Search

There is an increased demand for detecting online hate speech, especially with the recent changing policies of hate content and free-of-speech right of online social media platforms. Detecting hate speech will reduce its negative impact on social media users. A lot of effort in the Natural Language Processing (NLP) field aimed to detect hate speech in general or detect specific hate speech such as religion, race, gender, or sexual orientation. Hate communities tend to use abbreviations, intentional spelling mistakes, and coded words in their communication to evade detection, which adds more challenges to hate speech detection tasks. Word representation from its domain will play an increasingly pivotal role in detecting hate speech. This paper investigates the feasibility of leveraging domain-specific word embedding as features and a bidirectional LSTM-based deep model as a classifier to automatically detect hate speech. This approach guarantees that the word is assigned its negative meaning, which is a very helpful technique to detect coded words. Furthermore, we investigate the use of the transfer learning language model (BERT) on the hate speech problem as a binary classification task as it provides high-performance results for many NLP tasks. The experiments showed that domain-specific word embedding with the bidirectional LSTM-based deep model achieved a 93% f1-score, while BERT achieved 96% f1-score on a combined balanced dataset from available hate speech datasets. The results proved that the performance of pre-trained models is influenced by the size of the trained data. Although there is a huge variation in the corpus size, the first approach achieved a very close result compared to BERT, which is trained on a huge data corpus, this is because it is trained on data related to the same domain. The first approach was very helpful to detect coded words while the second approach achieved better performance because it is trained on much larger data. To conclude, it is very helpful to build large pre-trained models from rich domains specific content in current social media platforms. [ABSTRACT FROM AUTHOR]

Social media platforms have revolutionized traditional communication techniques by allowing people to connect instantaneously, openly, and frequently. As people use social media to share personal stories and express their opinions, negative emotions such as thoughts of death, self-harm, and hardship are commonly expressed, particularly among younger generations. Accordingly, the use of social media to detect suicidality may help provide proper intervention that will ultimately deter the spread of self-harm and suicidal ideation on social media. To investigate the automated detection of suicidal thoughts in Arabic tweets, we developed a novel Arabic suicidal tweet dataset, examined several machine learning models trained on word frequency and embedding features, and investigated the performance of pre-trained deep learning models in identifying suicidal sentiment. The results indicate that the support vector machine trained on character n-gram features yields the best performance among conventional machine learning models, with an accuracy of 86% and F1 score of 79%. In the subsequent deep learning experiment, AraBert outperformed all other machine and deep learning models with an accuracy of 91% and F1-score of 88%, significantly improving the detection of suicidal ideation in the dataset. To the best of our knowledge, this study represents the first attempt to compile an Arabic suicidality detection dataset from Twitter and to use deep learning to detect suicidal sentiment in Arabic posts. [ABSTRACT FROM AUTHOR]

Social networks have generated immense amounts of data that have been successfully utilized for research and business purposes. The approachability and immediacy of social media have also allowed ill-intentioned users to perform several harmful activities that include spamming, promoting, and phishing. These activities generate massive amounts of low-quality content that often exhibits duplicate, automated, inappropriate, or irrelevant content that subsequently affects users' satisfaction and imposes a significant challenge for other social media-based systems. Several real-time systems were developed to tackle this problem by focusing on filtering a specific kind of low-quality content. In this paper, we present a fine-grained real-time classification approach to identify several types of low-quality tweets (i.e., phishing, promoting, and spam tweets) written in Arabic. The system automatically extracts textual features using deep learning techniques without relying on hand-crafted features that are often time-consuming to be obtained and are tailored for a single type of low-quality content. This paper also proposes a lightweight model that utilizes a subset of the textual features to identify spamming Twitter accounts in a real-time setting. The proposed methods are evaluated on a real-world dataset (40, 000 tweets and 1, 000 accounts), showing superior performance in both models with accuracy and F1-scores of 0.98. The proposed system classifies a tweet in less than five milliseconds and an account in less than a second. • This research shows that training a deep learning model on a dataset that includes several types of low-quality tweet can be an efficient solution to filter such content on a real-time setting. • Two embedding methods (word- and character-level) are compared for the task of classifying tweets in either a legitimate or low-quality class using a dataset (40,000 tweets) collected through this project. • We also show that Twitter account can be efficiently classified into spam or genuine profile using only the textual data of its recent tweets and a deep learning model. [ABSTRACT FROM AUTHOR]

Distracted driving plays a significant role in road accidents worldwide. Detecting distracted driving in real time is a significant challenge faced by law enforcement officers. Although efforts are continuing to identify this type of infraction in an automated manner, most of the techniques proposed for this problem use images obtained internally from vehicles. Few studies have looked into using roadway images collected in naturalistic situations. We propose in this research a complete and fully automated deep-learning approach that locates vehicles in roadway images, detects and extracts license plate numbers, detects the windshield region, and classifies images into predefined violations. The proposed approach is complete in that it does not rely on roadway license plate systems to localize the vehicle of interest and extract the license plate numbers. The model performance-both overall and in each stage-was evaluated, achieving 90 % overall classification accuracy. The model is trained and tested on a real-world local dataset of 10,000 images. The dataset used in this study is the first dataset acquired from roadway license plate cameras in the Middle Eastern region showing unique variations of plate forms, driving habits, regional attire, and weather conditions. [ABSTRACT FROM AUTHOR]

This paper describes a novel method for building affectively intelligent human-interactive agents. The method is based on a key sociological insight that has been developed and extensively verified over the last twenty years, but has yet to make an impact in artificial intelligence. The insight is that resource bounded humans will, by default, act to maintain affective consistency. Humans have culturally shared fundamental affective sentiments about identities, behaviours, and objects, and they act so that the transient affective sentiments created during interactions confirm the fundamental sentiments. Humans seek and create situations that confirm or are consistent with, and avoid and suppress situations that disconfirm or are inconsistent with, their culturally shared affective sentiments. This “ affect control principle ” has been shown to be a powerful predictor of human behaviour. In this paper, we present a probabilistic and decision-theoretic generalisation of this principle, and we demonstrate how it can be leveraged to build affectively intelligent artificial agents. The new model, called BayesAct , can maintain multiple hypotheses about sentiments simultaneously as a probability distribution, and can make use of an explicit utility function to make value-directed action choices. This allows the model to generate affectively intelligent interactions with people by learning about their identity, predicting their behaviours using the affect control principle, and taking actions that are simultaneously goal-directed and affect-sensitive. We demonstrate this generalisation with a set of simulations. We then show how our model can be used as an emotional “plug-in” for artificially intelligent systems that interact with humans in two different settings: an exam practice assistant (tutor) and an assistive device for persons with a cognitive disability. [ABSTRACT FROM AUTHOR]