A framework for sentiment analysis on schema-based research content via lexica analysis

Authors:

Abstract
Sentiment Analysis is one of the significant issues in the area of natural language processing, computational linguistics and text mining. It has also become a potential research area in bibliographic search and opinion mining, which is our main focus in this paper. Sentiment analysis of citations on schema-based research contents, such as scientific articles and reports, may not only makes an appropriate understanding of the issue, but also enhances the level of reliability of citations in those contents. Taking the above points into account, in this paper, a framework for sentiment analysis based on lexica analysis is proposed to determine the polarity of author's opinion about citations in content. This framework consists of three major steps: subjectivity detection, polarity classification and strength determination of opinion. In each step, some considerations are considered to improve the capability of proposed framework. We have applied SentiWordNet, AFINN and Bingliu lexica sets for this purpose. The schema-based research contents that have been considered for this experiment is PubMed, which consists of open source medical articles of US digital library of medicine. Experimental results reveal the superiority of SentiWordNet lexica set compared to the other existing sets in sentiment analysis of PubMed articles.

This publication has 14 references indexed in Scilit: