Title Author Year Objectives Concept Dataset Algorithm Methodology Attribute Adavantages Summary Limitation Performance metrics Link
Concept Dataset Algorithm Methodology Attribute Adavantages Summary Limitation Performance metrics Link of Paper
Response Time (how quickly a person Depression affects millions worldwide
Step 1: Data Collection Use AI to detect mental health issues from speech and text data. Early detection Data privacy concerns Accuracy
replies) and is a major health issue.
1. To analyze the features and techniques used in current ADD 1. DAIC-WOZ (Distress Analysis
Step 2: Preprocessing
studies that support clinical depression diagnosis. Interview Corpus - Wizard of Oz)
Traditional diagnosis methods rely on https://ieeexplore.ieee.org/ielx8/6221020/6363502/11005879.
Collect voice recordings and text messages from users or public Consistency Across Sessions self-reporting and clinical interviews, Limited accuracy due to noisy or biased pdf?
V Kshirsagar, Step 3: Feature Extraction Non-invasive Precision
datasets. (changes over time) which can be biased and limited. data tp=&arnumber=11005879&isnumber=6363502&ref=aHR0cHM6
Ly9zY2hvbGFyLmdvb2dsZS5jb20v
Al for mental health monitoring via S.Suryavamshi,R patil,T Research Paper Year is 2. To evaluate the effectiveness of machine learning methods in "AI for mental health monitoring via
2. E-DAIC (Extended DAIC-WOZ) Step 4: Model Building
speech and text cues pawar 2025 detecting depression across various data sources. speech and text cues"
AI techniques using speech and text
Label the data with mental health conditions like depression or Contextual Cues (background or Difficulty in capturing complex human
Step 5: Training Continuous monitoring cues provide scalable, objective, and Recall (Sensitivity)
anxiety. situation info if available) emotions
early detection of depression.
3. To explore strategies addressing the challenge of data scarcity 3. CMDC (Chinese Multimodal
Step 6: Evaluation
in ADD research Depression Corpus)
Clean and preprocess speech by removing noise and extracting Cultural and language differences affecting
Step 7: Deployment Accessible F1-Score
features like pitch. performance
4. To assess the strengths and limitations of recent unimodal
4. RSDD (Reddit Self-reported
(text/audio) and bimodal (text + audio) ADD methods for scalable Tools You Can Use : librosa, pyAudioAnalysis, wav2vec2
Depression Diagnosis)
mental health assessment.
Clean and preprocess text by removing unnecessary words and
Cost-effective Risk of misdiagnosis or false positives Area Under the ROC Curve (AUC-ROC)
converting to numerical form.
Extract important features from both speech and text that indicate Dependence on quality and quantity of
Objective analysis Specificity
mental health status. labeled data
Build AI models that analyze speech and text separately or
Personalized feedback
together.
Supports mental health professionals
