Multilingual Sentiment Analysis for Detecting Mental Health Problems using a Hybrid algorithm combining RNN and Bi-LSTM
Chitralekha Dwivedi, Jyoti Rao · 2023
Sentiment Analysis is gaining traction nowadays owing to its diversified applications, one of them being identifying people’s mental health issues. Researchers have done great work decoding the actual meaning of human sentiments with the help of machine learning and NLP However, multilingual sentiment analysis and detection of depression from the online comments of a user is still demanding the attention of researchers worldwide. This study offers a novel hybrid recurrent neural network (RNN) and bidirectional long short-term memory (Bi-LSTM) architecture for multilingual sentiment analysis to detect mental health issues. RNN captures sequential information, while Bi-LSTM captures bidirectional dependencies, improving performance. The algorithm is comprised of multiple steps. Tokenization and word embeddings are used to preprocess the dataset. Splitting the preprocessed dataset into training and testing sets, sentiment labels are applied to each text sample and encoded numerically. The hybrid RNN and Bi-LSTM model captures sentiment patterns well. The RNN component evaluates sequential information, while the Bi-LSTM component collects forward and backward dependencies to improve contextual knowledge. To train the model, back propagation and gradient descent minimize a loss function like cross-entropy loss.