Detection of Depression-Related Tweets in Mexico Using Crosslingual Schemes and Knowledge Distillation.

Area of Science:

• Computational linguistics
• Artificial Intelligence in Healthcare
• Mental Health Informatics

Background:

• Mental health issues, particularly depression, pose significant global public health challenges.
• Early diagnosis of depression is crucial to prevent severe outcomes like suicidal ideation.
• Current diagnostic tools lack homogeneity, driving interest in AI for timely detection.

Purpose of the Study:

• To address the scarcity of Spanish language data for AI-driven depression detection.
• To develop and validate a cross-lingual methodology for classifying depression in short texts.
• To analyze depression trends in Mexico during the COVID-19 pandemic using public data.

Main Methods:

• Implemented a cross-lingual scheme to process both Spanish and English texts.
• Utilized existing English language databases to build classification models for Spanish text.
• Validated the methodology with public data to study depression behavior during the COVID-19 pandemic in Mexico.

Main Results:

• The proposed cross-lingual methodology demonstrated high effectiveness, achieving an F1-score of 0.95.
• The approach successfully enables the classification of depression in Spanish tweets by reusing English datasets.
• Analysis of Mexican public data revealed insights into depression patterns during the COVID-19 pandemic.

Conclusions:

• The developed methodology offers a viable solution for depression classification in low-resource languages like Spanish.
• This approach supports the creation of more efficient AI-powered diagnostic tools for mental health.
• The findings highlight the potential of cross-lingual AI in building diverse language databases for mental health research.