Multi-aspect Sentiment Analysis with Topic Models Modeling

Abstract

We have tried to examine the viability of topic
model based methodologies to two multi-aspect sentiment
analysis tasks: multi- aspect sentence labeling and multiaspect rating prediction. For one of the tas ks of sentence
labeling, we propose a weakly-supervised approach that
utilizes only minimal prior knowledge—in the form
of seed words— to uphold an immediate correspondence
between topics and aspects. This correspondence will be
utilized to name sentences with execution that approaches
a fully supervised standard. For multi-aspect rating
prediction, we find that general evaluations can be utilized as
a part of conjunction with our sentence labeling to
accomplish sensible execution contrasted with a fully
supervised baseline. At the point when highest level
perspective evaluations are accessible, we find that topic
model based characteristics can be utilized to enhance
unsophisticated supervised pattern execution, in
concurrence with past multi-aspect rating prediction work...