“…Since recent work has shown that encoding handcrafted affective knowledge (e.g., sentiment lexicons) can effectively enhance the training of deep DNNs for polarity classification 33 , our solution adds a new branch of sentiment attention upon the EFL-based DNN model 28 to generate a polarity-sensitive embedding space. Figure 5 The structure of EFL-based polarity classification model: 1) the boxes circled in red dashed lines represent the newly added branch of sentiment attention encoding affective knowledge; 2) the orange blocks represent the components of the newly added sentiment learning layer, the green blocks are the ones present in the original EFL model, and the blue blocks are the shared components between these two branches.…”
Section: Supervised Gml Solutionmentioning
confidence: 99%
“…Specifically, we measure the attention weight of each sentiment word by the weighted sum of its sentiment dimension values as follows: in which denotes the attention weight of a sentiment word, denotes the d -th affective dimension weight, and denotes the word’s sentiment value in the d -th affective dimension. Note that the values of represents the weights of six dimensions (namely evaluation, potency, activity, valence, arousal and dominance); in our implementation, we set their values at [0.2, 0.2, 0.3, 0.3, 0.2, 0.2] as suggested by 33 . The value of denotes a word’s sentiment value in the d -th affective dimension, which can be directly extracted from the EPA and VAD lexicons.…”
Sentence-level sentiment analysis (SLSA) aims to identify the overall sentiment polarity conveyed in a given sentence. The state-of-the-art performance of SLSA has been achieved by deep learning models. However, depending on the i.i.d (independent and identically distributed) assumption, the performance of these deep learning models may fall short in real scenarios, where the distributions of training and target data are almost certainly different to some extent. In this paper, we propose a supervised solution based on the non-i.i.d paradigm of gradual machine learning (GML) for SLSA. It begins with some labeled observations, and gradually labels target instances in the order of increasing hardness by iterative knowledge conveyance. It leverages labeled samples for supervised deep feature extraction, and constructs a factor graph based on the extracted features to enable gradual knowledge conveyance. Specifically, it employs a polarity classifier to detect polarity similarity between close neighbors in an embedding space, and a separate binary semantic network to extract implicit polarity relations between arbitrary instances. Our extensive experiments on benchmark datasets show that the proposed approach achieves the state-of-the-art performance on all benchmark datasets. Our work clearly demonstrates that by leveraging DNN for feature extraction, GML can easily outperform the pure DNN solutions.
“…Since recent work has shown that encoding handcrafted affective knowledge (e.g., sentiment lexicons) can effectively enhance the training of deep DNNs for polarity classification 33 , our solution adds a new branch of sentiment attention upon the EFL-based DNN model 28 to generate a polarity-sensitive embedding space. Figure 5 The structure of EFL-based polarity classification model: 1) the boxes circled in red dashed lines represent the newly added branch of sentiment attention encoding affective knowledge; 2) the orange blocks represent the components of the newly added sentiment learning layer, the green blocks are the ones present in the original EFL model, and the blue blocks are the shared components between these two branches.…”
Section: Supervised Gml Solutionmentioning
confidence: 99%
“…Specifically, we measure the attention weight of each sentiment word by the weighted sum of its sentiment dimension values as follows: in which denotes the attention weight of a sentiment word, denotes the d -th affective dimension weight, and denotes the word’s sentiment value in the d -th affective dimension. Note that the values of represents the weights of six dimensions (namely evaluation, potency, activity, valence, arousal and dominance); in our implementation, we set their values at [0.2, 0.2, 0.3, 0.3, 0.2, 0.2] as suggested by 33 . The value of denotes a word’s sentiment value in the d -th affective dimension, which can be directly extracted from the EPA and VAD lexicons.…”
Sentence-level sentiment analysis (SLSA) aims to identify the overall sentiment polarity conveyed in a given sentence. The state-of-the-art performance of SLSA has been achieved by deep learning models. However, depending on the i.i.d (independent and identically distributed) assumption, the performance of these deep learning models may fall short in real scenarios, where the distributions of training and target data are almost certainly different to some extent. In this paper, we propose a supervised solution based on the non-i.i.d paradigm of gradual machine learning (GML) for SLSA. It begins with some labeled observations, and gradually labels target instances in the order of increasing hardness by iterative knowledge conveyance. It leverages labeled samples for supervised deep feature extraction, and constructs a factor graph based on the extracted features to enable gradual knowledge conveyance. Specifically, it employs a polarity classifier to detect polarity similarity between close neighbors in an embedding space, and a separate binary semantic network to extract implicit polarity relations between arbitrary instances. Our extensive experiments on benchmark datasets show that the proposed approach achieves the state-of-the-art performance on all benchmark datasets. Our work clearly demonstrates that by leveraging DNN for feature extraction, GML can easily outperform the pure DNN solutions.
“…7 The scores of the three dimensions for each word provide direct links to the social perceptions, actions, and emotional experiences of people for the social events. Such indexes have been proven effective for sentiment analysis [63]. By employing such indexes for the distinct words in COVID-19 myths, we are able to probe into the respective sociopsychological dimensions of the salient lexicon and to account for the social behavior of people in disseminating the COVID-19 infodemic.…”
The rampant of COVID-19 infodemic has almost been simultaneous with the outbreak of the pandemic. Many concerted efforts are made to mitigate its negative effect to information credibility and data legitimacy. Existing work mainly focuses on fact-checking algorithms or multi-class labeling models that are less aware of the intrinsic characteristics of the language. Nor is it discussed how such representations can account for the common psycho-socio-behavior of the information consumers. This work takes a data-driven analytical approach to (1) describe the prominent lexical and grammatical features of COVID-19 misinformation; (2) interpret the underlying (psycho-)linguistic triggers in terms of sentiment, power and activity based on the affective control theory; (3) study the feature indexing for anti-infodemic modeling. The results show distinct language generalization patterns of misinformation of favoring evaluative terms and multimedia devices in delivering a negative sentiment. Such appeals are effective to arouse people's sympathy toward the vulnerable community and foment their spreading behavior.
“…These energies are positive and negative, while events with negative energy are more likely to cause large-scale discussions. At this time, local problems may become public topics in the country, causing huge social panic, and sometimes even requiring government intervention [ 2 , 3 ]. The process of urbanization and the innovation of social system have increased farmers' enthusiasm to move to cities.…”
Text makes up a large portion of network data because it is the vehicle for people’s direct expression of emotions and opinions. How to analyze and mine these emotional text data has become a hot topic of concern in academia and industry in recent years. The online LDA (Latent Dirichlet Allocation) model is used in this paper to train the social hot topic data of professional migrant workers on the same time slice, and the subtopic evolution and intensity are obtained. The topic development is divided into four categories, and the classification model is created using SVM (Support Vector Machine). Instead of decision makers, a virtual human with sensibility and rationality is built using a hierarchical emotional cognitive model to solve multiobjective optimization problems interactively. It analyzes human body structure and emotional signals, and then combines them with visual and physiological signals to create multimodal emotional data. An example is used to demonstrate the effectiveness of the proposed model.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
