Leveraging Transformer with Self-Attention for Multi-Label Emotion Classification in Crisis Tweets

BackgroundBreast cancer affects the lives of not only those diagnosed but also the people around them. Many of those affected share their experiences on social media. However, these narratives may differ according to who the poster is and what their relationship with the patient is; a patient posting about their experiences may post different content from someone whose friends or family has breast cancer. Weibo is 1 of the most popular social media platforms in China, and breast cancer–related posts are frequently found there.ObjectiveWith the goal of understanding the different experiences of those affected by breast cancer in China, we aimed to explore how content and language used in relevant posts differ according to who the poster is and what their relationship with the patient is and whether there are differences in emotional expression and topic content if the patient is the poster themselves or a friend, family member, relative, or acquaintance.MethodsWe used Weibo as a resource to examine how posts differ according to the different poster-patient relationships. We collected a total of 10,322 relevant Weibo posts. Using a 2-step analysis method, we fine-tuned 2 Chinese Robustly Optimized Bidirectional Encoder Representations from Transformers (BERT) Pretraining Approach models on this data set with annotated poster-patient relationships. These models were lined in sequence, first a binary classifier (no_patient or patient) and then a multiclass classifier (post_user, family_members, friends_relatives, acquaintances, heard_relation), to classify poster-patient relationships. Next, we used the Linguistic Inquiry and Word Count lexicon to conduct sentiment analysis from 5 emotion categories (positive and negative emotions, anger, sadness, and anxiety), followed by topic modeling (BERTopic).ResultsOur binary model (F1-score=0.92) and multiclass model (F1-score=0.83) were largely able to classify poster-patient relationships accurately. Subsequent sentiment analysis showed significant differences in emotion categories across all poster-patient relationships. Notably, negative emotions and anger were higher for the “no_patient” class, but sadness and anxiety were higher for the “family_members” class. Focusing on the top 30 topics, we also noted that topics on fears and anger toward cancer were higher in the “no_patient” class, but topics on cancer treatment were higher in the “family_members” class.ConclusionsChinese users post different types of content, depending on the poster- poster-patient relationships. If the patient is family, posts are sadder and more anxious but also contain more content on treatments. However, if no patient is detected, posts show higher levels of anger. We think that these may stem from rants from posters, which may help with emotion regulation and gathering social support.

Feeling Down: Idiom or Nature?

Although more and more researchers pay attention to the emotion classification, traditional emotion classification methods can not embrace changes in the global and local areas of the human brain after being stimulated. We propose an emotion classification method based on SVM combining brain functional connectivity. Firstly, the nonlinear phase-locked value (PLV) is used to calculate the multiband brain functional connectivity network, which is then converted into a binary brain network, and seven features of binary brain network are calculated. Secondly, support vector machines (SVM) are used to classify positive and negative emotions at the valence dimension and arousal dimension in the multiband. Experimental results on DEAP show that the best emotion classification accuracy of the proposed method is 86.67% in the arousal dimension, and 84.44% in the valence dimension. The results demonstrate that the classification accuracy of the arousal dimension is better than the valence dimension and the Beta2 frequency band is more suitable for emotion classification. Finally, several findings on brain functional connectivity network is discussed. The left and right areas of brain functional connectivity network are unbalanced in the low frequency band, and the feature values of clustering coefficient, average shortest path length, global efficiency, local efficiency, node degree are positively correlated with the arousal degree in the arousal dimension. Humans emotions are suppressed in the low frequency band, and the brain functional connectivity network after emotional stimulation is strengthened in the high frequency band. Our findings on emotion classification are valuable and consistent with the study of neural mechanisms.

Methods The study sample included 43 children (37 female) ages 8-17 years (M = 12.77 years, SD = 2.29) previously diagnosed with Juvenile Idiopathic Arthritis (JIA) and having active arthritis in the past 6 months. During an initial study visit, participants were trained in the use of a Smartphone for answering questions about pain, activity limitations, emotions, and use of emotion regulation strategies. Participants then were cued via audible alerts to respond to these questions three times per day for 4 weeks. Hierarchical linear models were used to evaluate the following: (a) whether pain and functional limitations were reliably greater for more emotionally labile children (i.e., children with high variability in positive and negative emotions); (b) whether pain and functional limitations reliably changed during moments when positive and negative emotions were higher or lower than typical for a given child; and (c) whether pain and functional limitations reliably changed during intervals when positive and negative emotions were “successfully” regulated (i.e., either recovered to or maintained within .5 standard deviations of a child’s typical positive or negative emotion intensity following use of one or more emotion regulation strategies). Results Pain intensity was significantly higher for children with greater negative and positive emotion lability (b=55.68, t(41)=2.17, p=.02; b=60.29, t(41)=2.18, p=.04) and at moments when negative and positive emotion intensities were higher or lower (respectively) than a given child’s typical level (b=3.14, t(2389)=-3.13, p<.01; b=-5.49, t(2389)=-8.01, p<.01). Pain intensity was significantly lower during intervals when negative and positive emotions were successfully maintained at adaptive levels (b=-2.51, t(1992)=-2.68, p<.01; b=-5.12, t(1870)=-5.82, p<.01) and when positive emotions were recovered to adaptive levels following a significant drop (b=-5.14, t(1549)=-4.71, p<.01). Limitations in daily activities were reliably greater for children with greater negative emotion lability (b=12.93, t(41)=2.09, p=.02) and at moments when negative emotion intensities were lower than a child’s typical level (b=1.11, t(2389)=-3.14, p<.01). Activity limitations were significantly lower during intervals when positive emotions were successfully maintained at adaptive levels (b=-.62, t(1870)=-2.00, p=.04) or negative emotions were recovered to adaptive levels following a significant rise (b=-1.42, t(804)=-2.59, p=.01).

Emotional information plays an important role in various multimedia applications. Movies, as a widely available form of multimedia content, can induce multiple positive emotions and stimulate people's pursuit of a better life. Different from negative emotions, positive emotions are highly correlated and difficult to distinguish in the emotional space. Since different positive emotions are often induced simultaneously by movies, traditional single-target or multi-class methods are not suitable for the classification of movie-induced positive emotions. In this paper, we propose TransEEG, a model for multi-label positive emotion classification from a viewer's brain activities when watching emotional movies. The key features of TransEEG include (1) explicitly modeling the spatial correlation and temporal dependencies of multi-channel EEG signals using the Transformer structure based model, which effectively addresses long-distance dependencies, (2) exploiting the label-label correlations to guide the discriminative EEG representation learning, for that we design an Inter-Emotion Mask for guiding the Multi-Head Attention to learn the inter-emotion correlations, and (3) constructing an attention score vector from the representation-label correlation matrix to refine emotion-relevant EEG features. To evaluate the ability of our model for multi-label positive emotion classification, we demonstrate our model on a state-of-the-art positive emotion database CPED. Extensive experimental results show that our proposed method achieves superior performance over the competitive approaches.

Multi-label emotion classification in texts using transfer learning

While negative emotions are associated with risk behaviors and risk avoidance among people with HIV, emerging evidence indicates that negative self-conscious emotions, those evoked by self-reflection or self-evaluation (e.g., shame, guilt, and embarrassment), may differentially influence health-risk behaviors by producing avoidance or, conversely, pro-social behaviors. Positive emotions are associated with beneficial health behaviors, and may account for inconsistent findings related to negative self-conscious emotions. Using multinomial logistic regression, we tested whether positive emotion moderated the relationships between negative emotion and negative self-conscious emotions and level of condomless sex risk: (1) seroconcordant; (2) serodiscordant with undetectable viral load; and (3) serodiscordant with detectable viral load [potentially amplified transmission (PAT)] among people recently diagnosed with HIV (n=276). While positive emotion did not moderate the relationship between negative emotion and condomless sex, it did moderate the relationship between negative self-conscious emotion and PAT (AOR=0.60; 95% CI 0.41, 0.87); high negative self-conscious and high positive emotion were associated with lower PAT risk. Acknowledgment of both positive and negative self-conscious emotion may reduce transmission risk behavior among people with HIV.

This study classifies and estimates the intensity of multiple emotional states using physiological signals. We employed a jigsaw puzzle task to elicit both positive and negative emotions in participants. Mood states were assessed using the profile of mood states 2nd Edition (POMS2), while electroencephalogram (EEG) and heart rate variability (HRV) signals were recorded simultaneously. Support vector machines (SVMs) were used for emotion classification. Feature extraction techniques were applied to enhance classification accuracy, including principal component analysis (PCA) and autoencoders (AE). Recursive feature elimination (RFE) was utilized to identify key physiological indicators. When PCA or AE preprocessing was applied, the classification model achieved a κ coefficient of over 0.9 for all emotions. The key features for emotion classification were identified as mean RR interval (MRRI), low-frequency power (LF), high-frequency power (HF), ratio, and prefrontal alpha asymmetry (Fp1α-Fp2α), whereas HF, standard deviation of RR intervals, LF, and F7α-F8α showed lower importance. The findings suggest that EEG and HRV signals can classify and estimate multiple emotional states simultaneously. These results contribute to developing objective emotion recognition systems for applications in mental health monitoring and affective computing.Clinical Relevance- Accurately assessing emotional states is crucial for mental health care, stress management, and affective computing applications. The proposed emotion classification model utilizing EEG and HRV signals provides an objective and quantitative approach to evaluating mood states. This study demonstrates the feasibility of non-invasive physiological monitoring for mental well-being assessment, offering potential applications in workplace stress management, early detection of mood disorders, and human-computer interaction systems.

Previous research shows that product design affects consumer emotions and behavior, and thus, generates competitive advantage. To date, however, most of the studies on product design focus on positive emotions and neglect the influence of product design on negative emotions (e.g., Fokkinga & Desmet, 2012). Negative emotions and the coexistence of positive and negative emotions as well as their effects on consumer behavior have been largely neglected (Chitturi, 2009). Similarly, little research in industrial design and marketing has explored the role of single emotions (e.g., joy, boredom). Results of an online experiment (n = 179, Mage = 33.00 years, 53.10% male) reveal that the influence of the three product design dimensions aesthetics, functionality, and symbolism on consumer purchase intention and word-of-mouth behavior (WOM) is mediated by positive consumer emotions. Negative consumer emotions also mediate the relationship between product design and purchase intention, while they are of little relevance for WOM. To gain deeper insights into the mediating effect of single emotions, a disaggregated perspective with 14 single emotions is applied. It becomes apparent that desire, hope, pride, and satisfaction, induced by product design, impact consumer behavior, while joy has the strongest influence. Regarding negative emotions, particularly boredom, in addition to shame and contempt, inhibits purchase intention and WOM for all three dimensions of product design. This paper examines the entire chain of effects from product design over consumer emotions to relevant success metrics (i.e., purchase intention and WOM) and shows that in addition to positive emotions, negative emotions evoked by product design also shape consumer behavior. While WOM is almost exclusively driven by positive emotions, both positive and negative emotions affect purchase intention. Results further reveal joy as a driver and boredom as an inhibitor of purchase intention and WOM. To create joy in consumers, product design should be innovative in aesthetics and functionality to meet consumers’ aim for novelty and change. Marketers should also ensure that the product design generates interest and excitement through aesthetic and functional design features and provides consumers with the opportunity to express themselves through product design, to particularly counteract the negative emotion of boredom. This research crosses the bridge between product design research, emotions, and consumer behavior to generate novel knowledge for academia and management.

A LSTM based deep learning network for recognizing emotions using wireless brainwave driven system

Emotion generation and emotion regulation: The role of emotion beliefs

Rumination has been consistently implicated in the onset and maintenance of depression. Less work has examined rumination in the context of bipolar disorder, especially rumination about positive emotion. The present study examined rumination about negative and positive emotion in interepisode bipolar disorder (BD; n = 39) and healthy controls (CTL; n = 34). Trait rumination about positive and negative emotion, as well as experiential and physiological responses to a rumination induction, was measured. Illness course was also assessed for the BD group. Results indicated that the BD group reported greater trait rumination about positive and negative emotion compared with the CTL group, though no group differences emerged during the rumination induction. For the BD group, trait rumination about positive and negative emotion, as well as increased cardiovascular arousal (i.e., heart rate), was associated with greater lifetime depression frequency; trait rumination about positive emotion was associated with greater lifetime mania frequency. These findings suggest that interepisode BD is associated with greater rumination about positive and negative emotion, which in turn is associated with illness course.

The purpose of this study was to examine how positive and negative emotions fluctuate over time within one workday and to investigate the moderating effects of neuroticism and job satisfaction. Data were obtained from 201 Seoul citizens in Korea using the Day Reconstruction Method (Kahneman, Krueger, Schkade, Schwarz, &amp; Stone, 2004). Data revealed that negative emotions increased over time; positive emotions did not show such a pattern. Job satisfaction correlated positively with average positive emotions and negatively with average negative emotions. Neuroticism correlated significantly and in opposite directions with average positive and negative emotions, but did not correlate significantly with the variability of emotions within a work day. Additionally, neuroticism had a significant moderating effect on the changing pattern of negative (but not positive) emotions over time, such that the negative emotions of workers with high levels of neuroticism increased more sharply than the negative emotions of workers with low levels of neuroticism. Contrary to expectation, job satisfaction did not moderate the pattern of positive or negative emotions at work. Changing patterns of negative emotions may be predictive of occupational accidents and diurnal patterns of positive emotions may be predictive of optimal concentration and efficiency at work. These patterns may also have implications for when we administer surveys in the workplace, when a boss should share bad news with his/her employees.

This study explores customer experience formation in an online shopping context by investigating the causes of customers’ positive and negative emotions during their visit to an online store. Survey data collected from 1786 Finnish online customers was used to identify individuals who experienced strong positive (N = 138) or negative emotions (N = 215) during their visit. The causes of negative and positive emotions were studied by analyzing customers’ open-ended, written explanations attributed to their emotions. Attribution theory is utilized to explain how individuals make sense of their emotions. The findings show that customers offer various explanations for the emotions evoked during a visit to an online store. Three main themes were identified with respect to the causes of such emotions and related to: (1) the online store, (2) the socio-material environment, and, (3) the customer her/himself. Customers generally blame the online store for negative emotions, whereas positive emotions are mostly associated with oneself and one’s success as a consumer. Both negative and positive emotions are to some extent explained by the sociomaterial environment. The findings demonstrate the complexity of customer experience formation. Further investigation of the topic is therefore warranted.

EPA-1062 – Acceptance and suppression of negative and positive emotions in patients with depressive disorders