A Novel Framework Using Neutrosophy for Integrated Speech and Text Sentiment Analysis

Vasantha Kandasamy W B,Florentin Smarandache,Kritika Mishra,Ilanthenral Kandasamy

doi:10.3390/sym12101715

Vasantha Kandasamy W B, Florentin Smarandache + Show 2 more

Open Access

https://doi.org/10.3390/sym12101715

Copy DOI

Abstract

With increasing data on the Internet, it is becoming difficult to analyze every bit and make sure it can be used efficiently for all the businesses. One useful technique using Natural Language Processing (NLP) is sentiment analysis. Various algorithms can be used to classify textual data based on various scales ranging from just positive-negative, positive-neutral-negative to a wide spectrum of emotions. While a lot of work has been done on text, only a lesser amount of research has been done on audio datasets. An audio file contains more features that can be extracted from its amplitude and frequency than a plain text file. The neutrosophic set is symmetric in nature, and similarly refined neutrosophic set that has the refined indeterminacies I1 and I2 in the middle between the extremes Truth T and False F. Neutrosophy which deals with the concept of indeterminacy is another not so explored topic in NLP. Though neutrosophy has been used in sentiment analysis of textual data, it has not been used in speech sentiment analysis. We have proposed a novel framework that performs sentiment analysis on audio files by calculating their Single-Valued Neutrosophic Sets (SVNS) and clustering them into positive-neutral-negative and combines these results with those obtained by performing sentiment analysis on the text files of those audio.

Highlights

While many algorithms and techniques were developed for sentiment analysis in the previous years, from classification into just positive and negative categories to a wide spectrum of emotions, less attention has been paid to the concept of indeterminacy
Speech sentiment analysis using neutrosophic sets has not been done to date, whereas it can provide excellent results
There has been no way to accommodate the neutrosophy in the sentiment analysis of audio

Summary

Introduction

While many algorithms and techniques were developed for sentiment analysis in the previous years, from classification into just positive and negative categories to a wide spectrum of emotions, less attention has been paid to the concept of indeterminacy. Stages of work were inclined towards Boolean logic which meant an absolute classification into positive or negative classes, 1 for positive and 0 for negative. Fuzzy logic uses the memberships of positive and negative that can vary in the range 0 to 1. Neutrosophy is the study of indeterminacies, meaning that not every given argument can be distinguished as positive or negative, it emphasizes the need for a neutral category. Neutrosophy theory was introduced in 1998 by Smarandache [1], and it is based on truth membership T, indeterminate membership I and false membership F that satisfies 0 ≤ T + I + F ≤ 3, and the memberships are independent of each other.

Objectives

Results

Discussion

Conclusion