Abstract:

This paper examines the various applications of sentiment analyses in code-mixed text, from annotating reviews by

users to identifying political or societal sentiments among certain subgroups. We propose an ensemble architecture to conduct

sentiment analysis of code-mixed tweets. The deep learning architecture of this project combines deep-learning technologies

such as CNN or self-attention-based LSTM neural networks. CNN is a key component in our architecture, allowing us to

differentiate between positive and negatively-toned tweets. Convolutional layers are excellent at identifying features in text

documents and accurately classifying sentiments for this polarized statement. The LSTM (Long-Term Memory), neural network

component for Neural tweets, has the ability to distinguish the correct sentiment in texts that include multiple units that express

that sentiment. It can also navigate code mixed text that contains mixed sentiments and accurately classify tweets as neural

Tweets.