Predicting Mental Health Problems with Automatic Identification of Metaphors

Area of Science:

• Psycholinguistics
• Computational Psychiatry
• Natural Language Processing

Background:

• Mental health diagnosis is challenging due to cost, time, and delays.
• Previous research links metaphor use in text to author's mental health status.

Purpose of the Study:

• To propose and evaluate an automated method for detecting metaphors in text.
• To predict specific mental health problems including anxiety, depression, inferiority, sensitivity, social phobias, and obsession.

Main Methods:

• Developed an algorithm for automatic metaphor detection in written texts.
• Tested the method on a dataset of second-language student compositions and the eRisk2017 Social Media dataset.

Main Results:

• The proposed approach effectively predicts mental health problems from written texts.
• The metaphor detection algorithm outperforms existing state-of-the-art methods.
• Metaphor usage in non-native languages is also indicative of mental health conditions.

Conclusions:

• Automated metaphor detection is a viable tool for mental health assessment.
• This method offers a promising, scalable approach to early identification of mental health issues.
• The findings highlight the cross-linguistic relevance of metaphor analysis in psychiatry.