Hindi/Bengali sentiment analysis using transfer learningand joint dual input learning with self-attention

Sentiment analysis typically refers to using natural language processing, text analysis, and computationallinguistics to extract effect and emotion-based information from text data. Our work explores how we caneffectively use deep neural networks in transfer learning and joint dual input learning settings to effectively classifysentiments and detect hate speech in Hindi and Bengali data. We start by training Word2Vec word embeddingsfor Hindi HASOC data set and Bengali hate speech (1) and then train long short-term memory and subsequentlyemploy parameter sharing-based transfer learning to Bengali sentiment classifiers, by reusing and fine-tuning thetrained weights of Hindi classifiers, with both classifiers being used as the baseline in our study. Finally, we useBiLSTM with self-attention in a joint dual-input learning setting where we train a single neural network on the Hindiand Bengali data sets simultaneously using their respective embeddings.