Integrating Text and Emoticons for Detecting Extremist Affiliations on Twitter using Deep Learning

The objective of this research paper is to apply opinion mining on Tweets fetched from twitter that would include both emoticons and text to gauge whether the tweets support or relate to extremist affiliations in any way. The study proposes a deep learning classification model that utilizes Distil BERT algorithm. The accuracy, precision, and F1-score are used to compare the performance of the different algorithms. The proposed model combines bi-grams, tri-grams, and Distil BERT model to capture sentiment expressed in text and emoticons. The study spotlights the significance of including emoticons in sentiment analysis, as they provide valuable information that can improve classification accuracy. The dataset is extracted from Twitter. This research provides an insight into the effectiveness of Distil BERT when compared to CNN and LSTM for improving classification accuracy. The implications of the study can be applied to sentiment analysis in social media and extended to other social media platforms that use emojis to express opinions and emotions.