HEDL: Deep learning multiple approaches for early detection of depression using sarcastic text
Sarcasm is particularly notorious towards mental health, and thus it is quite essential for early identification of depressive indicators. This paper introduces the Hybrid Ensemble Deep Learning model as novel for the task of the detection of sarcasm task, targeting the weaknesses which were found i...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | MethodsX |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S221501612500216X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Sarcasm is particularly notorious towards mental health, and thus it is quite essential for early identification of depressive indicators. This paper introduces the Hybrid Ensemble Deep Learning model as novel for the task of the detection of sarcasm task, targeting the weaknesses which were found in traditional approaches of SVC, DT, RF, and LR by using unique combinations of CNN, LSTM, and GRU to capture the sarcasm patterns that appear fine feature representation and enhanced robustness and accuracy. Our model uniquely integrates the architectures of CNN, LSTM, and GRU into one framework for capturing more complex patterns in feature representation, accuracy, and robustness. We tested it on a news headline dataset; HEDL gained 84 % accuracy along with marked reduction in false positives compared to baseline models, which improved the accuracy as well as the recall. Results of the experiment do support that the HEDL model is indeed much more accurate and reliable sarcastic detection methodology; it can have applications such as monitoring mental health or analysing sentiment. • Proposed the Hybrid Ensemble Deep Learning Algorithm (HEDL) for text data. • The proposed model outperforms traditional models in cognitive skill impairment detection. • Demonstrated scalability for diverse healthcare datasets. |
|---|---|
| ISSN: | 2215-0161 |