A Comparative Analysis of RNNs, GRUs and LSTMs in Machine Translation and Sentiment Analysis

• Published in: 2024 4th International Conference on Pervasive Computing and Social Networking (ICPCSN) pp. 293 - 303
• Main Authors: Gupta, Amit, Shastri, Bhavya, Nautiyal, Utsav, Kanupriya, Garg, Navin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 03.05.2024
• Subjects: Accuracy; Analytical models; Artificial neural networks; Computational modeling; Computer architecture; Gated Recurrent Unit(GRU); Long Short-Term Memory(LSTM); Machine Learning; Machine Translation; Natural Language Processing(NLP); Recurrent Neural Network(RNN); Sentiment analysis; Transfer learning
• DOI: 10.1109/ICPCSN62568.2024.00054

Machine translation is the act of translating from one language to another using a computer program or algorithm. The need for translation from one language to another to facilitate communication between different cultures and peoplei s one that we have had since antiquity. Traditional methods of translation are quite slow and cumbersome and are susceptible to biases from the translator. The advent of computers has led us to explore ways in which this process may be automated. Another long-standing problem is the detection and classification of emotions in textual data. The capacity to detect and recognize human emotions is one that is particularly fascinating and important in the field of Artificial Intelligence. Both problems are NLP(natural language processing) problems that involve analyzing textual data which is inherently sequential in nature. Traditional deep neural networks have proven inefficient and inaccurate in tasks involving sequential data such as natural language processing, time series analysis etc. Over the years numerous deep learning models have been developed to address these issues. The advent of RNNs(Recurrent Neural Networks), LSTMs(Long Short-Term Memory) and GRUs(Gated Recurrent Units), which are all deep neural networks built to handle sequential data, has made great progress in machine translation and sentiment analysis. Each of these models has its own unique characteristics and features. This paper aims to compare each model in terms of accuracy, performance, architecture, computational complexity, etc. and thus outline situations in which each model may prove suitable based on its particular set of strengths and weaknesses.