holders, fine-grained analysis of feature-based sentiments, detection of opinion spam, and application of opinion analysis to real-world problem solving or decision making. As a matter of fact, there are many opportunities and challenges for extensive research in the field of opinion mining.

Being inter-related, topic feature discovery and opinion mining are highly challenging topics in modern information analysis, from both an empirical and a theoretical perspective. They are also the important issues and the critical steps for Web personalization applications and recommender systems. The research problems related to these two topics have attracted increasingly more attention from researchers in the communities of data mining, Web intelligence, text mining, machine learning, natural language processing, and information retrieval. By highly focusing on these two challenging research topics and their related areas, this workshop aims to advance the theories and techniques for text mining in general and opinion mining in particular, and to explore novel methodologies for the discovery and interpretation of useful and interesting knowledge embedded in textual documents.

Topics include, but are not limited to:

• Relevant feature discovery
• Opinion mining and sentiment analysis
• Multilingual opinion summarization
• Sentiment and subjectivity classification
• Feature-based sentiment analysis
• Information filtering and retrieval
• Text mining
• Text categorizations
• Ontology mining and ontology merging
• Information extraction
• Recommender systems
• Web personalization and opinion analysis
• Evaluation methodologies for topic feature discovery and opinion mining
• Industrial applications of topic feature discovery and sentiment analysis

Contact:

• Yuefeng Li (y2.li AT qut.edu.au)
• Xiaohui (Daniel) Tao (x.tao AT qut.edu.au)