Feature Engineering for Telugu Sentiment Analysis A Study of Machine Learning and Deep Learning Models

Abstract

Recent advancements in NLP have made sentiment analysis an essential component newlineof a variety of NLP jobs, including recommendation systems, question answering, and business intelligence products. While sentiment analysis research has been extensively studied in English, it has rarely been studied in Telugu. The majority of research focuseson measuring the sentiments stated in Tweets, news, or reviews that use both Hindi and English phrases. Academicians are increasingly interested in analyzing people s ideas and opinions that were expressed in Indian languages such as Bengali, Telugu, Malayalam, newlineand Tamil. Microscopic study on Indian languages has just lately been published newlinedue to a lack of designated resources. Researchers are focusing more on unique regional languages in India. Analyzing sentiments in regional languages is a complex task that necessitates a standard corpus. Telugu is a regional language with a wealth of data readily available on social media, but finding class labels of sentences for Telugu Sentiment Analysis is difficult. The investigation of sentiment in the Telugu language, on the other hand, is recorded quite infrequently. Deployment of various deep learning techniques in Telugu sentiment analysis has shown only modest results because of the scarcity of annotated datasets. The motive behind this research is to analyze the sentiment analysis newlinein regional languages due to the massive increase of content produced in social media. newlineAnalyzing sentiments in Telugu definitely empowers examination of social media newlineplatforms for various activities like feedback, emerging trends and mentioning the new incoming brands. In this regard, the Sentiraama dataset which is publicly available has been considered and the deep learning models such as BiLSTM and BiGRU have been applied to obtain the results. In this work, the contextual information has been incorporated so that the sentiment analysis could be enhanced with the help of the models. In this study, various domains have been considered and as well as co