NTU » Sentiment Analysis

Please review Prof. Qiang YANG's invited talk slides:

http://www.cse.ust.hk/~qyang/Docs/2009/ACL-IJCNLP09_QYang.pdf

and read up the associated papers on domain adaptation (there is a part that is related to 2 view learning, instance alignment).

Look also at pivot Feature approach (page 21 of slides), and see how we can apply it to extract the right sentiment features in one domain (cellphone), and apply the learnt knowledge to a new domain.

It could also be applied to the same domain, but learning the explicit noun 重量, and apply it to extracting the implicit features from adjectives only 太轻.

Formulate the problem and give us a report by the end of August.

This direction is related to your previous two view approach, and also in the same application area of sentiment feature extraction.

Most researchers are interested in classification, and not the more practical sentiment feature extraction.