First International CIKM Workshop on Topic-Sentiment Analysis for Mass Opinion Measurement
November 6, 2009, Hong Kong
This workshop seeks to bring together researchers in both computer science and social sciences who are interested in developing and using topic-sentiment analysis methods to measure mass opinion, and to foster communications between the research community and industry practitioners as well.
The increasing amount of user-generated content on the Internet and social media and the digitization of large number of government and institutional documents provide new opinion-rich data sources for researchers to examine individual and group perceptions on products, organizations, and social issues at a large scale, and thus contribute to the research and practice in the areas of political science, social policy, communications, and business intelligence.
On the other hand, researchers are tackling the problem of processing large amount of opinion-rich data using various approaches. The increasing number of relevant publications in top data mining, information retrieval and natural language processing conferences (KDD, SIGIR, ACL, WWW, etc.) has witnessed the growing interest in automatic opinion analysis. Both TREC and TAC (Text Analysis Conference) have set up individual tracks for opinion retrieval and analysis tasks.
In recent years topic detection and tracking techniques have been well developed to identify the issues discussed in a large text collection. Sentiment analysis is catching up to detect the polarity of opinions expressed in texts. However, many times real-world applications have to take into consideration of both topics and sentiments for precise opinion measurement. Topic and sentiment alignment is crucial for opinion retrieval, extraction, categorization, and aggregation on various issues. Topics and sentiments could also have sophisticated interactions. For example, the choice of topics and the attention distribution among topics might bear hidden opinions as well.
How do we build synergistic topic and sentiment models for text documents? How do we tackle the domain-dependency problem of sentiment analysis? How do we identify users' needs and integrate them into the design of opinion analysis systems? What are the successful applications of topic-sentiment analysis for mass opinion measurement? What lessons have the pioneers learned? How do we evaluate the automatic mass opinion measuring tools with regard to the reliability and validity? This workshop solicits submissions to address these problems and more.
We hope this workshop can advance research in topic-sentiment analysis, make connections between research community and industry practitioners and encourage development of high performance tools and systems that can work at the web scale for real world applications.
Suggested topics include, but are not limited to:
- Opinion retrieval, extraction, categorization, and aggregation
- Topic and sentiment alignment in opinion analysis
- Applications of topic-sentiment analysis, e.g. corporate reputation measurement, political orientation categorization, customer preference study, public opinion study
- Issues in using topic-sentiment analysis as a new research method for mass opinion estimation, such as reliability, validity, sample bias, etc.
- Sentiment identification and filtering at various text granularity
- Domain-dependency of sentiment analyzers
- Evaluation methodologies
- Performance issues, scalability and efficiency
- Web-based system demonstration
- Novel algorithms, tools and systems
- Construction of benchmark data sets
- Individual workshop papers due: July 20, 2009
- Notification of Acceptance: August 10, 2009
- Camera ready: August 15, 2009
- Early registration deadline: August 15, 2009
- Workshop: November 6, 2009
Seen through the SIG-IRList.