Anotace:
This research aims to create and apply efficient sentiment analysis methods for the Bengali language. It also aims to investigate how people in Bangladesh communicate their feelings and mental health issues on social media platforms with a particular emphasis on depression and suicidal thoughts. The process of applying deep learning models to sentiment analysis of suicidal and depressing writing in Bangla entails a few thorough stages. First a dataset of 1076 data points is created by carefully classifying data from a variety of sources including news articles, Facebook, YouTube, and any other online resources into three categories: depressive, non-depressive, and suicidal. Tokenization, stop word removal, and stemming are important preprocessing techniques that help to improve the text. The dataset is split into training and testing sets to train various algorithms. Confusion metrics are used for evaluation and LSTM has the best accuracy (92.01%). This study advances the understanding of sentiment analysis in Bengali by exploring various methodologies and addressing specific challenges in this area. The usefulness of LSTM models is notably highlighted, and it shows that deep learning may be used to achieve accurate sentiment classification. The study compares the simplicity of use of machine learning with the superior performance of deep learning in managing contextual information. The goals are to employ sentiment analysis more widely in interdisciplinary fields and to improve existing methods.