Text Sentiment Analysis using Different Types of Recurrent Neural Networks
Tetyana O. Filimonova, Oleg I. Pursky, Vitalina Babenko, Andrey Nechepourenko, Victoria Shvets, Volodymyr Gamaliy · 2024
This work presents the results of research aimed at comparing and analyzing Recurrent Neural Networks (RNNs), specifically LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) models to identify the effectiveness of these models in the sentiment analysis of text messages. LSTM and GRU sentiment analysis models have been developed. For model training, a public dataset from the social media platform Twitter, consisting of text messages and corresponding sentiment labels is used. Obtained study results allow to conclude that compared to LSTM, the GRU model has proved to be more effective specifically in the task of text classification, which resulted in higher accuracy and lower losses during training.