ABSTRACT Aspect‐level sentiment analysis (ALSA) is the process of determining the emotional polarity that people have towards aspects of topics or entities expressed in their opinions. ALSA is increasingly integrated into many practical applications to make them more user‐friendly and suitable for users' psychological and emotional trends. Therefore, the performance of ALSA methods is increasingly being studied by scientists for improvement. Various approaches have been proposed for ALSA, the latest of which is Graph Convolutional Networks (GCNs). Although they have performed well, previous GCN‐based methods still fail to capture all important features from opinions. This raises the question of whether combining ALSA‐based GCNs can improve the ability of previous methods to capture important features. This motivates us to propose the ALSA method based on the Ensemble Graph Convolutional Networks (EGCNs). The objective of the proposed method is to capture features in a manner that is both independent and joint, in order to leverage the advantages of jointly learning features while also benefiting from the strengths of learning features independently. The proposed method includes the following main steps: (i) data representation based on the BERT model; (ii) extracting syntactic, semantic and contextual features based on the ASGCN, ATGCN and ASCNN models, respectively; (iii) combining the extracted feature vectors into a general feature vector based on the fusion mechanism; (iv) sentiment analysis based on the Softmax function. To demonstrate the performance of the EGCNs model, it is experimented on three benchmark datasets and compared with the previous methods before being combined.

Sentence-level sentiment analysis is a natural language processing model growing rapidly and strongly due to its role in artificial intelligence systems. There are many approaches to developing and improving the performance of sentence-level sentiment analysis with satisfactory performance, among which deep neural network methods are notable. However, sentence-level sentiment analysis methods based on deep neural networks often have two limitations: (i) The system architecture is not deep enough; (ii) Sentences containing unclear sentiments cannot be processed. To solve the above two challenges, in this paper, we propose a new sentence-level sentiment analysis method called Fuzzy-enhanced Deep Neural Networks (FeDN2) by deepening deep neural networks by adding fuzzy and defuzzy classes. FeDN2 includes the following main layers: (i) BERT layer to convert sentences into vectors. (ii) Fuzzification layer to blur sentence vectors. (iii) Deep convolutional layers to extract high-level features. (iv) Defuzzification layer to convert high-level feature vectors into clear values. (v) Fully connected layer to learn non-linear combinations of these high-level features. (vi) Classification layer to identify the sentiment polarity of sentences. FeDN2 was tested on two benchmark datasets. The results demonstrated that it can improve the performance of previous sentence-level sentiment analysis methods based on deep neural networks.

ABSTRACT With the emergence and rapid development of social networks, an increasing amount of news has been spreading. In addition to the benefits of factual information, there are always risks associated with the dissemination of fake news and preventing the spread of fake news has been a concern for researchers. Many methods have been proposed to detect fake news, but they do not fully extract important information related to news content and context, and rarely consider modelling the simultaneous exploitation of the news context and content in fake news detection. This study proposes a method to improve the performance of fake news detection by modelling features related to news context and content. First, we combine contextualised embeddings (e.g., BERT) and dependency‐based embeddings (e.g., dependency‐based GCN) to enhance the performance of the content representations of news and reviews posting them. Second, we combine all available review texts related to news belonging to the user. Third, we explore all the reviews that other users had posted about current news by clearly creating review representations posted by the same user about the same news. This leads the model to quickly memorise all reviews related to news from one user. Finally, we model the news content features and the modelled news context features to enhance the richness of the news feature representations. Experimental results on the PolitiFact and GossipCop datasets show improvement to the state‐of‐the‐art method of more than three percentage points in the best case.

Sentiment analysis in the text plays a more and more significant role in many systems, e.g., sentence-level sentiment analysis (SLSA) in recommender and decision-making systems. Various methods have been developed to improve the performance of SLSA, the newest as graph convolutional networks (GCNs)-based methods with promising accuracy. However, it often happens that many sentences in the text contain high ambiguity of sentiment. GCNs are not capable of capturing these inherent ambiguities with performance. Meanwhile, the fuzzy logic theory can improve knowledge representation under uncertainty. These facts motivate us to propose a novel SLSA method by integrating fuzzy logic into GCNs, called the fuzzy graph convolutional network (FGCN). In this novel model, the BERT+BiLSTM model is first used to convert sentences into a matrix of contextualized vectors. Second, the fuzzy membership function is integrated into the contextualized matrix to transform it into the fuzzy contextualized representation. Third, the sentence adjacency matrix combines the syntactic information extracted from the dependency tree. Fourth, the fuzzy membership function is continuously used to transform the sentence adjacency matrix into the fuzzy adjacency matrix. After that, the defuzzy membership function is used to transform the fuzzy adjacency matrix to continuous values before deriving significant features. Next, the fuzzy adjacency matrix and the fuzzy contextualized representation are concatenated to create the final representation and fed into GCN layers to capture the high-level features of the sentence. Finally, the sentiment classifier is constructed to learn the output distribution by applying the softmax function over the final representation. Unlike conventional GCNs, the FGCN integrates fuzzy membership functions into graph convolutional layers to reduce the ambiguities of sentiment in sentence representation. This enables to achieve efficiently extracting high level sentiment features in sentences. The experimental results on benchmark datasets prove that the FGCN can enhance the performance in terms of accuracy and$F_{1}$score of SLSA in comparison with the state-of-the-art methods.

Recently, aspect-level sentiment analysis methods using graph convolutional network (GCN)-based structures with fairly good performance have been introduced. However, previous GCN-based methods often experience one of the following limitations. First, GCNs usually use edges with binary weights. However, binary weights are not helpful in many tasks. Second, these GCNs only focus on extracting node features from some single words or phrases and ignore their context in the entire sentence or paragraph or only consider the information of independent phrases when determining the relation between two graph edges overlooking the semantic relation among these phrases. Finally, no studies simultaneously use the information on the context, the semantic relation, and the sentiment knowledge among words or phrases to build GCNs for aspect-level sentiment analysis. Therefore, to resolve these limitations, in this study, we propose a new method, the CANN-SSCG model, as follows. First, we built three separate heterogeneous graphs, namely, syntax-based, semantic-based, and context-based graphs. Second, we constructed a general heterogeneous graph (SSC graph) by combining the three constructed graphs. We then converted the nodes of the SSC graph into sentence vectors using a GCN with two layers (creating an SSC-GCN). Finally, we used a convolutional neural network algorithm with attention to position embeddings (CANN) on the output of the SSC-GCN model for aspect-level sentiment analysis. The experiments, which used three different datasets, including reviews and tweets, showed that the proposed method yields promising results based on the F1score.

From the end of 2019, numerous comments and opinions relating to the COVID-19 pandemic have been posted on Twitter. The number of opinions rapidly increased since the countries began implementing social isolation and reduction. In these comments, users often express different emotions regarding COVID-19 signs and symptoms, the majority of which are sadness and fear sentiments. It is important to determine the symptom effect level for the emotions of symptomatic persons based on their opinions. However, no study analyzes the tweets' sentiment related to the COVID-19 topic to predict the symptoms effect level. Therefore, in this study, we present a method to predict the symptoms effect level based on the sentiment analysis of symptomatic persons according to the following steps. First, the sentiments in tweets are analyzed by using a combination of the text representation model and convolutional neural network. Second, a topic modeling model is built based on the latent Dirichlet allocation algorithm to group symptoms into small clusters that conform to sadness and fear sentiments. Finally, the symptom effect level is predicted based on the probability distribution of the symptoms in each sentiment cluster. Experiments using tweets promise that the proposed method achieves significant results toward the accuracy and obtained information.

Social networks are a very popular channel for people to communicate with, to find, to reference other users before making decisions, especially those concerning purchase. How can users’ opinions within social networks be used in making decisions cost-effective and reliable? In this paper, we propose an approach for supporting decision-making based on measuring the user satisfaction level by analyzing the sentiment of aspects and mining the fuzzy decision trees. Our proposal has been proved to overcome some of the disadvantages of previous methods. Specifically, we consider the fuzzy sentiments of users for aspects and the effects of user satisfaction, dissatisfaction, and hesitation for decision-making. The proposed method comprises four main stages. The first stage identifies a topic, which the user is interested. In the second stage, aspects of the topic and their sentiments within tweets are extracted. At the third stage, the user satisfaction level is calculated according to each kind of sentiment identified in the second step. Finally, a decision matrix is constructed, and the fuzzy decision tree is built to generate a set of rules for supporting users in decision-making. The experiments using tweets show that the proposed method achieves promising results regarding the accuracy and gained information.

Twitter is an online social networking service with millions of users and an impressive flow of messages that are published and spread daily through interactions among users. There are different types of users on Twitter; therefore, determining the most important users in each topic is highly challenging. Hence, it is necessary to define efficient computed measures to classify users according to the criteria of relevance and the possibility of representing reality. Although several studies have considered identifying the user influence, user popularity, or user activity in a social network, relatively less focus has been on measuring and predicting important users in case of a topic. In this study, we have proposed a method to determine an important user based on the activities related to each topic on Twitter by combining the measures related to user influence, user activity, and user popularity. The results verified the effectiveness of our proposed approach for the identification of important users in each topic.

Owing to the development and dissemination of Twitter, an increasing number of users' opinions about various topics are being published on Twitter and have become a significant data source for numerous applications; one of the most popular is tweet sentiment analysis. Many researchers have tried to solve this problem with different methods. However, previous studies have only focused on sentiment analysis of general tweets without considering a divide-and-conquer strategy. Meanwhile, a large number of tweets contains fuzzy sentiment phrases. Thus, effectively solving fuzzy sentiment phrases may help to significantly improve the performance of sentiment analysis methods. In this study, we concentrate only on the detection and sentiment analysis problem of a specific tweet type that contains fuzzy sentiment phrases. The results show that the proposed method performs relatively well in both tasks.