Multi-Lingual Sentiment Analysis of Urdu and English Tweets Using RNN with Bidirectional LSTM

In this research paper, we introduce a cross-lingual sentiment analysis model designed to evaluate sentiments expressed in tweets associated with Pakistan’s political parties, across three languages: Urdu, transliterated Urdu, and English. Our model leverages a recurrent neural network (RNN) featuring a bidirectional Long Short-Term Memory (BiLSTM) architecture. Demonstrating remarkable proficiency, our model achieved an impressive accuracy rate of 95.64% on the test dataset, surpassing the performance of two other prominent cross-lingual sentiment analysis models, namely mBERT and RoBERTa. This study’s outcomes underscore the effectiveness of our proposed RNN-BiLSTM model for cross-lingual sentiment analysis. The model’s application extends to sentiment assessment of tweets concerning Pakistan’s political landscape, irrespective of language, be it Urdu, transliterated Urdu, or English. This analytical tool holds substantial potential for enhancing our comprehension of public sentiment and tracking evolving political trends within Pakistan.