News Dataset Research Articles

A hybrid model for text classification using part-of-speech features

In the field of text classification, current research ignores the role of part-of-speech features, and the multi-channel model that can learn richer text information compared to a single model. Moreover, the method based on neural network models to achieve final classification, using fully connected layer and Softmax layer can be further improved and optimized. This paper proposes a hybrid model for text classification using part-of-speech features, namely PAGNN-Stacking1. In the text representation stage of the model, introducing part-of-speech features facilitates a more accurate representation of text information. In the feature extraction stage of the model, using the multi-channel attention gated neural network model can fully learn the text information. In the text final classification stage of the model, this paper innovatively adopts Stacking algorithm to improve the fully connected layer and Softmax layer, which fuses five machine learning algorithms as base classifier and uses fully connected layer Softmax layer as meta classifier. The experiments on the IMDB, SST-2, and AG_News datasets show that the accuracy of the PAGNN-Stacking model is significantly improved compared to the benchmark models.

A Novel Taxonomy for Arabic Fake News Datasets

A Novel Taxonomy for Arabic Fake News Datasets

An Innovative Approach for Fake News Detection using Machine Learning

This research aims to increase people's awareness of fake news on online social networks and help them determine the reliability of information they consume. It investigates methods for detecting fake news sources, authors, and subjects on online social networks. The project uses an open-source online dataset of fake and real news to determine the credibility of news. Various text feature extraction techniques and classification algorithms are reviewed, with the Support Vector Machine (SVM) linear classification algorithm using TF-IDF feature extraction achieving the highest accuracy of 99.36%. Random Forest (RF) and Naive Bayes (NB) had accuracy scores of 98.25% and 94.74%, respectively.

Concept-drift detection index based on fuzzy formal concept analysis for fake news classifiers

Unpredictable changes in the underlying distribution of the streaming data over time are known as concept drift. The development of procedures and techniques for drift detection, interpretation, and adaptation is central to concept-drift research. Data research has demonstrated that machine learning in a concept-drift environment produces poor learning results if drift is not handled. This study focuses on defining the concept-drift detection index to predict when the performance of a machine learning model for text-stream classifiers is low. It proposes an index that relies on the Fuzzy Formal Concept Analysis theory. The index exploits the formal lattice to understand whether new incoming facts (e.g., news) are well represented in the training data used to build the machine-learning model. Fake news was deemed ideal for testing this new measure because its typical application scenario required handling a stream of unstructured content and concept-drift awareness. Experiments on three news datasets revealed a relevant correlation (i.e., 73.9 %, 80.8 %, and 81 %) between the Accuracy of Random Forest (RF), Naïve Bayes (NB), and Passive Aggressive (PA) models, respectively, and the proposed index. This strong correlation suggests that the new index can avoid incorrect classifications and help in retraining decisions.

LANCER: A Lifetime-Aware News Recommender System

From the observation that users reading news tend to not click outdated news, we propose the notion of 'lifetime' of news, with two hypotheses: (i) news has a shorter lifetime, compared to other types of items such as movies or e-commerce products; (ii) news only competes with other news whose lifetimes have not ended, and which has an overlapping lifetime (i.e., limited competitions). By further developing the characteristics of the lifetime of news, then we present a novel approach for news recommendation, namely, Lifetime-Aware News reCommEndeR System (LANCER) that carefully exploits the lifetime of news during training and recommendation. Using real-world news datasets (e.g., Adressa and MIND), we successfully demonstrate that state-of-the-art news recommendation models can get significantly benefited by integrating the notion of lifetime and LANCER, by up to about 40% increases in recommendation accuracy.

LFWE: Linguistic Feature Based Word Embedding for Hindi Fake News Detection

It is essential for research communities to investigate ways for authenticating news. The use of linguistic feature based analysis to automatically detect false news is gaining popularity among the scientific community. However, such techniques are exclusively created for English, leaving low-resource languages like Hindi behind. To address this issue, we constructed a novel annotated Hindi Fake News (HinFakeNews) dataset of roughly 33,300 articles that can be utilized to develop autonomous fake news detection systems. This work provides a two-stage benchmark model for identifying fake news in Hindi using machine learning. The proposed model, LFWE (Linguistic Feature Based Word Embedding), generates word embedding over linguistic features. This article focuses on 23 key linguistic features (15 extracted and 08 derived) for successful detection of Hindi fake news. These features are grouped as lexical, semantic, syntactic, psycho-linguistic, readability, and quantity features. The contribution is twofold. In the first phase, the dataset is preprocessed and linguistic features are extracted. In the second phase, feature sets are generated as word embeddings, and an Ensemble voting classification is carried out on the feature sets. According to experimental findings, the LFWE model accurately detects and classifies fake news in Hindi with an accuracy of 98.49%.

Identifying multimodal misinformation leveraging novelty detection and emotion recognition.

With the growing presence of multimodal content on the web, a specific category of fake news is rampant on popular social media outlets. In this category of fake online information, real multimedia contents (images, videos) are used in different but related contexts with manipulated texts to mislead the readers. The presence of seemingly non-manipulated multimedia content reinforces the belief in the associated fabricated textual content. Detecting this category of misleading multimedia fake news is almost impossible without relevance to any prior knowledge. In addition to this, the presence of highly novel and emotion-invoking contents can fuel the rapid dissemination of such fake news. To counter this problem, in this paper, we first introduce a novel multimodal fake news dataset that includes background knowledge (from authenticate sources) of the misleading articles. Second, we design a multimodal framework using Supervised Contrastive Learning (SCL) based novelty detection and Emotion Prediction tasks for fake news detection. We perform extensive experiments to reveal that our proposed model outperforms the state-of-the-art (SOTA) models.

Fake news detection based on a hybrid BERT and LightGBM models

With the rapid growth of social networks and technology, knowing what news to believe and what not to believe become a challenge in this digital era. Fake news is defined as provably erroneous information transmitted intending to defraud. This kind of misinformation poses a serious threat to social cohesion and well-being, since it fosters political polarisation and can destabilize trust in the government or the service provided. As a result, fake news detection has emerged as an important field of study, with the goal of identifying whether a certain piece of content is real or fake. In this paper, we propose a novel hybrid fake news detection system that combines a BERT-based (bidirectional encoder representations from transformers) with a light gradient boosting machine (LightGBM) model. We compare the performance of the proposed method to four different classification approaches using different word embedding techniques on three real-world fake news datasets to validate the performance of the proposed method compared to other methods. The proposed method is evaluated to detect fake news based on the headline-only or full text of the news content. The results show the superiority of the proposed method for fake news detection compared to many state-of-the-art methods.

Towards COVID-19 fake news detection using transformer-based models

The COVID-19 pandemic has resulted in a surge of fake news, creating public health risks. However, developing an effective way to detect such news is challenging, especially when published news involves mixing true and false information. Detecting COVID-19 fake news has become a critical task in the field of natural language processing (NLP). This paper explores the effectiveness of several machine learning algorithms and fine-tuning pre-trained transformer-based models, including Bidirectional Encoder Representations from Transformers (BERT) and COVID-Twitter-BERT (CT-BERT), for COVID-19 fake news detection. We evaluate the performance of different downstream neural network structures, such as CNN and BiGRU layers, added on top of BERT and CT-BERT with frozen or unfrozen parameters. Our experiments on a real-world COVID-19 fake news dataset demonstrate that incorporating BiGRU on top of the CT-BERT model achieves outstanding performance, with a state-of-the-art F1 score of 98%. These results have significant implications for mitigating the spread of COVID-19 misinformation and highlight the potential of advanced machine learning models for fake news detection.

The Effects of Preprocessing on Turkish and English News Data

In a standard text classification (TC) study, preprocessing is one of the key components to improve performance. This study aims to look at how preprocessing effects TC according to news text, text language, and feature selection. All potential combinations of commonly used preprocessing techniques are compared on one domain, namely news data, and in two different news datasets for this aim. Preprocessing technique contributions to classification performance at multiple feature sizes, possible interconnections among these techniques, and technique dependency on corresponding languages are all evaluated in this way. Using best combinations of preprocessing techniques rather than using or not using them all, experimental studies on public datasets reveals that, choosing best combinations of preprocessing techniques can improve classification accuracy significantly.

F-FNC: Privacy concerned efficient federated approach for fake news classification

Fake news and manipulated information affect the social, economic and emotional growth of the world's population.For the identification of fake news, several classification systems are available, but no such system was found fast, secure and reliable as per the need of the hour. In this work, an efficient framework based on the federated architecture for the classification of fake news was proposed, while maintaining the data privacy constraints for sensitive text news datasets. The proposed federated-Fake New Classification (f-FNC) framework utilized the distributed client–server architecture with data privacy of all client or connected edge devices. For the testing and evaluation of the proposed f-FNC framework, the non-identical data was gathered from several online resources and was disseminated in a pre-processed format. To test the validity of federated deep learning models, the experiments were performed under various scenarios such as traditional learning, federated learning single client, and federated learning multi-clients. The performance of f-FNC framework was evaluated through various computational parameters such as accuracy and loss validation along with available resource parameters including CPU and RAM utilization. It was observed from the resultant outcome that the proposed f-FNC framework worked significantly well in both single-client and multi-client (N-clients) scenarios in comparison to traditional distributed deep learning based classifiers. The additional features of low cost and data-privacy of edge devices with limited resources made this proposed framework unique and the best alternative to existing fake news classifier tools.

SSM: Stylometric and semantic similarity oriented multimodal fake news detection

Over the years, there has been a rise in the number of fabricated and fake news stories that utilize both textual and visual information formats. This coincides with the increased likelihood that users will acquire their news from websites and social media platforms. While there has been various research into the detection of fake news in text using machine learning techniques, less attention has been paid to the problem of multimedia data fabrication. In this paper, we propose a Stylometric, and Semantic similarity oriented for Multimodal Fake News Detection (SSM). There are five distinct modules that make up our methodology: Firstly, we used a Hyperbolic Hierarchical Attention Network (Hype-HAN) for extracting stylometric textual features. Secondly, we generated the news content summary and computed the similarity between Headline and summary. Thirdly, semantic similarity is computed between visual and textual features. Fourthly, images are analyzed for forgery. Lastly, the extracted features are fused for final classification. We have tested SSM framework on three standard fake news datasets. The results indicated that our suggested model has outperformed the base line and state-of-the-art methods and is more likely to detect fake news in complex environments.

A Novel Approach to Detection of Fake News in Online Communities

Fake news serving various political and commercial agendas has emerged on the web and spread rapidly in recent years, thanks in large part to the proliferation of online social networks. People who use informal online groups are especially vulnerable to the sneaky effects of deceptive language used in fake news on the internet, which has far-reaching effects on real society. To make information in informal online communities more reliable, it is important to be able to spot fake news as soon as possible. The goal of this study is to look at the criteria, methods, and calculations that are used to find and evaluate fake news, content, and topics in unstructured online communities. This research is mostly about how vague fake news is and how many connections there are between articles, writers, and topics. In this piece, we introduce FAKEDETECTOR, a novel controlled graph neural network. FAKEDETECTOR creates a deep diffusive organization model based on a wide range of explicit and specific attributes extracted from the textual content, allowing it to simultaneously learn the models of reports, authors, and topics. The complete version of this paper provides exploratory results from extensive experiments on a real fake news dataset designed to distinguish FAKEDETECTOR from two state-of-the-art algorithms.

Abstractive Text Summarization using Pre-Trained Language Model "Text-to-Text Transfer Transformer (T5)"

Automatic Text Summarization (ATS) is one of the utilizations of technological sophistication in terms of text processing assisting humans in producing a summary or key points of a document in large quantities. We use Indonesian language as objects because there are few resources in NLP research using Indonesian language. This paper utilized PLTMs (Pre-Trained Language Models) from the transformer architecture, namely T5 (Text-to-Text Transfer Transformer) which has been completed previously with a larger dataset. Evaluation in this study was measured through comparison of the ROUGE (Recall-Oriented Understudy for Gisting Evaluation) calculation results between the reference summary and the model summary. The experiments with the pre-trained t5-base model with fine tuning parameters of 220M for the Indonesian news dataset yielded relatively high ROUGE values, namely ROUGE-1 = 0.68, ROUGE-2 = 0.61, and ROUGE-L = 0.65. The evaluation value worked well, but the resulting model has not achieved satisfactory results because in terms of abstraction, the model did not work optimally. We also found several errors in the reference summary in the dataset used.

Multiverse: Multilingual Evidence for Fake News Detection

The rapid spread of deceptive information on the internet can have severe and irreparable consequences. As a result, it is important to develop technology that can detect fake news. Although significant progress has been made in this area, current methods are limited because they focus only on one language and do not incorporate multilingual information. In this work, we propose Multiverse-a new feature based on multilingual evidence that can be used for fake news detection and improve existing approaches. Our hypothesis that cross-lingual evidence can be used as a feature for fake news detection is supported by manual experiments based on a set of true (legit) and fake news. Furthermore, we compared our fake news classification system based on the proposed feature with several baselines on two multi-domain datasets of general-topic news and one fake COVID-19 news dataset, showing that (in combination with linguistic features) it yields significant improvements over the baseline models, bringing additional useful signals to the classifier.

One-class learning for fake news detection through multimodal variational autoencoders

Machine learning methods to detect fake news typically use textual features and Binary or Multi-class classification. However, accurately labeling a large news set is still a very costly process. On the other hand, one of the prominent approaches is One-Class Learning (OCL). OCL requires only the labeling of fake news, minimizing data labeling efforts. Although we eliminate the need to label non-interest news, the efficiency of OCL algorithms depends directly on the data representation model adopted. Most existing methods in the OCL literature explore representations based on one modality to detect fake news. However, different text features can be the reason for the news to be fake, such as topic or linguistic features. We model this behavior as different modalities for news to represent different textual feature sets. Thus, we present the MVAE-FakeNews, a multimodal method to represent the texts in the fake news detection through OCL that learns a new representation from the combination of promising modalities for news data: text embeddings, topic, and linguistic information. We used real-world fake news datasets in Portuguese and English in the experimental evaluation. Results show that MVAE-FakeNews obtained a better F1-Score and AUC-ROC, outperforming another fourteen methods in three datasets and getting competitive results on the other three. Moreover, our MVAE-FakeNews, with only 3% of labeled fake news, obtained comparable or higher results than other methods. To improve the experimental evaluation, we also propose the Multimodal LIME for OCL to identify how each modality is associated with the fake news class.

Which Emotions of Social Media Users Lead to Dissemination of Fake News: Sentiment Analysis Towards Covid-19 Vaccine

The use of social media as a news source is quite common today. However, the fact that the news encountered on social media are accepted as true without questioning or checking their validity is one of the main reasons for the dissemination of fake news. For the social media ecosystem, the question arises as to which emotion is more effective in spreading fake news, as the accuracy and validity of the news are under the control of opinions and emotions rather than evidence-based data. From this point of view, our study investigates whether there is a relationship between users’ reaction to the news and the prevalence of the news. In our study, sentiment analysis was conducted on the reactions of Twitter users to fake news about the COVID-19 vaccine between December 31, 2019 and July 30, 2022. To fully assess whether there is a relationship between the reactions and the prevalence of the news, the spread of real news published in the same period in addition to fake news is also taken into consideration. Fake and real news comments, which were selected in different degrees of prevalence from the most to the least, were examined comparatively. In the study, where text mining techniques were used for text pre-processing, analysis was carried out with NLP techniques. In 83% of the fake news datasets and 91% of the overall news datasets considered in the study, negative emotion was more dominant than other emotions, and it was observed that as negative comments increased, fake news spread more as well as real news. While neutral comments have no effect on prevalence, users who comment on fake news for fun significantly increase the prevalence. Finally, to reveal bot activity NLP techniques were applied.

Detecting Fake News Using Machine Learning

Fake news has had a significant effect on society and politics. To aid in combating the spread of misinformation, we worked to develop a machine learning algorithm that could detect fake news based on textual data. We used a count vectorizer to vectorize our text which we then inputted into Logistic Regression, Support Vector Machine (SVM), and Linear Support Vector Classifier (SVC) models. The greatest accuracy score achieved was 99.97% with the Linear SVC. We discovered however that there was a significant difference in how the real and fake news datasets were constructed that would not translate into real life: the true news articles contained quotation marks, apostrophes, and dashes while these characters were not present in the fake news articles. Because of this, we also developed a more applicable Logistic Regression model removing these specific characters from the dataset all together with an accuracy score of 98.4%.

APSN: Adversarial Pseudo-Siamese Network for Fake News Stance Detection

Fake news is a longstanding issue that has existed on the social network, whose negative impact has been increasingly recognized since the US presidential election. During the election, numerous fake news about the candidates distributes vastly in the online social networks. Identifying inauthentic news quickly is an essential purpose for this research to enhance the trustworthiness of news in online social networks, which will be the task studied in this paper. The fake news stance detection can contribute to detect a startling amount of fake news, which aims at evaluating the relevance between the headline and text bodies. There exists a significant difference between news article headline and text body, since headlines with several key phrases are usually much shorter than the text bodies. Such an information imbalance challenge may cause serious problems for the stance detection task. Furthermore, news article data in online social networks is usually exposed to various types of noise and can be contaminated, which poses more challenges for the stance detection task. In this paper, we propose a novel fake news stance detection model, namely Adversarial Pseudo-Siamese Network model (APSN), to solve these challenges. With coupled input components with imbalanced parameters, APSN can learn and compute feature vectors and similarity score of news article headlines and text bodies simultaneously. In addition, by adopting adversarial setting, besides the regular training set, a set of noisy training instances will be generated and fed to APSN in the learning process, which can significantly enhance the robustness of the model. Extensive experiments have been conducted on a real-world fake news dataset, and the experimental results reveal that the presented model exceeds compared suspicious information detection models with significant advantages.

Canopy-MMD Text Clustering Algorithm Based on Simulated Annealing and Canopy Optimization

<p>Aiming at the problems that traditional K-means text clustering cannot automatically determine the number of clusters and is sensitive to initial cluster centers, this paper proposes a Canopy-MMD text clustering algorithm based on simulated annealing and silhouette coefficient optimization. The algorithm uses the simulated annealing algorithm combined with the silhouette coefficient to optimize the Canopy algorithm to find the optimal number of clusters, and uses the optimal number of clusters to determine the scale coefficient in the MMD algorithm, and finally achieves a better text clustering effect. The Sohu News dataset of Sogou Lab is experimentally analyzed and compared with the clustering results obtained by traditional K-means and algorithms in the literature. The experimental results show that the clustering performance of the algorithm is better than the traditional K-means algorithm and the algorithm in the literature, and the accuracy, precision, recall and F value are improved by 8.02%, 8.91%, 8.02%, 9.51% compared with the traditional K-means algorithm, which can be widely used in fields such as text mining, knowledge graph and natural language processing.</p> <p> </p>