Sunday, November 11, 2012

At the ICCBR 2012 Conference

Back in September I took part in the ICCBR 2012 conference in Lyon. I had a chance to present our research in Case-Based methods for sentiment classification as a main conference paper, and also take part in the doctoral consortium. All in all, a great experience.

Our research paper focus on the following question (which is also a key component of my PhD research): How can case-based methods help in cross-domain sentiment classification? Broadly speaking, the key motivators for exploring these methods are (a) the fact that there are countless strategies for applying a lexicon to extract sentiment orientation from a document - not to mention the choice of sentiment lexicons available is also big -  coupled with (b) no single method is likely to win in all domains an all types of documents one might encounter. Thus, it may be advantageous to somehow reuse the knowledge of which methods worked on past predictions when attempting to classify documents on unseen domains. Enter Case-Based Reasoning!

This idea, along with a sentiment classification experiment across domains is discussed in our paper: A Case-Based Approach to Cross Domain Sentiment Classification (full paper here). I have also posted the PPT of my presentation at the conference.

No comments:

Post a Comment