A Framework for Sentiment Analysis in Hindi using HSWN

Abstract

Due to increase in amount of Hindi content on the web in past years, there are more requirements to perform sentiment analysis for Hindi Language. Sentiment Analysis (SA) is a task which finds orientation of one's opinion in a piece of information with respect to an entity. It deals with analyzing emotions, feelings, and the attitude of a speaker or a writer from a given piece of information. Sentiment Analysis involves capturing of user's behavior, likes and dislikes of an individual from the text. The work of most of the SA system is to identify the sentiments express over an entity, and then classify it into either positive or negative sentiment. Our proposed system for sentiment analysis of Hindi movie review uses HindiSentiWordNet (HSWN) to find the overall sentiment associated with the document; polarity of words in the review are extracted from HSWN and then final aggregated polarity is calculated which can sum as either positive, negative or neutral. Synset replacement algorithm is used to find polarity of those words which don't have polarity associated with it in HSWN. Negation and discourse relations which are mostly present in Hindi movie review are also handled to improve the performance of the system.

Sentiment Analysis on Indian Indigenous Languages: A Review on Multilingual Opinion Mining

“…Pandey et al [13] defined a framework to carry out the SA task on the Hindi movie reviews. [13] observed that the lower accuracy was obtained by using SWN as a classification technique and hence suggested using synset replacement algorithm along with the SWN. Synset replacement algorithms groups the synonymous words having same concepts together.…”

“…Synset replacement algorithms groups the synonymous words having same concepts together. It helped in increasing the accuracy of the system because if the word was not present in the Hindi SWN then it found the closest word and assigned the score of that word [13]. In the study, Bhargava et al [14] completed the SA task on the FIRE 2015 dataset.…”

“…Some researchers have extracted their data from Twitter [10,17,18,21,24,25,26], while some have opted for extracting the data manually or by performing web scrapping on different websites like Facebook, microblogs, e-commerce websites, YouTube etc. [8,9,12,13,15,23]. Authors in [14] have accessed the FIRE 2015 dataset.…”

A Frame Study on Sentiment Analysis of Hindi Language Using Machine Learning

“…It manages examining individual feelings, sentiments, state of mind and conclusion of a speaker or an author over a question. The essential focus of SA is to discover the assessments communicated by individual over a data or element [2].…”

Sentiment Analysis of Mixed Code for The Transliterated Hindi and Marathi Texts

“…Nagesh and Ravi [26] gave a way to perform language identification using multi class regression classifiers and was able to get nearly 54% accuracy. Pandey and Sharvari [27] applied HSWN along with negation discourse for sentiment analysis of Hindi language text corpora, with the accuracy of near 80%. Srinivas, Sharma and Balbantray [28] demonstrated that text normalization can be achieved using techniques such phonetics based, slang and spelling correction approaches in another limited work [29] they demonstrated application of sentiment analysis techniques on transliterated text by using bilingual dictionary methods and HSWN for sentiment score calculation with 80% accuracy being achieved.…”