Opinion analysis in social networks
At present day, there have been a large number of social networks such as Facebook, MySpace, Habbo, Windows live spaces, Orkut, Qzone, and many others. Many users may be interested in finding out new topics and opinions in those social networks. This is a new application area for some theoretical work on information retrieval, machine learning, text document categorization, etc. There has been some preliminary research on this issue (e.g., see references [1-3]), but more sophisticated methods are very much in demanding.
Usually, an opinion-finding system is built on the top of a conventional
information retrieval system, which retrieves a list of relevant documents with scores for a given query, but not care if they are opinionated or not. Then an opinion finding subsystem is used to score all those documents retrieved. Finally, all the retrieved documents are re-ranked using the combination of scores obtained at the first and the second stages.
This project is aimed to investigate the technology that can be used to identify the most useful newtorks for a given topic or find out the most prevalent issues discussed in some or most of them. It is also aimed to find some representative posters (such as positive or negative or mixed opinions) for the selected topics. To take on this research project, some knowledge on information access and retrieval, machine learning, text document categorization, statistics, or numerical optimisation methods, would be useful.
References:
[1] Shengli Wu: Applying the data fusion technique to blog opinion retrieval. Expert Systems with Applications, Volume 39, Issue 1, January 2012, Pages 1346-1353.
[2] Mostafa Keikha, Fabio Crestani: Effectiveness of Aggregation Methods in Blog Distillation. Proceedings of the 8th International Conference on Flexible Query Answering Systems, 2009: 157-167.
[3] Yuchul Jung, Yoonjung Choi, Sung-Hyon Myaeng: Determining Mood for a Blog by Combining Multiple Sources of Evidence. Web Intelligence 2007: 271-274
Personnel Involved
First Supervisor: Wu, S Dr
Second Supervisor: Bi, Y Dr
Collaboration: This project does not involve collaboration with another establishment