A Study Of Mental Health Prediction Based On Metaphorical Features

With the great richness of modern material life,people are more and more concerned about their personal psychological and mental health.Psychological disorders not only affect a person’s work and study,but in serious cases can even endanger their own and others’ lives and property,leading to suicide,crime and other behaviors.In recent years,how to efficiently and accurately identify psychological disorders has gradually become a hot issue for scholars to study.Most of the current approaches focus on the exploration of explicit features of texts and the optimization of representational models,with little work paying attention to deeper linguistic expressions.As a frequently used linguistic expression in daily life,metaphor is closely related to the emotional,cognitive and psychological states of individuals.Previous research has confirmed that there are differences in metaphor use among people in different mental health states.This paper examines the value of metaphorical features in predicting mental health based on differences in metaphor use among individuals with psychological problems in terms of implicit text features.The work in this paper consists of three parts:(1)In response to the current lack of mental health analysis datasets,this paper constructs a mental health dataset with students’ handwritten text and mental health labels as the main corpus sources.The collection,annotation and inspection of the dataset strictly follow the principles and processes of corpus construction,and the validity of the data is ensured through quality control.This mental health dataset has certain application value in the fields of psychology,education,and human resources.(2)In response to the current weak analysis of the validity of metaphorical features in mental health analysis,this paper gives the variability in the use of metaphors by people in different mental health states and the necessity of metaphors as an important feature in mental health analysis.A metaphor-emotion based prediction model is constructed,and the experimental results on student mental health dataset and social media dataset illustrate the relevance of metaphor and mental health and the application value of metaphor features.(3)To address the shortcomings of the current mental health prediction using only metaphorical shallow information,this paper proposes a metaphor-based mental health prediction algorithm.The algorithm uses deep learning convolutional-recursive neural networks to extract textual information,and the metaphorical attention mechanism enables the model to focus on the content in the text that is relevant to metaphorical use.Experimental results on social media datasets show that the algorithm has some advantages in the task of predicting users with mental health problems.In this paper,we experimentally illustrate the association between deep textual information and mental health,verify the validity and value of metaphors in prediction tasks,and propose a metaphor-based mental health prediction algorithm.The work in this paper provides textual resources and new algorithm design ideas for the field of mental health research and contributes to the development of mental health prediction.