Improving Mental Health Screening and Early Risk Detection in Spanish

JOURNAL OF COMMUNICATION, LANGUAGE AND CULTURE

The use of large language models in automated depression detection.

Current performance estimates of LLMs with respect to depression screening are most likely optimistic, but when restricted to smaller models that could be locally deployed (for privacy protection) in a clinical setting, LLMs do not detect depression with sufficient accuracy, sensitivity, or specificity to be used in a screening programme.

Mental Health Disorder Indications Detection Based on Text Using NLP with EDA Augmentation Techniques

This study aims to build a classification model for the early screening of mental health disorders from social media text data using the CRISP-DM framework. The primary issue of data imbalance between categories was addressed using the Easy Data Augmentation (EDA) technique. Logistic Regression algorithm and TF-IDF feature extraction were used to classify six categories of mental conditions. Test results showed that the model with EDA experienced a slight decrease in global accuracy to 0.74 (compared to 0.76 without EDA) but successfully increased the Recall for the minority class, Mentalillness, significantly from 0.28 to 0.56. This improvement proves that EDA effectively enriches linguistic variation in limited data. The model has been validated by a psychologist and implemented into a web-based application as an indicative early detection tool, not a clinical medical diagnosis.  

The Effect of Hyperparameter Settings on BERT-Based Detection of Mental Health Support Needs on X (Twitter)

Mental health issues, particularly indications of the need for professional support, have become increasingly prominent in the digital era, requiring effective and efficient detection methods. This study develops a text classification model to identify indications of the need for professional mental health support from Indonesian-language posts on the X platform. The dataset consists of 40,000 scraped tweets collected from July 2023 to July 2025, followed by preprocessing (noise removal, slang and bilingual term normalization, and stopword removal) and automatic labeling using TF-IDF scoring and rule-based cue detection. The dataset was split using stratified sampling into train/validation/test sets (70%/20%/10%), with a label distribution of 73% non-needs_help and 27% needs_help. The model was built by fine-tuning BERT (bert-base-uncased) for binary classification (needs_help, non-needs_help) and evaluated using accuracy, precision, recall, and F1-scores (macro and weighted). Hyperparameter experiments across 12 combinations (learning rate 2e-5, 3e-5, 5e-5; batch size 16, 32; epochs 3, 5) show that the best configuration is learning rate 5e-5, batch size 32, and 3 epochs, achieving an accuracy of 0.8608 and a macro F1-score of 0.8017 (macro precision 0.8585; macro recall 0.7733). Epoch-wise stability analysis indicates a consistent decrease in evaluation loss up to epoch 3, while macro F1 peaks at epoch 2 and remains stable at epoch 3, confirming that hyperparameter selection affects both performance and training stability when fine-tuning BERT on informal social media text.

CounselingQA: scaling professional mental-health guidance via dense retrieval and large language models

Results show that dense retrieval provides strong candidates and that transformer/LLM-driven reranking further elevates relevant, on-topic advice to the top positions, indicating that retrieval-first pipelines can help scale access to professional guidance.

AI-Driven Depression Level Predection for Enhanced Mental Health Diagnostics

AI is significantly impacting the way that mental health diagnostic tools are developed through their ability to provide affordable, accessible and efficient means of detecting psychological disorders like depression. While the current state of screening includes many effective tools (i.e., Clinical Interviews and Self- Reported Questionnaires) they have some inherent shortcomings; these include, but are not limited to, being subjective and/or delayed diagnoses, lack of access to individuals who may be experiencing difficulties with their mental health, and dependency on the availability of professional interventions. The shortfalls identified above demonstrate the need for the development of intelligent systems, capable of conducting rapid and accurate evaluations of an individualʼs mental health. An AI-driven Depression Level Prediction System was therefore created to collect structured clinical information, as well as unstructured textual input in order to create a full and complete assessment of an individualʼs mental health condition. Utilizing the PHQ 9 survey instrument as the basis for collecting clinical information, the system utilizes Natural Language Processing techniques to evaluate user-generated text, thereby gaining further insight into an individualʼs emotional and psychological trends. The system described herein utilizes three machine learning-based predictive models Random Forest, SVM, XGBoost) to predict an individualʼs level of depression as one of four categories (minimal, mild, moderate or severe). Unlike prior binary prediction models utilized in the context of mental health evaluations, the described model provides fine-tuned evaluations that can be more effectively used in practical applications of mental health monitoring. Additionally, Explainable AI techniques were incorporated into the design of the system to improve transparency and interpretability of the results produced by the system. Such capabilities enable both patients and clinicians to identify specific variables within the results that contributed to the systemʼs predictions. The modular nature of the system enables scalability, flexibility and efficient operation of the system even when utilizing lightweight hardware that does not require extensive computing capabilities. Experimental validation demonstrated that the described system achieved greater accuracy and better generalization than other systems currently available. Through its ability to process both behavioral inputs, questionnaire responses and textual sentiment analysis, the system offers a holistic view of an individualʼs mental health status. Beyond improving early detection, the described system can assist clinicians and patients in making informed decisions regarding treatment options for issues related to mental health. Therefore, the system serves as a connection between traditional healthcare practices and emerging AI technology to provide a private and secure method for evaluating mental health conditions.