?Applying EMONO Variants to Multi-Class Sentiment Analysis for Short-Distance Inter-Class Frequency of Term

Authors

  • Cristopher C. Abalorio, Ariel M. Sison, Ruji P. Medina, Gleen A. Dalaorao

DOI:

https://doi.org/10.17762/msea.v71i4.721

Abstract

Opinion mining has become one of the most sought-after interests by researchers because of the advent of the internet and relevant technologies. Analyzing people's opinions and emotions or sentiment analysis is a subdomain of text classification under NLP. Feature vectorizer is a technique in sentiment analysis frequently used in machine learning approaches to improve classification performance. However, working with multiple categories of sentiments becomes challenging in the machine learning approach using TF-IDF vectorizer as a word tends to be spread out in various classes. In this paper, EMONO, a supervised feature vectorizer with variants TF and SRTF, was implemented to answer the problems in the appropriate representation of terms in multiple sentiments due to a term's short-distance frequency. Results showed for Sentiment Analysis that an EMO value of 3 obtained 74% for KNN and 82% for SVM using EMONO variants compared with 69% of KNN and 81% for SVM in TF-IDF, respectively, on the Commodity News (Gold) dataset. It is evident that EMO that sets extensions of inter-classes coverage in max-occurrence values in EMONO vectorizer improves the classification performance in sentiment analysis with multi-classes.

Downloads

Published

2022-09-05

How to Cite

Cristopher C. Abalorio, Ariel M. Sison, Ruji P. Medina, Gleen A. Dalaorao. (2022). ?Applying EMONO Variants to Multi-Class Sentiment Analysis for Short-Distance Inter-Class Frequency of Term. Mathematical Statistician and Engineering Applications, 71(4), 1938–1947. https://doi.org/10.17762/msea.v71i4.721

Issue

Section

Articles