Sentiment Classification of Low Resource Language Tweets Using Machine Learning Algorithms

Sentiment analysis is a way of figuring out if a piece of writing or a sentence is positive or negative. Over the past few years, it has garnered a significant amount of interest in the English language. As a result of the rapid expansion of social media in the modern day, an increasing amount of content can be found in languages apart from English. Telugu represents one of the many native languages for which there is comparatively little information on social media. Finding labeled data of sentences in Telugu for sentiment analysis is becoming increasingly challenging. This study provides Sentikanna, a standard Telugu speech corpus built to assist Telugu Sentiment Analysis across multiple domains, including Recipe, Tourism, and Movie reviews. Sentikanna is a corpus including preprocessed and manually annotated Telugu sentences from a variety of sources, annotated according to annotation rules. All the scripts written in the Telugu language have been used. Apart from the creation of the corpora, polarities have been assigned to the sentences. In total, 3113 Reviews are annotated that have been extracted from various domains. Proposed dataset is used to evaluate different sentiment analysis methods and its performances. Later, sentiment evaluation algorithms using support vector machines (SVM) and Naive Bayes (NB) were applied to the proposed dataset in addition to Sentirama data samples. Performance evaluation has been done and it has been observed that SVM and NB showed promising results on proposed dataset comparatively.