Sentiment Analysis Using Emojis

Jeba Rexciya M.; Miraclin Joyce Pamila J.C.

doi:10.37591/jocta.v13i1.914

Abstract

Sentiment analysis is a fast-growing research part in NLP (Natural Language Processing). It is fully focused on categorizing customer’s opinion about a particular product, blogs or comments etc. Public opinion has a significant impact on people's desire to contact with businesses, as well as overall brand perception. According to a Podium research, 93% of buyers believe online reviews affect their shopping decisions. Users may not give you another chance after reading a few unfavourable reviews. They are not going to look into whether or not the feedback was phoney. They will go with something else. Companies that keep a close watch on their public image can quickly solve problems and enhance operations based on input in this environment. In the information era, sentiment analysis makes it possible to measure people's attitudes towards a company. In the existing work, deep learning and machine learning algorithms like ‘Long short-term memory’ and ‘Support vector machine’ were used to classify the sentiments from the text. Identifying a sarcastic word from a text is a difficult task. To address the issue, non-verbal cues like emojis are used. Emojis are actual icons that appear on physical or virtual keyboards and can be used across various platforms and it is reliable ground truth for sentiment. Feature extraction is performed using the deep Residual network ResNet-101 version 2, which incorporates convolutional neural networks for facial emotion recognition in this suggested work. Finally, the proposed ResNet-101 achieved accuracy of 95%. After then, further performance metrics including accuracy, precision, recall, and F-Score are generated. The chances of sarcasm will be less by expressing sentiments through the emojis.

Abstract

Keywords