JMIR Publications

ABSTRACT

Background:

NA

Objective:

Studying the impact of COVID-19 on mental health is both compelling and imperative for the healthcare system preparedness development. Discovering how the pandemic conditions and the governmental strategies and measures have impacted the mental health is a challenging task. In this study, our aim is to detect suicidal ideation by mining textual content extracted from social media by leveraging state of the art NLP techniques.

Methods:

Despite easy expression via social media, suicidal thought remains sensitive and complex to comprehend and detect. Indeed, detecting suicidal ideation take capturing the new suicidal statements employed during the Covid-19 circumstances that represents a different context of expressions. For this task we propose hybrid deep neural network approach CNN-LSTM, in which CNN detects local features such as word sequences, n-grams, and syntactic patterns which are crucial to understand the semantics of the posts. LSTM captures long-range dependencies between words and phrases that are crucial for understanding the context.

Results:

Two state of art deep learning approaches CNN and LSTM are combined based on the features selected from TFIDF, Word2vec and BERT.TFIDF+CNN achieve superior performance, with precision of 94%, recall of 95%, F1-Score of 94%, and accuracy of 93.65%.

Conclusions:

Considering the dynamic nature of suicidal behavior posts we propose a fused architecture which captures both local and global contextual information which is important for understanding the language patterns and predict the evolution of suicidal ideation over the time. The work is divided into two major phases, one to classify the suicidal ideation posts and the other to extract the factors that cause the suicidal ideation. We propose a novel deep learning-based method to classify suicidal and non-suicidal posts. Explainable artificial intelligence (XAI) is used to extract the key factors that contribute to suicidal ideation in order to provide a reliable and sustainable solution. According to LIME and Shap XAI algorithms, there is a drift in the features during and before COVID-19. Due to the COVID-19 pandemic new features have been added which leads to suicidal tendencies. In Future strategies need to be formed to combat this deadly disease. Clinical Trial: NA