Open AccessStatistically weighted reviews to enhance sentiment classification
Author(s) -
S. R. Prakash,
T. Chakravarthy,
E. Kaveri
Publication year - 2015
Publication title -
karbala international journal of modern science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.345
H-Index - 17
eISSN - 2405-6103
pISSN - 2405-609X
DOI - 10.1016/j.kijoms.2015.07.001
Subject(s) - weighting , sentiment analysis , categorical variable , computer science , artificial intelligence , word (group theory) , support vector machine , natural language processing , the internet , data mining , machine learning , information retrieval , world wide web , mathematics , medicine , geometry , radiology
AbstractThe exponential growth of Internet content, due to social networks, blogs and forums necessitate the research of processing the information in a meaningful way. The research area, Opinion mining is at the cross roads of computation linguistic, machine learning and data mining, which analyze the shared online reviews. Reviews may be about a product, service, events or even a person. Word weighting is a technique that provides weights to words in these reviews to enhance the performance of opinion mining. This study proposes a supervised word weighting method that combined, Word Weighting (WW) and Sentiment Weighting (SW). For WW and SW two function each applied based on word frequency. So totally four statistical functions are applied and checked on categorical labels. Support Vector Machine is used to classify the weighted reviews and it outperforms the existing weighting methods. Two different sizes of corpus are used for the verification