Abstract:
The sentiment discourse (i.e. responses and replay) on the social media such as, Facebook,
Twitter, YouTube, Forums, etc., forms the outbreaks of ample of opinionated thread chains.To
drive a complete message, from social Network discourses every single opinionated text under
opinion thread has to be seen interdependently. But, the methods of straight forward sentiment
mining or computational linguistics being applied, cannot notice the interdependency between
two nodes in the absence of opinion oriented graph. In This study, a graph-based opinion
summarizing model, whose vertices contain message objects or topic under discussion and its
reply nodes that are labeled with opinion polarity is anticipated. The major contribution of this
study is the use of back-trace enabled rule based applied on opinion-oriented graph. The total
Data set used to undertake this study ware collected from social Network site Facebook, from
sport domain and annotated experts. The proposed model extracts the summary of opinions
polarity from the corpus of opinion-oriented graph. Hence, it is possible to achieve enhanced
decisions by summarized sentiments polarity derived from this graph. Experiments are
conducted and have confirmed that the proposed model provided an encouraging result, The
result from the model show that the entropy of thread of discussions are between 0
and 0.5 this indicates that the opinion posted were the same type and positive. This reveals
that the feedbacks given on sport event are positive feedbacks that encourage National team.
However, the graph-based opinion mining model by itself does not automatically identify the
orientation of a text. For this, we put forward the automatic sentiment annotation for better
performance and the use of separate model for local language for enhanced decisions on
domain of the study
