A Novel Method for Hostility Management

With the rise of social networking websites, there is a wave of user opinions expressed in text, image or different content. Different words and images are used to express one’s personal feeling, but sometimes words and sentences are not used properly which results in an offensive or inappropriate language. Therefore, there is a need to look into the problem. There are many techniques to mine the sentiment from the most well-known social media, Twitter, where users post real-time reactions and opinions relating to everything. This is done by first analyzing the words properly by semantic analysis technique. Throughout this paper, we’ve a tendency to create a hybrid approach to exploitation of every word primarily based and lexicon-based types in which to figure out the linguistics orientation of the opinion words in tweets. Our aim is to optimize the classification of sentiments and prevent increasing hostility on social media platforms. Here, we used the machine learning technique to classify the text. Hostility management is finding out if a text is hostile or not. It can occur anytime or at any place. For cleaning text, we use lemmatization and stemming. To extract the features, we have different techniques in consideration that are bag of words, tfidfvectorizer, Word2Vec, Doc2vec and TF-IDF. We have developed an original technique that mixes existing approaches, providing the simplest coverage results and competitive agreement.