Revealing semantic and emotional structure of suicide notes with cognitive network science

Gendered Patterns in Manifest and Latent Mental Health Indicators Among Suicide Decedents: 2003-2020 National Violent Death Reporting System (NVDRS)

Objectives. To investigate differences in the documentation of mental health symptomology between male and female suicide decedents in the 2003-2020 US National Violent Death Reporting System (NVDRS). Methods. Using information on 271 998 suicides in the 2003-2020 NVDRS, we evaluated precoded mental health-related variables and topic model-derived latent mental health themes in the law enforcement and coroner or medical examiner death narratives compiled by trained public health workers. Results. Public health records of male compared with female suicides were less likely to include notations of mental health conditions or treatment interventions. However, topic modeling of death summaries revealed that male suicide decedents were more likely to evidence several subclinical cognitive and emotional indicators of distress. Conclusions. Suicide death records vary by gender, both in recorded evidence for mental health conditions at time of death and in accompanying narratives describing proximal circumstances surrounding these deaths. Our findings hint that patterns of subclinical mental health changes among men might be less well captured in commonly used mental health indicators, suggesting that prevention efforts may benefit from measures that also target assessment of subclinical distress. (Am J Public Health. 2024;114(S3):S268-S277. https://doi.org/10.2105/AJPH.2023.307427).

An Exploratory Study in the Portuguese Population on Writing a Suicide Note: Correlates in the Suicide Spectrum and Qualitative Analysis

Suicide notes are an important warning sign for suicidal behaviors. The aim of this exploratory research is (a) to contribute to understanding the place of suicide notes in the spectrum of suicidal behaviors in Portugal, and (b) to analyze the content remembered by individuals regarding a suicide note. Two complementary studies were carried out. In the first, a quantitative investigation, the statistical relationship of writing a suicide note: To the occurrence of lifetime suicide attempts, to lifetime self-harming behaviors, and to suicidal ideation in the two weeks prior to the assessment was evaluated. In the second study, a qualitative investigation, the content recalled by individuals regarding a suicide note was analyzed. In the quantitative study, 841 adults aged between 18 and 65 years, and 1,012 young adults participated. In the qualitative study, 18 young adults participated. Findings of the quantitative study reveal that writing a suicide note significantly related to the lifetime presence of a suicide attempt and to self-harming behaviors and to suicidal ideation in the two weeks prior to the assessment. Findings of the qualitative study suggest that individuals who have written suicide notes exhibit a significant self-oriented focus, yet they also demonstrate a strong sense of concern for the survivors.

Understanding for Prevention: Qualitative and Quantitative Analyses of Suicide Notes and Forensic Reports

Suicide risk is associated with vulnerabilities and specific life events. The study's objective was to explore the relevance of data from forensic documentation on suicide deaths to the design of person-centered preventive strategies. Descriptive and thematic analyses were conducted of forensic observations of 286 deaths by suicide, including some with suicide notes. Key findings included the influence of health-and family-related adverse events, emotional states of loss and sadness, and failures of the health system to detect and act on signs of vulnerability, as confirmed by the suicide notes. Forensic documentation provides useful information to improve the targeting of preventive campaigns.

Thinking about life in COVID-19: An exploratory study on the influence of temporal framing on streams-of-consciousness

The COVID-19 global pandemic led to major upheavals in daily life. As a result, mental health has been negatively impacted for many, including college students who have faced increased stress, depression, anxiety, and social isolation. How we think about the future and adjust to such changes may be partly mediated by how we situate our experiences in relation to the pandemic. To test this idea, we investigate how temporal framing influences the way participants think about COVID life. In an exploratory study, we investigate the influence of thinking of life before versus during the pandemic on subsequent thoughts about post-pandemic life. Participants wrote about their lives in a stream-of-consciousness style paradigm, and the linguistic features of their thoughts are extracted using Linguistic Inquiry and Word Count (LIWC). Initial results suggest principal components of LIWC features can distinguish the two temporal framings just from the content of their post-pandemic-oriented texts alone. We end by discussing theoretical implications for our understanding of personal experience and self-generated narrative. We also discuss other aspects of the present data that may be useful for investigating these thought processes in the future, including document-level features, typing dynamics, and individual difference measures.

Cognitive networks detect structural patterns and emotional complexity in suicide notes

Communicating one's mindset means transmitting complex relationships between concepts and emotions. Using network science and word co-occurrences, we reconstruct conceptual associations as communicated in 139 genuine suicide notes, i.e., notes left by individuals who took their lives. We find that, despite their negative context, suicide notes are surprisingly positively valenced. Through emotional profiling, their ending statements are found to be markedly more emotional than their main body: The ending sentences in suicide notes elicit deeper fear/sadness but also stronger joy/trust and anticipation than the main body. Furthermore, by using data from the Emotional Recall Task, we model emotional transitions within these notes as co-occurrence networks and compare their structure against emotional recalls from mentally healthy individuals. Supported by psychological literature, we introduce emotional complexity as an affective analog of structural balance theory, measuring how elementary cycles (closed triads) of emotion co-occurrences mix positive, negative and neutral states in narratives and recollections. At the group level, authors of suicide narratives display a higher complexity than healthy individuals, i.e., lower levels of coherently valenced emotional states in triads. An entropy measure identified a similar tendency for suicide notes to shift more frequently between contrasting emotional states. Both the groups of authors of suicide notes and healthy individuals exhibit less complexity than random expectation. Our results demonstrate that suicide notes possess highly structured and contrastive narratives of emotions, more complex than expected by null models and healthy populations.

Emotional profiling and cognitive networks unravel how mainstream and alternative press framed AstraZeneca, Pfizer and COVID-19 vaccination campaigns

COVID-19 vaccines have been largely debated by the press. To understand how mainstream and alternative media debated vaccines, we introduce a paradigm reconstructing time-evolving narrative frames via cognitive networks and natural language processing. We study Italian news articles massively re-shared on Facebook/Twitter (up to 5 million times), covering 5745 vaccine-related news from 17 news outlets over 8 months. We find consistently high trust/anticipation and low disgust in the way mainstream sources framed “vaccine/vaccino”. These emotions were crucially missing in alternative outlets. News titles from alternative sources framed “AstraZeneca” with sadness, absent in mainstream titles. Initially, mainstream news linked mostly “Pfizer” with side effects (e.g. “allergy”, “reaction”, “fever”). With the temporary suspension of “AstraZeneca”, negative associations shifted: Mainstream titles prominently linked “AstraZeneca” with side effects, while “Pfizer” underwent a positive valence shift, linked to its higher efficacy. Simultaneously, thrombosis and fearful conceptual associations entered the frame of vaccines, while death changed context, i.e. rather than hopefully preventing deaths, vaccines could be reported as potential causes of death, increasing fear. Our findings expose crucial aspects of the emotional narratives around COVID-19 vaccines adopted by the press, highlighting the need to understand how alternative and mainstream media report vaccination news.

Lexicons of Key Terms in Scholarly Texts and Their Disciplinary Differences: From Quantum Semantics Construction to Relative-Entropy-Based Comparisons

Complex networks are often used to analyze written text and reports by rendering texts in the form of a semantic network, forming a lexicon of words or key terms. Many existing methods to construct lexicons are based on counting word co-occurrences, having the advantage of simplicity and ease of applicability. Here, we use a quantum semantics approach to generalize such methods, allowing us to model the entanglement of terms and words. We show how quantum semantics can be applied to reveal disciplinary differences in the use of key terms by analyzing 12 scholarly texts that represent the different positions of various disciplinary schools (of conceptual change research) on the same topic (conceptual change). In addition, attention is paid to how closely the lexicons corresponding to different positions can be brought into agreement by suitable tuning of the entanglement factors. In comparing the lexicons, we invoke complex network-based analysis based on exponential matrix transformation and use information theoretic relative entropy (Jensen–Shannon divergence) as the operationalization of differences between lexicons. The results suggest that quantum semantics is a viable way to model the disciplinary differences of lexicons and how they can be tuned for a better agreement.

Contextual Graph Attention Network for Aspect-Level Sentiment Classification

Aspect-level sentiment classification aims to predict the sentiment polarities towards the target aspects given in sentences. To address the issues of insufficient semantic information extraction and high computational complexity of attention mechanisms in existing aspect-level sentiment classification models based on deep learning, a contextual graph attention network (CGAT) is proposed. The proposed model adopts two graph attention networks to aggregate syntactic structure information into target aspects and employs a contextual attention network to extract semantic information in sentence-aspect sequences, aiming to generate aspect-sensitive text features. In addition, a syntactic attention mechanism based on syntactic relative distance is proposed, and the Gaussian function is cleverly introduced as a syntactic weight function, which can reduce computational complexities and effectively highlight the words related to aspects in syntax. Experiments on three public sentiment datasets show that the proposed model can make better use of semantic information and syntactic structure information to improve the accuracy of sentiment classification.

Establishment and psychometric characteristics of emotional words list for suicidal risk assessment in speech emotion recognition

BACKGROUND

Emotional disturbance is an important risk factor of suicidal behaviors. To ensure speech emotion recognition (SER), a novel technique to evaluate emotional characteristics of speech, precision in labeling emotional words is a prerequisite. Currently, a list of suicide-related emotional word is absent. The aims of this study were to establish an Emotional Words List for Suicidal Risk Assessment (EWLSRA) and test the reliability and validity of the list in a suicide-related SER task.

METHODS

Suicide-related emotion words were nominated and discussed by 10 suicide prevention professionals. Sixty-five tape-recordings of calls to a large psychological support hotline in China were selected to test psychometric characteristics of the EWLSRA.

RESULTS

The results shows that the EWLSRA consists of 11 emotion words which were highly associated with suicide risk scores and suicide attempts. Results of exploratory factor analysis support one-factor model of this list. The Fleiss' Kappa value of 0.42 indicated good inter-rater reliability of the list. In terms of criteria validities, indices of despair (Spearman ρ = 0.54, P < 0.001), sadness (ρ = 0.37, P = 0.006), helplessness (ρ = 0.45, P = 0.001), and numbness (ρ = 0.35, P = 0.009) were significantly associated with suicidal risk scores. The index of the emotional word of numbness in callers with suicide attempt during the 12-month follow-up was significantly higher than that in callers without suicide attempt during the follow-up (P = 0.049).

CONCLUSION

This study demonstrated that the EWLSRA has adequate psychometric performance in identifying suicide-related emotional words of recording of hotline callers to a national wide suicide prevention line. This list can be useful for SER in future studies on suicide prevention.

DASentimental: Detecting Depression, Anxiety, and Stress in Texts via Emotional Recall, Cognitive Networks, and Machine Learning

Most current affect scales and sentiment analysis on written text focus on quantifying valence/sentiment, the primary dimension of emotion. Distinguishing broader, more complex negative emotions of similar valence is key to evaluating mental health. We propose a semi-supervised machine learning model, DASentimental, to extract depression, anxiety, and stress from written text. We trained DASentimental to identify how N = 200 sequences of recalled emotional words correlate with recallers’ depression, anxiety, and stress from the Depression Anxiety Stress Scale (DASS-21). Using cognitive network science, we modeled every recall list as a bag-of-words (BOW) vector and as a walk over a network representation of semantic memory—in this case, free associations. This weights BOW entries according to their centrality (degree) in semantic memory and informs recalls using semantic network distances, thus embedding recalls in a cognitive representation. This embedding translated into state-of-the-art, cross-validated predictions for depression (R = 0.7), anxiety (R = 0.44), and stress (R = 0.52), equivalent to previous results employing additional human data. Powered by a multilayer perceptron neural network, DASentimental opens the door to probing the semantic organizations of emotional distress. We found that semantic distances between recalls (i.e., walk coverage), was key for estimating depression levels but redundant for anxiety and stress levels. Semantic distances from “fear” boosted anxiety predictions but were redundant when the “sad–happy” dyad was considered. We applied DASentimental to a clinical dataset of 142 suicide notes and found that the predicted depression and anxiety levels (high/low) corresponded to differences in valence and arousal as expected from a circumplex model of affect. We discuss key directions for future research enabled by artificial intelligence detecting stress, anxiety, and depression in texts.

Cognitive Network Science for Understanding Online Social Cognitions: A Brief Review

Social media are digitalizing massive amounts of users' cognitions in terms of timelines and emotional content. Such Big Data opens unprecedented opportunities for investigating cognitive phenomena like perception, personality, and information diffusion but requires suitable interpretable frameworks. Since social media data come from users' minds, worthy candidates for this challenge are cognitive networks, models of cognition giving structure to mental conceptual associations. This work outlines how cognitive network science can open new, quantitative ways for understanding cognition through online media like: (i) reconstructing how users semantically and emotionally frame events with contextual knowledge unavailable to machine learning, (ii) investigating conceptual salience/prominence through knowledge structure in social discourse; (iii) studying users' personality traits like openness-to-experience, curiosity, and creativity through language in posts; (iv) bridging cognitive/emotional content and social dynamics via multilayer networks comparing the mindsets of influencers and followers. These advancements combine cognitive-, network- and computer science to understand cognitive mechanisms in both digital and real-world settings but come with limitations concerning representativeness, individual variability, and data integration. Such aspects are discussed along with the ethical implications of manipulating sociocognitive data. In the future, reading cognitions through networks and social media can expose cognitive biases amplified by online platforms and relevantly inform policy-making, education, and markets about complex cognitive trends.