{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,3]],"date-time":"2025-11-03T09:01:13Z","timestamp":1762160473788,"version":"build-2065373602"},"reference-count":38,"publisher":"PeerJ","license":[{"start":{"date-parts":[[2025,11,3]],"date-time":"2025-11-03T00:00:00Z","timestamp":1762128000000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Deanship of Graduate Studies and Scientific Research at University of Bisha through the Fast-Track Research Support Program"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>The detection of hate speech on social media has become a pressing challenge, particularly in multilingual and low-resource language settings such as Roman Urdu, where informal grammar, code-switching, and inconsistent orthography hinder accurate classification. Despite progress in hate speech detection for high-resource languages, limited research exists for Roman Urdu content. This study addresses this gap by proposing a computationally efficient deep learning framework based on a hybrid convolutional neural network and bidirectional long short-term memory (CNN-BiLSTM) architecture. The model leverages FastText pre-trained embeddings to capture subword-level semantics and combines convolutional layers for local feature extraction with BiLSTM for global context modeling. We evaluate our approach on a labeled Roman Urdu dataset and compare it with traditional machine learning models and deep learning baselines. Our proposed CNN-BiLSTM model achieves the highest performance with an accuracy of 80.67% and an F1-score of 81.47%, outperforming competitive baselines. These findings demonstrate the effectiveness and practicality of our lightweight architecture in detecting hate speech in Roman Urdu, offering a scalable solution for multilingual and resource-constrained environments.<\/jats:p>","DOI":"10.7717\/peerj-cs.3342","type":"journal-article","created":{"date-parts":[[2025,11,3]],"date-time":"2025-11-03T08:47:31Z","timestamp":1762159651000},"page":"e3342","source":"Crossref","is-referenced-by-count":0,"title":["Detecting hate speech in roman Urdu using a convolutional-BiLSTM-based deep hybrid neural network"],"prefix":"10.7717","volume":"11","author":[{"given":"Muhammad","family":"Zohaib","sequence":"first","affiliation":[{"name":"Information Systems, University of Management & Technology, Lahore, Lahore, Pakistan"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ghulam","family":"Farooque","sequence":"additional","affiliation":[{"name":"Computer Science and Information Technology, University of Lahore, Lahore, Pakistan"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5765-1291","authenticated-orcid":true,"given":"Mohammad","family":"Alsulami","sequence":"additional","affiliation":[{"name":"Computer Science, Najran University, Najran, Saudi Arabia"},{"name":"Emerging Technologies Research Lab (ETRL), Najran University, Najran, Saudi Arabia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Fazeel","family":"Abid","sequence":"additional","affiliation":[{"name":"Computer Science and Information Technology, University of Lahore, Lahore, Pakistan"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7721-0450","authenticated-orcid":true,"given":"Ali","family":"Alqazzaz","sequence":"additional","affiliation":[{"name":"College of Computing and Information Technology, University of Bisha, Bisha, Saudi Arabia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2266-9608","authenticated-orcid":true,"given":"Mana Saleh Al","family":"Reshan","sequence":"additional","affiliation":[{"name":"Emerging Technologies Research Lab (ETRL), Najran University, Najran, Saudi Arabia"},{"name":"Information Systems, Najran University, Najran, Saudi Arabia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3761-1641","authenticated-orcid":true,"given":"Jawad","family":"Rasheed","sequence":"additional","affiliation":[{"name":"Software Engineering, Istanbul Nisantasi University, Istanbul, Turkey"},{"name":"Applied Science Research Center, Applied Science Private University, Amman, Jordan"},{"name":"Research Institute, Istanbul Medipol University, Istanbul, Turkey"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4806-6159","authenticated-orcid":true,"given":"Asadullah","family":"Shaikh","sequence":"additional","affiliation":[{"name":"Emerging Technologies Research Lab (ETRL), Najran University, Najran, Saudi Arabia"},{"name":"Information Systems, Najran University, Najran, Saudi Arabia"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"4443","published-online":{"date-parts":[[2025,11,3]]},"reference":[{"issue":"6","key":"10.7717\/peerj-cs.3342\/ref-1","doi-asserted-by":"publisher","first-page":"1963","DOI":"10.1007\/s00530-020-00742-w","article-title":"Detection of hate speech in Arabic tweets using deep learning","volume":"28","author":"Al-Hassan","year":"2022","journal-title":"Multimedia Systems"},{"issue":"1","key":"10.7717\/peerj-cs.3342\/ref-2","doi-asserted-by":"publisher","first-page":"19","DOI":"10.1080\/13600834.2018.1494417","article-title":"Hate speech on social media networks: towards a regulatory framework?","volume":"28","author":"Alkiviadou","year":"2019","journal-title":"Information & Communications Technology Law"},{"key":"10.7717\/peerj-cs.3342\/ref-3","doi-asserted-by":"publisher","first-page":"28590","DOI":"10.1038\/s41598-024-79106-7","article-title":"Roman Urdu hate speech detection using hybrid machine learning models and hyperparameter optimization","volume":"14","author":"Ashiq","year":"2024","journal-title":"Scientific Reports"},{"issue":"3","key":"10.7717\/peerj-cs.3342\/ref-4","doi-asserted-by":"publisher","first-page":"e896","DOI":"10.7717\/peerj-cs.896","article-title":"Multi-label emotion classification of Urdu tweets","volume":"8","author":"Ashraf","year":"2022","journal-title":"PeerJ Computer Science"},{"key":"10.7717\/peerj-cs.3342\/ref-5","first-page":"759","article-title":"Deep learning for hate speech detection in tweets","author":"Badjatiya","year":"2017"},{"key":"10.7717\/peerj-cs.3342\/ref-6","first-page":"2470","article-title":"Hate or non-hate: translation based hate speech identification in code-mixed Hinglish data set","author":"Biradar","year":"2021"},{"issue":"1","key":"10.7717\/peerj-cs.3342\/ref-7","doi-asserted-by":"publisher","first-page":"512","DOI":"10.1609\/icwsm.v11i1.14955","article-title":"Automated hate speech detection and the problem of offensive language","volume":"11","author":"Davidson","year":"2017","journal-title":"Proceedings of the International AAAI Conference on Web and Social Media"},{"key":"10.7717\/peerj-cs.3342\/ref-8","first-page":"29","article-title":"Hate speech detection with comment embeddings","author":"Djuric","year":"2015"},{"issue":"4","key":"10.7717\/peerj-cs.3342\/ref-9","doi-asserted-by":"publisher","first-page":"4001","DOI":"10.1007\/s13369-021-05383-3","article-title":"A deep learning framework for automatic detection of hate speech embedded in Arabic tweets","volume":"46","author":"Duwairi","year":"2021","journal-title":"Arabian Journal for Science and Engineering"},{"key":"10.7717\/peerj-cs.3342\/ref-10","first-page":"119","article-title":"At the lower end of language exploring the vulgar and obscene side of German","author":"Eder","year":"2019"},{"key":"10.7717\/peerj-cs.3342\/ref-11","first-page":"7","article-title":"When does a compliment become sexist? analysis and classification of ambivalent sexism using twitter data","author":"Jha","year":"2017"},{"issue":"5","key":"10.7717\/peerj-cs.3342\/ref-12","doi-asserted-by":"publisher","first-page":"2694","DOI":"10.3390\/app12052694","article-title":"Deep sentiment analysis using CNN-LSTM architecture of English and Roman Urdu text shared in social media","volume":"12","author":"Khan","year":"2022a","journal-title":"Applied Sciences"},{"issue":"1","key":"10.7717\/peerj-cs.3342\/ref-13","doi-asserted-by":"publisher","first-page":"5436","DOI":"10.1038\/s41598-022-09381-9","article-title":"Multi-class sentiment analysis of Urdu text using multilingual BERT","volume":"12","author":"Khan","year":"2022b","journal-title":"Scientific Reports"},{"key":"10.7717\/peerj-cs.3342\/ref-14","doi-asserted-by":"publisher","first-page":"97803","DOI":"10.1109\/access.2021.3093078","article-title":"Urdu sentiment analysis with deep learning methods","volume":"9","author":"Khan","year":"2021","journal-title":"IEEE Access"},{"key":"10.7717\/peerj-cs.3342\/ref-15","doi-asserted-by":"publisher","first-page":"10","DOI":"10.1007\/s40747-024-01631-9","article-title":"Empowering Urdu sentiment analysis: an attention-based stacked CNN-Bi-LSTM DNN with multilingual BERT","volume":"11","author":"Khan","year":"2025","journal-title":"Complex & Intelligent Systems"},{"issue":"1","key":"10.7717\/peerj-cs.3342\/ref-16","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3414524","article-title":"Hate speech detection in roman Urdu","volume":"20","author":"Khan","year":"2021","journal-title":"ACM Transactions on Asian and Low-Resource Language Information Processing"},{"issue":"2","key":"10.7717\/peerj-cs.3342\/ref-17","doi-asserted-by":"publisher","first-page":"102233","DOI":"10.1007\/s42979-021-00457-3","article-title":"Challenges of hate speech detection in social media","volume":"2","author":"Kov\u00e1cs","year":"2021","journal-title":"SN Computer Science"},{"issue":"4","key":"10.7717\/peerj-cs.3342\/ref-18","doi-asserted-by":"publisher","first-page":"102233","DOI":"10.1016\/j.ipm.2020.102233","article-title":"Deep sentiments in Roman Urdu text using recurrent convolutional neural network model","volume":"57","author":"Mahmood","year":"2020","journal-title":"Information Processing & Management"},{"key":"10.7717\/peerj-cs.3342\/ref-19","first-page":"467","article-title":"Detecting hate speech in social media","author":"Malmasi","year":"2017"},{"key":"10.7717\/peerj-cs.3342\/ref-20","first-page":"14","article-title":"Overview of the HASOC track at FIRE 2019: hate speech and offensive content identification in indo-European languages","author":"Mandl","year":"2019"},{"key":"10.7717\/peerj-cs.3342\/ref-21","article-title":"Why you should use cross-entropy error instead of classification error or mean squared error for neural network classifier training","author":"McCaffrey","year":"2015"},{"key":"10.7717\/peerj-cs.3342\/ref-22","first-page":"52","article-title":"Abusive language detection on Arabic social media","author":"Mubarak","year":"2017"},{"key":"10.7717\/peerj-cs.3342\/ref-23","doi-asserted-by":"publisher","first-page":"88364","DOI":"10.1109\/access.2021.3089515","article-title":"Advances in machine learning algorithms for hate speech detection in social media: a review","volume":"9","author":"Mullah","year":"2021","journal-title":"IEEE Access"},{"key":"10.7717\/peerj-cs.3342\/ref-24","first-page":"145","article-title":"Abusive language detection in online user content","author":"Nobata","year":"2016"},{"key":"10.7717\/peerj-cs.3342\/ref-25","first-page":"13","article-title":"Code-switching in Urdu books of Punjab text book board, Lahore, Pakistan","volume":"1","author":"Noor","year":"2015","journal-title":"Communication and Linguistics Studies"},{"issue":"1","key":"10.7717\/peerj-cs.3342\/ref-26","doi-asserted-by":"publisher","first-page":"115","DOI":"10.3390\/s16010115","article-title":"Deep convolutional and LSTM recurrent neural networks for multimodal wearable activity recognition","volume":"16","author":"Ord\u00f3\u00f1ez","year":"2016","journal-title":"Sensors"},{"issue":"12","key":"10.7717\/peerj-cs.3342\/ref-27","doi-asserted-by":"publisher","first-page":"4730","DOI":"10.1007\/s10489-018-1242-y","article-title":"Effective hate-speech detection in twitter data using recurrent neural networks","volume":"48","author":"Pitsilis","year":"2018","journal-title":"Applied Intelligence"},{"issue":"3","key":"10.7717\/peerj-cs.3342\/ref-28","doi-asserted-by":"publisher","first-page":"032006","DOI":"10.1088\/1757-899x\/830\/3\/032006","article-title":"A comparison of classification algorithms for hate speech detection","volume":"830","author":"Putri","year":"2020","journal-title":"IOP Conference Series: Materials Science and Engineering"},{"key":"10.7717\/peerj-cs.3342\/ref-29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.05324","article-title":"Multilingual offensive language identification with cross-lingual embeddings","author":"Ranasinghe","year":"2020"},{"key":"10.7717\/peerj-cs.3342\/ref-30","volume-title":"New insights in machine learning and deep neural networks","author":"Renna","year":"2023"},{"key":"10.7717\/peerj-cs.3342\/ref-31","first-page":"2512","article-title":"Hate-speech and offensive language detection in roman urdu","author":"Rizwan","year":"2020"},{"key":"10.7717\/peerj-cs.3342\/ref-32","first-page":"903","article-title":"Adapting deep learning for sentiment classification of code-switched informal short text","author":"Shakeel","year":"2020"},{"issue":"7","key":"10.7717\/peerj-cs.3342\/ref-33","doi-asserted-by":"publisher","first-page":"76","DOI":"10.1016\/j.neucom.2016.12.088","article-title":"An analysis of convolutional long short-term memory recurrent neural networks for gesture recognition","volume":"268","author":"Tsironi","year":"2017","journal-title":"Neurocomputing"},{"key":"10.7717\/peerj-cs.3342\/ref-34","first-page":"86","article-title":"Hate me, hate me not: hate speech detection on facebook","author":"Vigna","year":"2017"},{"key":"10.7717\/peerj-cs.3342\/ref-35","first-page":"415","article-title":"Cursing in English on Twitter","author":"Wang","year":"2014"},{"key":"10.7717\/peerj-cs.3342\/ref-36","first-page":"138","article-title":"Are you a racist or am I seeing things? annotator influence on hate speech detection on Twitter","author":"Waseem","year":"2016"},{"key":"10.7717\/peerj-cs.3342\/ref-37","first-page":"88","article-title":"Hateful symbols or hateful people? predictive features for hate speech detection on Twitter","author":"Waseem","year":"2016"},{"issue":"1\u20132","key":"10.7717\/peerj-cs.3342\/ref-38","doi-asserted-by":"publisher","first-page":"298","DOI":"10.31577\/cai_2020_1-2_298","article-title":"Travel mode recognition from GPS data based on LSTM","volume":"39","author":"Zhu","year":"2020","journal-title":"Computing and Informatics"}],"container-title":["PeerJ Computer Science"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/peerj.com\/articles\/cs-3342.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3342.xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3342.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/peerj.com\/articles\/cs-3342.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,3]],"date-time":"2025-11-03T08:47:33Z","timestamp":1762159653000},"score":1,"resource":{"primary":{"URL":"https:\/\/peerj.com\/articles\/cs-3342"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,3]]},"references-count":38,"alternative-id":["10.7717\/peerj-cs.3342"],"URL":"https:\/\/doi.org\/10.7717\/peerj-cs.3342","archive":["CLOCKSS","LOCKSS","Portico"],"relation":{},"ISSN":["2376-5992"],"issn-type":[{"value":"2376-5992","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,3]]},"article-number":"e3342"}}