Hate Speech Detection In Italian Text Using Deep Learning Techniques

The purpose of this project is to develop a comprehensive Hate Speech Detection and Sentiment Analysis system using both Machine Learning and Deep Learning techniques. The project aims to create a robust and accurate system that can automatically identify hate speech in text data and perform sentiment analysis to determine the emotions and opinions expressed in the text. The project is designed to address the growing concern over the spread of hate speech and offensive content online. By implementing an automated detection system, it can help social media platforms, content moderators, and online communities to proactively identify and remove harmful content, fostering a safer and more inclusive online environment. Additionally, sentiment analysis plays a crucial role in understanding public opinions, customer feedback, and social media trends. By accurately predicting sentiment, businesses can make data-driven decisions, improve customer satisfaction, and gain valuable insights into consumer preferences. This project focuses on Hate Speech Detection and Sentiment Analysis using both Machine Learning and Deep Learning techniques. It begins with exploring the dataset, analyzing feature distributions, and predicting sentiment using Machine Learning models like Logistic Regression, Support Vector Machines, K-Nearest Neighbors, Decision Trees, Random Forests, Gradient Boosting, Extreme Gradient Boosting, Light Gradient Boosting, and AdaBoost, while optimizing their performance through Grid Search for hyperparameter tuning. Subsequently, Deep Learning LSTM and 1D CNN models are implemented for sentiment analysis to capture long-term dependencies and local patterns in the text data. The project starts with exploring the dataset, understanding its structure, and analyzing the distribution of classes for hate speech and sentiment labels. This initial step allows us to gain insights into the dataset and potential challenges. After exploring the data, the distribution of text features, such as word frequency and sentiment scores, is analyzed to identify any patterns or biases that could impact the model's performance. The dataset is then divided into training, validation, and testing sets to evaluate the models' generalization capabilities. Early stopping techniques are utilized during training to prevent overfitting and enhance model generalization. Performance evaluation involves calculating metrics like accuracy, precision, recall, and F1-score to gauge the models' effectiveness. Confusion matrices and visualizations provide further insights into model predictions and potential areas for improvement. A graphical user interface (GUI) is developed using PyQt to facilitate user interaction with the Hate Speech Detection and Sentiment Analysis system. Before training the Deep Learning models, the text data is tokenized and padded for uniform input sequences. The dataset is split into training and validation sets for model evaluation, and early stopping is used to prevent overfitting during training. The final system combines predictions from both Machine Learning and Deep Learning models to provide robust sentiment analysis results. The PyQt GUI allows users to input text and receive real-time sentiment analysis predictions. The LSTM and 1D CNN models, along with their optimized hyperparameters, are saved and deployed for future sentiment analysis tasks. Users can interact with the GUI, analyze sentiment in different texts, and provide feedback for continuous improvement of the Hate Speech Detection and Sentiment Analysis system.

We are witnessing a proliferation of hate speech on social media targeting individuals for their protected characteristics, including ethnicity, religion, gender, and nationality. Our research focuses on devising effective Arabic hate and offensive speech detection frameworks to address this serious issue. In the first part of the thesis, we aim to improve Arabic hate speech detection systems and present our efforts at building binary and multi-class (3-class and 6-class) hate and offensive speech datasets using four robust extraction strategies that we implement based on the four types of hate: religion, ethnicity, nationality, and gender. Next, we develop several 2-class, 3-class, and 6-class machine and deep learning classification models that we train on different feature spaces using a variety of feature extraction techniques. We also investigate how we can develop single and ensemble machine and deep learning models for hate speech detection and conduct extensive experiments to assess the performance of the various learned models on unseen data. The performance outcome is very encouraging compared to prior hate speech studies carried out on Arabic and English corpora. Furthermore, we examine the word-embedding models' effect on the neural network's performance since they were not adequately examined in the literature. Through 2-class, 3-class, and 6-class classification tasks, we investigate the impact of both word-embedding models and neural network architectures on predictive accuracy. We first train several word-embedding models on a large-scale Arabic text corpus. Next, based on our Arabic hate and offensive speech dataset, we train multiple neural networks for each detection task using the pre-trained word embeddings. This task yields a large number of learned models, which allows conducting an exhaustive comparison. One key for improving hate speech detection performance is to have a textual training corpus that is vast and confidently labeled. Thus, in the second part of this thesis, we explore how we can improve hate speech detection and leverage the abundant social media content based on the recent success of semisupervised learning techniques. In particular, we explore two new research directions: (1) adopting semi-supervised self-learning to create a large-scale hate speech corpus and use it to improve hate speech detection models; and (2) build ensemble-based semi-supervised learning systems based on the machine and deep learning models. We empirically demonstrate the effectiveness of these approaches and show that our semi-supervised approaches improve classification performance over supervised hate speech classification methods.

The ninth edition of the Italian Conference on Computational Linguistics (CLiC-it 2023) was held from 30th November to 2nd December 2023 at Ca’ Foscari University of Venice, in the beautiful venue of the Auditorium Santa Margherita - Emanuele Severino. After the edition of 2020, which was organized in fully virtual mode due to the health emergency related to Covid-19, and CLiC-it 2021, which was held in hybrid mode, with CLiC-it 2023 we are back to a fully in-presence conference. Overall, almost 210 participants registered to the conference, confirming that the community is eager to meet in person and to enjoy both the scientific and social events together with the colleagues.

The opportunities afforded by the Internet greatly overshadow the challenges. While not forgetting this, we can nevertheless still address some of the problems that arise. Hate speech online is one such problem. But what exactly is hate speech online, and how can we deal with it effectively? As with freedom of expression, on- or offline, UNESCO defends the position that the free flow of information should always be the norm. Counter-speech is generally preferable to suppression of speech. And any response that limits speech needs to be very carefully weighed to ensure that this remains wholly exceptional, and that legitimate robust debate is not curtailed.

This book constitutes the refereed proceedings of the 40th European Conference on IR Research, ECIR 2018, held in Grenoble, France, in March 2018. The 39 full papers and 39 short papers presented together with 6 demos, 5 workshops and 3 tutorials, were carefully reviewed and selected from 303 submissions. Accepted papers cover the state of the art in information retrieval including topics such as: topic modeling, deep learning, evaluation, user behavior, document representation, recommendation systems, retrieval methods, learning and classication, and micro-blogs.

This book contains original, peer-reviewed research articles from the Second International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications, held in March 28-29th 2021 at CMR Institute of Technology, Hyderabad, Telangana India. It covers the latest research trends and developments in areas of machine learning, artificial intelligence, neural networks, cyber-physical systems, cybernetics, with emphasis on applications in smart cities, Internet of Things, practical data science and cognition. The book focuses on the comprehensive tenets of artificial intelligence, machine learning and deep learning to emphasize its use in modelling, identification, optimization, prediction, forecasting and control of future intelligent systems. Submissions were solicited of unpublished material, and present in-depth fundamental research contributions from a methodological/application perspective in understanding artificial intelligence and machine learning approaches and their capabilities in solving a diverse range of problems in industries and its real-world applications.

Identifying and stopping the dissemination of fabricated news, hate speech, or deceptive information camouflaged as legitimate news poses a significant technological hurdle. This book presents emergent methodologies and technological approaches of natural language processing through machine learning for counteracting the spread of fake news and hate speech on social media platforms. • Covers various approaches, algorithms, and methodologies for fake news and hate speech detection. • Explains the automatic detection and prevention of fake news and hate speech through paralinguistic clues on social media using artificial intelligence. • Discusses the application of machine learning models to learn linguistic characteristics of hate speech over social media platforms. • Emphasizes the role of multilingual and multimodal processing to detect fake news. • Includes research on different optimization techniques, case studies on the identification, prevention, and social impact of fake news, and GitHub repository links to aid understanding. The text is for professionals and scholars of various disciplines interested in fake news and hate speech detection.

This book constitutes the proceedings of the 25th International Conference on Text, Speech, and Dialogue, TSD 2022, held in Brno, Czech Republic, in September 2022. The 43 papers presented in this volume were carefully reviewed and selected from 94 submissions. The topical sections "Text", "Speech", and "Dialogue" deal with the following issues: speech recognition; corpora and language resources; speech and spoken language generation; tagging, classification and parsing of text and speech; semantic processing of text and speech; integrating applications of text and speech processing; automatic dialogue systems; multimodal techniques and modelling.