Detection of Depression Severity Using Bengali Social Media Posts on Mental Health: Study Using Natural Language Processing Techniques
Bibliographic record
Abstract
BACKGROUND: There are a myriad of language cues that indicate depression in written texts, and natural language processing (NLP) researchers have proven the ability of machine learning and deep learning approaches to detect these cues. However, to date, these approaches bridging NLP and the domain of mental health for Bengali literature are not comprehensive. The Bengali-speaking population can express emotions in their native language in greater detail. OBJECTIVE: Our goal is to detect the severity of depression using Bengali texts by generating a novel Bengali corpus of depressive posts. We collaborated with mental health experts to generate a clinically sound labeling scheme and an annotated corpus to train machine learning and deep learning models. METHODS: We conducted a study using Bengali text-based data from blogs and open source platforms. We constructed a procedure for annotated corpus generation and extraction of textual information from Bengali literature for predictive analysis. We developed our own structured data set and designed a clinically sound labeling scheme with the help of mental health professionals, adhering to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) during the process. We used 5 machine learning models for detecting the severity of depression: kernel support vector machine (SVM), random forest, logistic regression K-nearest neighbor (KNN), and complement naive Bayes (NB). For the deep learning approach, we used long short-term memory (LSTM) units and gated recurrent units (GRUs) coupled with convolutional blocks or self-attention layers. Finally, we aimed for enhanced outcomes by using state-of-the-art pretrained language models. RESULTS: The independent recurrent neural network (RNN) models yielded the highest accuracies and weighted F1 scores. GRUs, in particular, produced 81% accuracy. The hybrid architectures could not surpass the RNNs in terms of performance. Kernel SVM with term frequency-inverse document frequency (TF-IDF) embeddings generated 78% accuracy on test data. We used validation and training loss curves to observe and report the performance of our architectures. Overall, the number of available data remained the limitation of our experiment. CONCLUSIONS: The findings from our experimental setup indicate that machine learning and deep learning models are fairly capable of assessing the severity of mental health issues from texts. For the future, we suggest more research endeavors to increase the volume of Bengali text data, in particular, so that modern architectures reach improved generalization capability.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".