Threatening URDU Language Detection from Tweets Using Machine Learning

Article Subjects > Engineering Europe University of Atlantic > Research > Scientific Production
Fundación Universitaria Internacional de Colombia > Research > Scientific Production
Ibero-american International University > Research > Scientific Production
Ibero-american International University > Research > Scientific Production
Universidad Internacional do Cuanza > Research > Articles and books
Abierto Inglés Technology’s expansion has contributed to the rise in popularity of social media platforms. Twitter is one of the leading social media platforms that people use to share their opinions. Such opinions, sometimes, may contain threatening text, deliberately or non-deliberately, which can be disturbing for other users. Consequently, the detection of threatening content on social media is an important task. Contrary to high-resource languages like English, Dutch, and others that have several such approaches, the low-resource Urdu language does not have such a luxury. Therefore, this study presents an intelligent threatening language detection for the Urdu language. A stacking model is proposed that uses an extra tree (ET) classifier and Bayes theorem-based Bernoulli Naive Bayes (BNB) as the based learners while logistic regression (LR) is employed as the meta learner. A performance analysis is carried out by deploying a support vector classifier, ET, LR, BNB, fully connected network, convolutional neural network, long short-term memory, and gated recurrent unit. Experimental results indicate that the stacked model performs better than both machine learning and deep learning models. With 74.01% accuracy, 70.84% precision, 75.65% recall, and 73.99% F1 score, the model outperforms the existing benchmark study. metadata Mehmood, Aneela and Farooq, Muhammad Shoaib and Naseem, Ansar and Rustam, Furqan and Gracia Villar, Mónica and Rodríguez Velasco, Carmen Lilí and Ashraf, Imran mail UNSPECIFIED, UNSPECIFIED, UNSPECIFIED, UNSPECIFIED, monica.gracia@uneatlantico.es, carmen.rodriguez@uneatlantico.es, UNSPECIFIED (2022) Threatening URDU Language Detection from Tweets Using Machine Learning. Applied Sciences, 12 (20). p. 10342. ISSN 2076-3417

[img] Text
applsci-12-10342-v3.pdf
Available under License Creative Commons Attribution.

Download (884kB)

Abstract

Technology’s expansion has contributed to the rise in popularity of social media platforms. Twitter is one of the leading social media platforms that people use to share their opinions. Such opinions, sometimes, may contain threatening text, deliberately or non-deliberately, which can be disturbing for other users. Consequently, the detection of threatening content on social media is an important task. Contrary to high-resource languages like English, Dutch, and others that have several such approaches, the low-resource Urdu language does not have such a luxury. Therefore, this study presents an intelligent threatening language detection for the Urdu language. A stacking model is proposed that uses an extra tree (ET) classifier and Bayes theorem-based Bernoulli Naive Bayes (BNB) as the based learners while logistic regression (LR) is employed as the meta learner. A performance analysis is carried out by deploying a support vector classifier, ET, LR, BNB, fully connected network, convolutional neural network, long short-term memory, and gated recurrent unit. Experimental results indicate that the stacked model performs better than both machine learning and deep learning models. With 74.01% accuracy, 70.84% precision, 75.65% recall, and 73.99% F1 score, the model outperforms the existing benchmark study.

Item Type: Article
Uncontrolled Keywords: threatening language detection; Urdu text classification; machine learning; stacking
Subjects: Subjects > Engineering
Divisions: Europe University of Atlantic > Research > Scientific Production
Fundación Universitaria Internacional de Colombia > Research > Scientific Production
Ibero-american International University > Research > Scientific Production
Ibero-american International University > Research > Scientific Production
Universidad Internacional do Cuanza > Research > Articles and books
Date Deposited: 26 Oct 2022 23:30
Last Modified: 18 Jul 2023 23:30
URI: https://repositorio.unic.co.ao/id/eprint/4194

Actions (login required)

View Item View Item

<a class="ep_document_link" href="/10290/1/Influence%20of%20E-learning%20training%20on%20the%20acquisition%20of%20competences%20in%20basketball%20coaches%20in%20Cantabria.pdf"><img class="ep_doc_icon" alt="[img]" src="/style/images/fileicons/text.png" border="0"/></a>

en

open

Influence of E-learning training on the acquisition of competences in basketball coaches in Cantabria

The main aim of this study was to analyse the influence of e-learning training on the acquisition of competences in basketball coaches in Cantabria. The current landscape of basketball coach training shows an increasing demand for innovative training models and emerging pedagogies, including e-learning-based methodologies. The study sample consisted of fifty students from these courses, all above 16 years of age (36 males, 14 females). Among them, 16% resided outside the autonomous community of Cantabria, 10% resided more than 50 km from the city of Santander, 36% between 10 and 50 km, 14% less than 10 km, and 24% resided within Santander city. Data were collected through a Google Forms survey distributed by the Cantabrian Basketball Federation to training course students. Participation was voluntary and anonymous. The survey, consisting of 56 questions, was validated by two sports and health doctors and two senior basketball coaches. The collected data were processed and analysed using Microsoft® Excel version 16.74, and the results were expressed in percentages. The analysis revealed that 24.60% of the students trained through the e-learning methodology considered themselves fully qualified as basketball coaches, contrasting with 10.98% of those trained via traditional face-to-face methodology. The results of the study provide insights into important characteristics that can be adjusted and improved within the investigated educational process. Moreover, the study concludes that e-learning training effectively qualifies basketball coaches in Cantabria.

Producción Científica

Josep Alemany Iturriaga mail josep.alemany@uneatlantico.es, Álvaro Velarde-Sotres mail alvaro.velarde@uneatlantico.es, Javier Jorge mail , Kamil Giglio mail ,

Alemany Iturriaga

<a href="/15198/1/nutrients-16-03859.pdf" class="ep_document_link"><img class="ep_doc_icon" alt="[img]" src="/style/images/fileicons/text.png" border="0"/></a>

en

open

Carotenoids Intake and Cardiovascular Prevention: A Systematic Review

Background: Cardiovascular diseases (CVDs) encompass a variety of conditions that affect the heart and blood vessels. Carotenoids, a group of fat-soluble organic pigments synthesized by plants, fungi, algae, and some bacteria, may have a beneficial effect in reducing cardiovascular disease (CVD) risk. This study aims to examine and synthesize current research on the relationship between carotenoids and CVDs. Methods: A systematic review was conducted using MEDLINE and the Cochrane Library to identify relevant studies on the efficacy of carotenoid supplementation for CVD prevention. Interventional analytical studies (randomized and non-randomized clinical trials) published in English from January 2011 to February 2024 were included. Results: A total of 38 studies were included in the qualitative analysis. Of these, 17 epidemiological studies assessed the relationship between carotenoids and CVDs, 9 examined the effect of carotenoid supplementation, and 12 evaluated dietary interventions. Conclusions: Elevated serum carotenoid levels are associated with reduced CVD risk factors and inflammatory markers. Increasing the consumption of carotenoid-rich foods appears to be more effective than supplementation, though the specific effects of individual carotenoids on CVD risk remain uncertain.

Producción Científica

Sandra Sumalla Cano mail sandra.sumalla@uneatlantico.es, Imanol Eguren García mail imanol.eguren@uneatlantico.es, Álvaro Lasarte García mail , Thomas Prola mail thomas.prola@uneatlantico.es, Raquel Martínez Díaz mail raquel.martinez@uneatlantico.es, Iñaki Elío Pascual mail inaki.elio@uneatlantico.es,

Sumalla Cano

<a class="ep_document_link" href="/15444/1/s41598-024-79106-7.pdf"><img class="ep_doc_icon" alt="[img]" src="/style/images/fileicons/text.png" border="0"/></a>

en

open

Roman urdu hate speech detection using hybrid machine learning models and hyperparameter optimization

With the rapid increase of users over social media, cyberbullying, and hate speech problems have arisen over the past years. Automatic hate speech detection (HSD) from text is an emerging research problem in natural language processing (NLP). Researchers developed various approaches to solve the automatic hate speech detection problem using different corpora in various languages, however, research on the Urdu language is rather scarce. This study aims to address the HSD task on Twitter using Roman Urdu text. The contribution of this research is the development of a hybrid model for Roman Urdu HSD, which has not been previously explored. The novel hybrid model integrates deep learning (DL) and transformer models for automatic feature extraction, combined with machine learning algorithms (MLAs) for classification. To further enhance model performance, we employ several hyperparameter optimization (HPO) techniques, including Grid Search (GS), Randomized Search (RS), and Bayesian Optimization with Gaussian Processes (BOGP). Evaluation is carried out on two publicly available benchmarks Roman Urdu corpora comprising HS-RU-20 corpus and RUHSOLD hate speech corpus. Results demonstrate that the Multilingual BERT (MBERT) feature learner, paired with a Support Vector Machine (SVM) classifier and optimized using RS, achieves state-of-the-art performance. On the HS-RU-20 corpus, this model attained an accuracy of 0.93 and an F1 score of 0.95 for the Neutral-Hostile classification task, and an accuracy of 0.89 with an F1 score of 0.88 for the Hate Speech-Offensive task. On the RUHSOLD corpus, the same model achieved an accuracy of 0.95 and an F1 score of 0.94 for the Coarse-grained task, alongside an accuracy of 0.87 and an F1 score of 0.84 for the Fine-grained task. These results demonstrate the effectiveness of our hybrid approach for Roman Urdu hate speech detection.

Producción Científica

Waqar Ashiq mail , Samra Kanwal mail , Adnan Rafique mail , Muhammad Waqas mail , Tahir Khurshaid mail , Elizabeth Caro Montero mail elizabeth.caro@uneatlantico.es, Alicia Bustamante Alonso mail alicia.bustamante@uneatlantico.es, Imran Ashraf mail ,

Ashiq

<a class="ep_document_link" href="/14584/1/s41598-024-73664-6.pdf"><img class="ep_doc_icon" alt="[img]" src="/style/images/fileicons/text.png" border="0"/></a>

en

open

Performance of the 4C and SEIMC scoring systems in predicting mortality from onset to current COVID-19 pandemic in emergency departments

The evolution of the COVID-19 pandemic has been associated with variations in clinical presentation and severity. Similarly, prediction scores may suffer changes in their diagnostic accuracy. The aim of this study was to test the 30-day mortality predictive validity of the 4C and SEIMC scores during the sixth wave of the pandemic and to compare them with those of validation studies. This was a longitudinal retrospective observational study. COVID-19 patients who were admitted to the Emergency Department of a Spanish hospital from December 15, 2021, to January 31, 2022, were selected. A side-by-side comparison with the pivotal validation studies was subsequently performed. The main measures were 30-day mortality and the 4C and SEIMC scores. A total of 27,614 patients were considered in the study, including 22,361 from the 4C, 4,627 from the SEIMC and 626 from our hospital. The 30-day mortality rate was significantly lower than that reported in the validation studies. The AUCs were 0.931 (95% CI: 0.90–0.95) for 4C and 0.903 (95% CI: 086–0.93) for SEIMC, which were significantly greater than those obtained in the first wave. Despite the changes that have occurred during the coronavirus disease 2019 (COVID-19) pandemic, with a reduction in lethality, scorecard systems are currently still useful tools for detecting patients with poor disease risk, with better prognostic capacity.

Producción Científica

Pedro Ángel de Santos Castro mail , Carlos del Pozo Vegas mail , Leyre Teresa Pinilla Arribas mail , Daniel Zalama Sánchez mail , Ancor Sanz-García mail , Tony Giancarlo Vásquez del Águila mail , Pablo González Izquierdo mail , Sara de Santos Sánchez mail , Cristina Mazas Pérez-Oleaga mail cristina.mazas@uneatlantico.es, Irma Dominguez Azpíroz mail irma.dominguez@unini.edu.mx, Iñaki Elío Pascual mail inaki.elio@uneatlantico.es, Francisco Martín-Rodríguez mail ,

de Santos Castro

<a href="/14915/1/s41598-024-74357-w.pdf" class="ep_document_link"><img class="ep_doc_icon" alt="[img]" src="/style/images/fileicons/text.png" border="0"/></a>

en

open

Enhanced detection of diabetes mellitus using novel ensemble feature engineering approach and machine learning model

Diabetes is a persistent health condition led by insufficient use or inappropriate use of insulin in the body. If left undetected, it can lead to further complications involving organ damage such as heart, lungs, and eyes. Timely detection of diabetes helps obtain the right medication, diet, and exercise plan to lead a healthy life. ML approach has been utilized to obtain rapid and reliable diabetes detection, however, existing approaches suffer from the use of limited datasets, lack of generalizability, and lower accuracy. This study proposes a novel feature extraction approach to overcome these limitations by using an ensemble of convolutional neural network (CNN) and long short-term memory (LSTM) models. Multiple datasets are combined to make a larger dataset for experiments and multiple features are utilized for investigating the efficacy of the proposed approach. Features from the extra tree classifier, CNN, and LSTM are also considered for comparison. Experimental results reveal the superb performance of CNN-LSTM-based features with random forest model obtaining a 0.99 accuracy score. This performance is further validated by comparison with existing approaches and k-fold cross-validation which shows the proposed approach provides robust results.

Producción Científica

Furqan Rustam mail , Ahmad Sami Al-Shamayleh mail , Rahman Shafique mail , Silvia Aparicio Obregón mail silvia.aparicio@uneatlantico.es, Rubén Calderón Iglesias mail ruben.calderon@uneatlantico.es, J. Pablo Miramontes Gonzalez mail , Imran Ashraf mail ,

Rustam