Screening Internet forum participants for depression symptoms by assembling and enhancing multiple NLP methods

Area of Science:

• Computational linguistics
• Mental health informatics
• Psychological assessment

Background:

• Depression significantly impacts quality of life and can prevent individuals from seeking help due to stigma.
• Social media offers anonymous platforms for sharing feelings, but users may underestimate their depression severity.
• There is a need for tools to identify depression symptoms in online text to encourage help-seeking behavior.

Purpose of the Study:

• To develop and evaluate a method for detecting depression symptoms in free text from online environments.
• To assist individuals in recognizing the significance of their depression and the need for professional intervention.

Main Methods:

• Utilized Natural Language Processing (NLP) to parse textual data into grammatical units.
• Analyzed text for depression indicators and their synonyms to quantify symptom frequency.
• Developed a depression scoring system, DepreSD (Depression Symptom Detection), analogous to paper-based scales like CES-D.

Main Results:

• The DepreSD system achieved an average precision of 0.84 (ranging from 0.72 to 1.0).
• The system demonstrated an average F-measure of 0.79 (ranging from 0.72 to 0.9).
• Evaluation indicates a high degree of accuracy in detecting depressive symptoms.

Conclusions:

• The developed NLP-based system effectively detects depression symptoms in online text.
• DepreSD has the potential to help individuals acknowledge their mental health status and seek necessary professional help.
• This approach offers a scalable method for mental health screening in digital communication.