ID | 118923 |
Author |
Dou, Rongyu
Tokushima University
|
Keywords | Neuro-Symbolic AI
Depression detection
Social media analysis
Early intervention
Sentiment analysis
|
Content Type |
Journal Article
|
Description | This paper introduces TAM-SenticNet, a Neuro-Symbolic AI framework uniquely designed for early depression detection through social media content analysis. Merging neural networks for feature extraction and sentiment analysis with advanced symbolic reasoning, TAM-SenticNet overcomes the limitations of traditional diagnostic tools, particularly in real-time responsiveness and interpretability. The symbolic reasoning, powered by SenticNet, provides a deep, structured understanding of emotional expressions, greatly enhancing model explainability and logical inference. Empirical evaluations reveal that TAM-SenticNet excels beyond existing models in performance metrics, achieving a Precision of 0.665, Recall of 0.881, and F1-score of 0.758, coupled with superior latency metrics, including ERDE5 and ERDE50 at 0.025, LatencyTP at 1.0, and Flatency at 0.675. These achievements highlight TAM-SenticNet’s cutting-edge approach to early depression detection, making it a pioneering tool in the application of AI for mental health informatics.
|
Journal Title |
Computers and Electrical Engineering
|
ISSN | 18790755
00457906
|
NCID | AA00613548
AA11527073
|
Publisher | Elsevier
|
Volume | 114
|
Start Page | 109071
|
Published Date | 2024-01-20
|
Remark | 論文本文は2026-01-20以降公開予定
|
Rights | © 2024. This manuscript version is made available under the CC-BY-NC-ND 4.0 license https://creativecommons.org/licenses/by-nc-nd/4.0/
|
EDB ID | |
DOI (Published Version) | |
URL ( Publisher's Version ) | |
language |
eng
|
TextVersion |
その他
|
departments |
Science and Technology
|