A Mechanism Of Text Pre-Processing For Sentiment Analysis On The Basis Of Comparative Study

Abstract

Sentiment analysis, also known as opinion mining, plays a crucial role in comprehending the sentiments and attitudes expressed in textual data. With the rapid expansion of social networking platforms in today's global society, effective communication has become paramount. This technology can be leveraged for business development, evaluating social activities, and capturing novel ideas through sentiment analysis. Consequently, preprocessing is an essential task in sentiment analysis. People's reviews hold significant importance in this process. Text preprocessing involves cleansing the textual data and preparing it for efficient information processing within the model. The accuracy and effectiveness of sentiment classification models are greatly influenced by text preprocessing. This study aims to investigate and propose a tailored text preprocessing mechanism designed specifically for sentiment analysis, focusing on a comparative study of existing techniques. The proposed mechanism consists of seven phases: data extraction, noise reduction, Text Preprocessing, language identification (Urdu and English), language translation, sentiment analysis, and scorin[1]g module. Our mechanism is applied to a dataset comprising 8000 tweets related to product reviews in Urdu and English. Several experiments are conducted using an unsupervised lexicon-based approach. This research includes a comparative study of various preprocessing techniques, comparing them with our approach. The results show a significant improvement in accuracy from 61.7% to 93.45%.