Sentiment Analysis From Urdu Language-based Text using Deep Learning Techniques

The advancement in technology has enabled people to give reviews in their native language. Urdu is spoken by millions of people worldwide and a substantial amount of textual data is generated in the Urdu language. Therefore, there is a need to explore Urdu language-based data to get valuable insights into public opinion in Urdu-speaking communities. In the past, machine learning, lexicon-based, and rule-based techniques have been employed in sentiment analysis. Recently, sentiments have been classified by using techniques based on transformers due to the integration of self-attention mechanisms. In this work, a framework based on deep learning techniques for sentiment analysis of Urdu language is presented that comprises data curation, pre-processing, and classification stages. We have used a publicly available IMDB dataset of movie reviews translated into Urdu. For sentiment classification, we have performed experiments using three deep learning models i.e., 1-dimensional convolutional neural network (1D-CNN), long short-term memory (LSTM), and Multilingual-MiniLM-L12-H384 transformer: Our experimental results show that the transformer architecture is well-suited for Urdu language sentiment classification and attained a significant accuracy of $89.36 \%$.