Linguistic features and psychological states: A machine-learning based approach.

Area of Science:

• Computational linguistics and psychology
• Natural Language Processing (NLP) for mental health analysis

Background:

• Prior research on psychological state detection relied on limited linguistic features.
• Existing methods often used simplistic measures, overlooking nuanced emotional and sentiment information.

Purpose of the Study:

• To propose and evaluate the use of additional linguistic features, specifically sentiment polarities and emotions, for classifying psychological states.
• To enhance the accuracy of detecting various psychological states in text data.

Main Methods:

• Utilized a large dataset of forum posts categorized by psychological states (anxiety, depression, suicide ideation, normal).
• Employed machine-learning algorithms, including Support Vector Machine (SVM) and Deep Learning (DL).
• Incorporated sentiment polarities and emotions as key linguistic features for classification.

Main Results:

• The proposed approach using sentiment polarities and emotions significantly improved the performance of psychological state detection.
• Both Support Vector Machine (SVM) and Deep Learning (DL) models achieved high accuracy in classifying texts related to psychological states.
• This study demonstrates the efficacy of advanced linguistic features in identifying mental health conditions from textual data.

Conclusions:

• The integration of sentiment polarities and emotions represents a novel and effective strategy for detecting psychological states in text.
• Findings suggest that machine learning models, when augmented with these linguistic features, can accurately identify conditions like anxiety, depression, and suicide ideation.
• This research contributes to advancing NLP applications in mental health and offers a pathway for enhanced diagnostic accuracy.