{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T06:52:51Z","timestamp":1785912771662,"version":"3.56.0"},"reference-count":50,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T00:00:00Z","timestamp":1785888000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Early identification of depression through social media analytics is frequently compromised by linguistic ambiguities, contextual complexities, and the opaque nature of conventional deep learning architectures. This study proposes Graph-RoBERTa-CL, a hybrid neuro-semantic framework that explicitly integrates three synergistic components: (i) RoBERTa-based Transformer semantic encoding for deep contextual representation, (ii) Graph Attention Networks (GAT) for relational contextual aggregation across semantically related posts, and (iii) Supervised Contrastive Learning (SCL) for latent space regularization under class imbalance, collectively enabling explainable depression risk trajectory mapping from social media corpora.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      The framework was trained and validated on the Suicide Detection Dataset (\n                      <jats:italic>N<\/jats:italic>\n                      \u202f=\u202f232,074 posts; 78,541 unique users) using rigorous user-level data splitting to ensure zero user overlap between training and test sets, with cross-dataset consistency assessed on an independent Combined Mental Health Dataset (\n                      <jats:italic>N<\/jats:italic>\n                      \u202f=\u202f3,900 posts) withheld entirely from training and model selection. Both datasets are drawn from Reddit communities with community-based labels; broader generalizability across platforms, populations, and clinically diagnosed cohorts requires further validation. User posts are represented as nodes in a multi-relational semantic graph, where a GAT selectively aggregates contextual embeddings to capture ruminative behavioral patterns and temporal trajectories. A supervised contrastive objective simultaneously reshapes the latent space to improve class separability under linguistic ambiguity, and SHAP-based explanations alongside graph attention visualizations provide structural transparency aligned with established psycholinguistic theories of depression.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>On the internal user-level test set, Graph-RoBERTa-CL achieved Accuracy 91.2%, Precision 89.7%, Recall 93.1%, F1-Score 0.914, and AUC 0.962, substantially outperforming BERT-base (F1: 0.798, AUC: 0.934) and LSTM-Attention (F1: 0.681, AUC: 0.856). On the external independent validation set, the model achieved F1 0.882 and AUC 0.941, with the modest cross-dataset reduction confirming genuine generalizability rather than dataset-specific overfitting. Ablation studies confirm that each architectural component \u2014 semantic encoding, graph attention aggregation, and contrastive optimization \u2014 contributes independently and synergistically to overall performance.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>The proposed framework establishes a rigorous, interpretable, and ethically grounded mechanism for proactive mental health screening in digital ecosystems, designed to function as a human-in-the-loop clinical screening aid rather than a standalone diagnostic tool.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2026.1854179","type":"journal-article","created":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T06:31:23Z","timestamp":1785911483000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Neuro-semantic graph fusion for explainable depression risk trajectory mapping using a graph-enhanced RoBERTa framework"],"prefix":"10.3389","volume":"9","author":[{"given":"M.","family":"Karthiga","sequence":"first","affiliation":[{"name":"Department of Computer Science and Engineering, Bannari Amman Institute of Technology","place":["Sathyamangalam, Tamilnadu, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Emerson Raja","family":"Joseph","sequence":"additional","affiliation":[{"name":"Faculty of Engineering and Technology Centre for Advanced Analytics, COE of Artificial Intelligence, Multimedia University","place":["Melaka, Malaysia"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Subhash","family":"Patil","sequence":"additional","affiliation":[{"name":"Surgery Department, International Medical University, Management and Science University","place":["Shah Alam, Malaysia"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"K.","family":"Saranya","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Engineering, Bannari Amman Institute of Technology","place":["Sathyamangalam, Tamilnadu, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,5]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"2581","DOI":"10.1109\/JBHI.2023.3274486","article-title":"Graph attention-based curriculum learning for mental healthcare classification","volume":"28","author":"Ahmed","year":"2023","journal-title":"IEEE J. Biomed. Health Inform."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"100375","DOI":"10.1016\/j.array.2025.100375","article-title":"Effective depression detection and interpretation: integrating machine learning, deep learning, language models, and explainable AI","volume":"25","author":"Al Masud","year":"2025","journal-title":"Array"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"12635","DOI":"10.3390\/ijerph191912635","article-title":"Detecting and analyzing suicidal ideation on social media using deep learning and machine learning models","volume":"19","author":"Aldhyani","year":"2022","journal-title":"Int. J. Environ. Res. Public Health"},{"key":"ref4","doi-asserted-by":"publisher","first-page":"676","DOI":"10.3390\/electronics11050676","article-title":"Deep learning for depression detection from textual data","volume":"11","author":"Amanat","year":"2022","journal-title":"Electronics"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"7346","DOI":"10.3390\/s23177346","article-title":"Drivers\u2019 mental engagement analysis using multi-sensor fusion approaches based on deep convolutional neural networks","volume":"23","author":"Aminosharieh Najafi","year":"2023","journal-title":"Sensors"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"47","DOI":"10.1007\/s13755-024-00303-9","article-title":"Explainable depression symptom detection in social media","volume":"12","author":"Bao","year":"2024","journal-title":"Health Inf. Sci. Syst."},{"key":"ref7","doi-asserted-by":"publisher","first-page":"100321","DOI":"10.1016\/j.osnem.2025.100321","article-title":"Detecting mental disorder on social media: a ChatGPT-augmented explainable approach","volume":"48","author":"Belcastro","year":"2025","journal-title":"Online Soc. Netw. Media"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"44","DOI":"10.46662\/jrp.v2i2.46","article-title":"The role of AI-driven tools for early detection of mental health disorders: a systematic review","volume":"2","author":"Chaudhry","year":"2024","journal-title":"J. Res. Psychol. (JRP)"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"930291","DOI":"10.3389\/fnhum.2022.930291","article-title":"Mental state classification using multi-graph features","volume":"16","author":"Chen","year":"2022","journal-title":"Front. Hum. Neurosci."},{"key":"ref10","volume-title":"Deep Learning and Natural Language Processing for Suicidal Ideation Using Instagram Posts","author":"Darade","year":"2023"},{"key":"ref11","first-page":"72","author":"Fan","year":"2024"},{"key":"ref12","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2021.emnlp-main.552","article-title":"Simcse: simple contrastive learning of sentence embeddings","author":"Gao","year":"2021"},{"key":"ref13","author":"Garg","year":"2022"},{"key":"ref14","author":"Go","year":"2009"},{"key":"ref15","doi-asserted-by":"publisher","first-page":"1627078","DOI":"10.3389\/frai.2025.1627078","article-title":"Explainable AI-driven depression detection from social media using natural language processing and black box machine learning models","volume":"8","author":"Hameed","year":"2025","journal-title":"Front. Artif. Intell."},{"key":"ref16","first-page":"1","author":"Haque","year":"2020"},{"key":"ref17","first-page":"7184","article-title":"Mentalbert: publicly available pretrained language models for mental healthcare","author":"Ji","year":"2022"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"208711","DOI":"10.1109\/ACCESS.2025.3636147","article-title":"Psychological stress inference for professional women via multimodal deep representations","volume":"13","author":"Jiahui","year":"2025","journal-title":"IEEE Access"},{"key":"ref19","first-page":"1","author":"Jo","year":"2024"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"36","DOI":"10.1111\/acps.13250","article-title":"Combining machine learning algorithms for prediction of antidepressant treatment response","volume":"143","author":"Kautzky","year":"2021","journal-title":"Acta Psychiatr. Scand."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"1219479","DOI":"10.3389\/fpsyt.2023.1219479","article-title":"Toward explainable AI (XAI) for mental health detection based on language behavior","volume":"14","author":"Kerz","year":"2023","journal-title":"Front. Psych."},{"key":"ref22","doi-asserted-by":"publisher","first-page":"92974","DOI":"10.1109\/ACCESS.2021.3093456","article-title":"Attention meets perturbations: robust and interpretable attention with adversarial training","volume":"9","author":"Kitada","year":"2021","journal-title":"IEEE Access"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"4397","DOI":"10.1109\/JBHI.2023.3277789","article-title":"Clinically labeled contrastive learning for oct biomarker classification","volume":"27","author":"Kokilepersaud","year":"2023","journal-title":"IEEE J. Biomed. Health Inform."},{"key":"ref24","doi-asserted-by":"publisher","first-page":"14742","DOI":"10.1109\/ACCESS.2023.3244390","article-title":"Multi-label multimodal emotion recognition with transformer-based fusion and emotion-level representation learning","volume":"11","author":"Le","year":"2023","journal-title":"IEEE Access"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"305","DOI":"10.3390\/ph13100305","article-title":"Prediction of antidepressant treatment response and remission using an ensemble machine learning framework","volume":"13","author":"Lin","year":"2020","journal-title":"Pharmaceuticals"},{"key":"ref26","first-page":"8730","article-title":"An effective deployment of contrastive learning in multi-label text classification","author":"Lin","year":"2023"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"4221","DOI":"10.3390\/s23094221","article-title":"Self-supervised contrastive learning for medical time series: a systematic review","volume":"23","author":"Liu","year":"2023","journal-title":"Sensors"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"1383648","DOI":"10.3389\/fpsyt.2024.1383648","article-title":"Leveraging ChatGPT to optimize depression intervention through explainable deep learning","volume":"15","author":"Liu","year":"2024","journal-title":"Front. Psych."},{"key":"ref29","article-title":"Roberta: a robustly optimized bert pretraining approach","author":"Liu","year":"2019"},{"key":"ref30","first-page":"168","author":"Lu","year":"2022"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"e1","DOI":"10.4108\/eai.13-7-2018.164259","article-title":"Multimodal deep learning based framework for detecting depression and suicidal behaviour by affective analysis of social media posts","volume":"6","author":"Malhotra","year":"2020","journal-title":"EAI Endorsed Trans. Pervasive Health Technol."},{"key":"ref32","doi-asserted-by":"publisher","first-page":"117","DOI":"10.1142\/S2196888822500269","article-title":"Tweets topic classification and sentiment analysis based on transformer-based language models","volume":"10","author":"Mandal","year":"2023","journal-title":"Vietnam J. Comput. Sci."},{"key":"ref33","doi-asserted-by":"publisher","first-page":"1612769","DOI":"10.3389\/fpsyg.2025.1612769","article-title":"Enhancing TextGCN for depression detection on social media with emotion representation","volume":"16","author":"Mao","year":"2025","journal-title":"Front. Psychol."},{"key":"ref34","article-title":"Performance evaluation of deep learning models on suicide ideation detection of reddit posts","volume-title":"NHSJS","author":"Meng","year":"2024"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"103189","DOI":"10.1016\/j.artmed.2025.103189","article-title":"An anxiety screening framework integrating multimodal data and graph node correlation","volume":"167","author":"Mo","year":"2025","journal-title":"Artif. Intell. Med."},{"key":"ref36","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3681794","article-title":"AI-assisted diagnosing, monitoring and treatment of mental disorders: a survey","volume":"5","author":"Muetunda","year":"2025","journal-title":"ACM Trans. Comput. Healthc."},{"key":"ref37","doi-asserted-by":"publisher","first-page":"58","DOI":"10.1016\/j.future.2020.06.050","article-title":"Transformer based deep intelligent contextual embedding for twitter sentiment analysis","volume":"113","author":"Naseem","year":"2020","journal-title":"Futur. Gener. Comput. Syst."},{"key":"ref38","doi-asserted-by":"crossref","DOI":"10.20944\/preprints202512.2393.v1","article-title":"A survey of contrastive learning in medical AI: foundations, biomedical modalities, and future directions","author":"Obaido","year":"2025"},{"key":"ref39","first-page":"11130","author":"Pan","year":"2022"},{"key":"ref40","doi-asserted-by":"publisher","first-page":"e63962","DOI":"10.2196\/63962","article-title":"Emotion forecasting: a transformer-based approach","volume":"27","author":"Paz-Arbaizar","year":"2025","journal-title":"J. Med. Internet Res."},{"key":"ref41","doi-asserted-by":"publisher","first-page":"114","DOI":"10.3390\/info16020114","article-title":"Detection of depression severity in social media text using transformer-based models","volume":"16","author":"Qasim","year":"2025","journal-title":"Information"},{"key":"ref42","doi-asserted-by":"publisher","first-page":"100311","DOI":"10.1016\/j.osnem.2025.100311","article-title":"Advancing depression detection on social media platforms through fine-tuned large language models","volume":"46","author":"Shah","year":"2025","journal-title":"Online Soc. Netw. Media"},{"key":"ref43","article-title":"Improving the robustness of transformer-based large language models with dynamic attention","author":"Shen","year":"2023"},{"key":"ref44","doi-asserted-by":"publisher","first-page":"253","DOI":"10.22266\/ijies2025.0930.16","article-title":"A deep learning framework for depression and major depressive disorder detection from social media text using MentalBERT and multilayer perceptron","volume":"18","author":"Shwetha","year":"2025","journal-title":"Int. J. Intell. Eng. Syst."},{"key":"ref45","author":"Vere","year":"2021"},{"key":"ref46","first-page":"373","author":"Verma","year":"2022"},{"key":"ref47","doi-asserted-by":"publisher","first-page":"109335","DOI":"10.1016\/j.knosys.2022.109335","article-title":"Sentiment lexical strength enhanced self-supervised attention learning for sentiment analysis","volume":"252","author":"Wang","year":"2022","journal-title":"Knowl.-Based Syst."},{"key":"ref48","doi-asserted-by":"publisher","first-page":"161203","DOI":"10.1109\/ACCESS.2024.3488081","article-title":"Integrating BERT with CNN and BiLSTM for explainable detection of depression in social media contents","volume":"12","author":"Xin","year":"2024","journal-title":"IEEE Access"},{"key":"ref49","first-page":"622","author":"Zheng","year":"2024"},{"key":"ref50","doi-asserted-by":"publisher","first-page":"372","DOI":"10.1016\/j.jad.2021.03.079","article-title":"Machine learning to predict clinical remission in depressed patients after acute phase selective serotonin reuptake inhibitor treatment","volume":"287","author":"Zhou","year":"2021","journal-title":"J. Affect. Disord."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1854179\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T06:31:25Z","timestamp":1785911485000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1854179\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,5]]},"references-count":50,"alternative-id":["10.3389\/frai.2026.1854179"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1854179","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,5]]},"article-number":"1854179"}}