An Automatic Approach for Interpreting the Sentiments of Short Texts

26 May 2017 Vo Duy Tin ISTD Machine Learning

PhD Student
Vo Duy Tin, Information Systems Technology and Design

Supervisor
Zhang Yue, Assistant Professor, Information Systems Technology and Design


This research project proposes an efficient method for machines to interpret writers’ attitudes, i.e. to learn information about sentiments, for pieces of short texts, such as tweets, automatically.
 
Currently, counting-based methods are used to train machines to interpret sentiments. By using emoticons from large tweets, a prediction-based method has been developed such that sentiments can be predicted by machines.
 
Due to the proposed method’s simplicity, it can interpret sentiments fast and does not take up huge data volume. It also provides for significantly better accuracy across multiple languages (e.g. English and Arabic) compared to even the best of the current methods.

Read more here.