Fig 1 - uploaded by Gaël Guibon
Content may be subject to copyright.
Text resemblance scores. The closer they are, the more difficult they are to discriminate.

Text resemblance scores. The closer they are, the more difficult they are to discriminate.

Source publication
Chapter
Full-text available
The information spread through the Web influences politics, stock markets, public health, people’s reputation and brands. For these reasons, it is crucial to filter out false information. In this paper, we compare different automatic approaches for fake news detection based on statistical text analysis on the vaccination fake news dataset provided...

Similar publications

Article
Full-text available
Multilingual based voice activated human computer interaction systems are currently in high demand. The Spoken Language Identification Unit (SPLID) is an inevitable front end unit of such a multilingual system. These systems will be a great boon to a country like India where around 24 official languages are spoken. Deep learning architectures for s...

Citations

... Another important task addressed in fake news detection is satire detection, with the methods ranging from convolutional neural networks (CNNs) (Guibon et al., 2019) to adversarial training (McHardy et al., 2019) and BERT-based architectures with long-short-term memory (LSTM) (Pandey and Singh, 2022;Liu and Xie, 2021) and CNN (Kaliyar et al., 2021) layers on top. ...
Preprint
Full-text available
This paper describes our approach for SemEval-2023 Task 3: Detecting the category, the framing, and the persuasion techniques in online news in a multi-lingual setup. For Subtask 1 (News Genre), we propose an ensemble of fully trained and adapter mBERT models which was ranked joint-first for German, and had the highest mean rank of multi-language teams. For Subtask 2 (Framing), we achieved first place in 3 languages, and the best average rank across all the languages, by using two separate ensembles: a monolingual RoBERTa-MUPPETLARGE and an ensemble of XLM-RoBERTaLARGE with adapters and task adaptive pretraining. For Subtask 3 (Persuasion Techniques), we train a monolingual RoBERTa-Base model for English and a multilingual mBERT model for the remaining languages, which achieved top 10 for all languages, including 2nd for English. For each subtask, we compare monolingual and multilingual approaches, and consider class imbalance techniques.
... Another important task addressed in fake news detection is satire detection, with the methods ranging from convolutional neural networks (CNNs) (Guibon et al., 2019) to adversarial training (McHardy et al., 2019) and BERT-based architectures with long-short-term memory (LSTM) (Pandey and Singh, 2022;Liu and Xie, 2021) and CNN (Kaliyar et al., 2021) layers on top. ...
Article
Full-text available
In many crops worldwide, including hazelnuts, the majority of stages in production and delivery to end-users are conducted either manually or with machine equipment lacking the advancements brought by technology. Non-destructive, fast, and reliable methods, particularly deep learning algorithms, have emerged as prominent techniques for determining product quality and classification in fruits, vegetables, and cereal products in recent years. This study aims to classify hazelnuts using deep learning algorithms, thereby minimizing the labor, time, and cost expended during the sorting process. Hazelnut images were obtained from Giresun, Ordu, and Van hazelnut varieties. The dataset consists of 1165 images of Giresun, 1324 images of Ordu, and 1138 images of Van hazelnut varieties. The classification was performed using deep learning models such as InceptionV3 and ResNet50. To combine the classification capabilities of the models, an InceptionV3 + ResNet50 data fusion model was created using the data fusion method. In addition, feature reduction processes were conducted by adding a convolutional layer to the data fusion model to decrease the number of features. The classification was conducted using a total of 3627 images, resulting in a 100% classification accuracy. Furthermore, the classification times of all models were analyzed. Based on these analyses, the 1024 reduced features data fusion model with 100% classification accuracy exhibited the shortest classification time. This model was selected, and a mobile application was developed for easy on-field hazelnut classification. The hazelnut classification performed using deep learning algorithms in the application will facilitate the work of both non-experts and professionals in industrial and personal domains. Through these methods, patents for products and devices developed for use in different industries can be obtained, thereby increasing the economic value added of our country.
Chapter
Fake news is a severe problem on social media networks, with confirmed detrimental consequences for individuals and organizations. As a result, detecting false news is a significant difficulty. In this way, the topic of new fakes and their proper detection are crucial. General knowledge states that the receiver of information must verify the sources. However, the creation of new information can be a difficult problem that requires more than a single viewpoint based on a news source. The objective of this paper is to evaluate the performance of six deep learning models for fake news detection including CNN, LSTM, Bi-LSTM, HAN, Conv-Han and Bert. We find that BERT and similar pre-trained models perform the best for fake news detection, the models are described below with their experimental setups.The models are examined against ISOT [22, 30] datasets. KeywordsFake news detectionNatural language processingDeep learning