International Journal in IT & Engineering
Year : 2014, Volume : 2, Issue : 12
First page : ( 95) Last page : ( 107)
Online ISSN : 2321-1776.

Performance evaluation of ABSA system for Restaurant Domain

Kanade Shyam A., M. Tech. Scholar, Chauhan Abhishek Singh, Asst. Professor

NIIST, RGPV, Bhopal, Madhya Pradesh, India

Online published on 22 April, 2015.

Abstract

Today we are having abundant sources and quantity of information available online. This is result of advancements in the field of computing and communication technologies. Due to the availability of restaurant choices, the restaurant selection has influenced the decision making. So, certain websites are dedicated to the task which provides restaurant summaries in the form of ratings, some textual reviews. But still it is challenge to get impressive (densely informative) summary of review without reading all the reviews. To tackle this situation here aspect based sentiment analysis (ABSA) system is introduced. The main objective of ABSA system is to identify aspects and their polarity. Here the task of aspect term extraction was accomplished by dictionary based approach. Then classifiers (SVM and KNN) were used to identify the polarity and the source group of aspect terms. The classifiers were trained from the training dataset (provided for task 4 in SemEval). In this paper the performance evaluation is described for ABSA system for the restaurant domain.

Top

Keywords

ABSA, dictionary based approach, SemEval, SVM and KNN.

Top