Sentiment Analysis on Social Topics

Overview

Topic status: We're looking for students to study this topic.

Project Summary

Nowadays, the vast proliferation of social media, such as facebook and twitter, and online news reports creates various easy-to-access platforms for human to express their sentimental feelings and opinions towards many topics.  Automatic analysis on the vast repository of sentimental opinions can lead to many useful commercial applications, such as quickly discovering customer concerns about a target product/brand, and which aspects they dislike. Sentiment analysis can also help to summarise the key topics that are most widely covered and/or associated with human emotions. Overtime, this will form a very important intelligence, such as politician’s sentiment towards a person or topic during a period of election campaign, which can be summarised to provide editorial insights and summarisations. 

This project will build on from our current technologies (“Scoop” and video-based emotion classification), exploring the temporal correlations of sentimental states (e.g. positive or negative) revealed by emotional keywords in text and facial expressions in image or video that extracted from on-line news reports and social media, spanning a certain length of period. The results will help to provide a deeper understanding on the potential benefits of combining text with facial expressions for more accurate sentiment analysis. An integrated systematic prototype will be developed that is ready for analysing public sentimental opinions on real-world hot topics.  

Expected outcomes, applications and/or benefits

  • A small dataset related to certain social topics will be collected;
  • An integrated sentiment analysis systematic prototype that combines text keywords with facial expressions will be implemented;
  • Novel knowledge will be derived from experimental comparisons on different modalities of the developed systematic prototype based on the collected dataset;
  • Demos, reports and papers in top quality journals/conferences will be produced.

(The student will collaborate with our team researchers for all the above work)

Required student skills/experience

Computer vision, opinion mining, text analysis, data collection, face emotion analysis from image/video;
MATLAB/C++, experience with similar project.

Study level
Vacation research experience scholarship
Supervisors
QUT
Organisational unit

Science and Engineering Faculty

Research area

Information Systems

Keywords
social, media, information, systems
Contact
Contact the supervisor for more information