A survey on transfer learning in sentiment analysis

PhD Qualifying Examination


Title: "A survey on transfer learning in sentiment analysis"

by

Mr. Zheng LI


Abstract:

Sentiment analysis (SA), characterizing human opinions, sentiments and 
attitudes towards entities such as products, services, or events from various 
scenarios, can greatly facilitate commercial applications and society. 
Supervised learning algorithms, especially deep neural networks that heavily 
depend on massive labeled data, have been successfully explored to build 
sentiment classifiers for a specific domain. Unfortunately, capturing all of 
the opinions across widely-varying domains involves labor-intensive and 
expensive labeling costs. Cross domain sentiment analysis, which leverages 
useful knowledge from related source domains with abundant labeled data to 
improve the learning of the target domain with few annotations, has become a 
promising direction. In this survey, we provide a systematic literature review 
on single-domain SA, and more challenging cross-domain SA, including 
traditional pivot based, auto-encoder based, embedding based, and adversarial 
learning based transfer for tackling the domain feature mismatch and semantic 
variation problems. We also introduce the pros and cons of the models in 
different perspectives and point out some promising research directions for 
further enhancement.


Date:			Thursday, 28 June 2018

Time:                  	1:30pm - 3:30pm

Venue:                  Room 5560
                         Lifts 27/28

Committee Members:	Prof. Qiang Yang (Supervisor)
 			Dr. Brian Mak (Chairperson)
 			Dr. Yangqiu Song
 			Dr. Ming Liu (ECE)


**** ALL are Welcome ****