Khairullah Khan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Khairullah Khan has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Fida Muhammad Khan
Department of Computer Science, Qurtuba University of Science & Information Technology, 25000 Peshawar, Pakistan
2 shared publications - Aftab Ahmad Khan
Department of Computer Science, University of Science and Technology Bannu, Bannu 28100, Pakistan
1 shared publication - Ahmad Ali
College of Mechatronics and Control Engineering, Shenzhen University, Shenzhen 518060, China
1 shared publication - Atta Ur Rahman
Center for Finance and Digital Economy, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
1 shared publication - Bilal Khan
Department of Computer Science and Information Technology, University of Science and Technology Bannu, Bannu, Pakistan
1 shared publication - Haseena Noureen
Department of Computer Science, University of Malakand, Pakistan
1 shared publication
Latest Publications
- Open Access
- 30 June 2025
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Advanced Computing and Systems
- Vol. 1 (2) , 106-116
Comparison of Machine Learning and Deep Learning Models for Part-of-Speech Tagging
Part-of-speech (POS) tagging—the automatic assignment of grammatical categories to every token in a text corpus—is a foundational preprocessing step for AI-driven language applications such as machine translation, sentiment analysis, and information retrieval. For morphologically complex, low-resource languages such as Pashto, the scarcity of annotated data and standardised tools makes this task particularly challenging. This paper presents a systematic comparative evaluation of six machine learning (ML) and deep learning (DL) algorithms—Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbor (KNN), Multi-Layer Perceptron (MLP), and Naïve Bayes (NB)—on a newly constructed 32,000-token CoNLL-formatted Pashto corpus (PashtoPoSTags). Each model is assessed by token-level accuracy, the standard metric for POS tagging evaluation. Decision Tree achieved the highest accuracy of 94.34%, followed by KNN with Jaccard distance (94.14%) and KNN with Euclidean distance (93.97%). Random Forest and SVM both exceeded the 90% threshold, while MLP with Tanh activation reached 87.25% and the best Naïve Bayes variant (Complement NB) attained 83.96%. Results show that classical ML algorithms with carefully engineered n-gram features provide computationally efficient and interpretable baselines for POS tagging in resource-constrainedsettings, and that PashtoPoSTags offers a reproducible benchmark for future Pashto NLP research.
- 30 June 2025
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Advanced Computing and Systems
- Vol. 1 (2) , 106-116
Open AccessComparison of Machine Learning and Deep Learning Models for Part-of-Speech Tagging
Part-of-speech (POS) tagging—the automatic assignment of grammatical categories to every token in a text corpus—is a foundational preprocessing step for AI-driven language applications such as machine translation, sentiment analysis, and information retrieval. For morphologically complex, low-resource languages such as Pashto, the scarcity of annotated data and standardised tools makes this task particularly challenging. This paper presents a systematic comparative evaluation of six machine learning (ML) and deep learning (DL) algorithms—Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbor (KNN), Multi-Layer Perceptron (MLP), and Naïve Bayes (NB)—on a newly constructed 32,000-token CoNLL-formatted Pashto corpus (PashtoPoSTags). Each model is assessed by token-level accuracy, the standard metric for POS tagging evaluation. Decision Tree achieved the highest accuracy of 94.34%, followed by KNN with Jaccard distance (94.14%) and KNN with Euclidean distance (93.97%). Random Forest and SVM both exceeded the 90% threshold, while MLP with Tanh activation reached 87.25% and the best Naïve Bayes variant (Complement NB) attained 83.96%. Results show that classical ML algorithms with carefully engineered n-gram features provide computationally efficient and interpretable baselines for POS tagging in resource-constrainedsettings, and that PashtoPoSTags offers a reproducible benchmark for future Pashto NLP research.
- Open Access
- 25 May 2025
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Advanced Computing and Systems
- Vol. 1 (2) , 78-96
Comparing Fine-Tuned RoBERTa with Traditional Machine Learning Models for Stance Detection in Political Tweets
Stance detection identifies a text’s position or attitude toward a given subject. A major challenge in Roman Urdu is the lack of a publicly available dataset for political stance detection. To address this gap, we constructed a high-quality dataset of 8,374 political tweets and comments using the Twitter API, annotated with stance labels: agree, disagree, and unrelated. The dataset captures diverse political viewpoints and user interactions. For feature representation, we employed TF-IDF due to its effectiveness in handling high-dimensional, context-sensitive Roman Urdu text. Several machine learning classifiers were evaluated, with Random Forest achieving the highest accuracy of 95%. Additionally, we fine-tuned the transformer-based RoBERTa model, which outperformed traditional methods with 97% accuracy. Our results demonstrate the potential of combining machine learning and deep learning for stance detection in low-resource languages. This study not only introduces a novel dataset but also provides a robust evaluation of methods, highlighting the importance of modern AI techniques in processing informal and multilingual text data.
- 25 May 2025
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Advanced Computing and Systems
- Vol. 1 (2) , 78-96
Open AccessComparing Fine-Tuned RoBERTa with Traditional Machine Learning Models for Stance Detection in Political Tweets
Stance detection identifies a text’s position or attitude toward a given subject. A major challenge in Roman Urdu is the lack of a publicly available dataset for political stance detection. To address this gap, we constructed a high-quality dataset of 8,374 political tweets and comments using the Twitter API, annotated with stance labels: agree, disagree, and unrelated. The dataset captures diverse political viewpoints and user interactions. For feature representation, we employed TF-IDF due to its effectiveness in handling high-dimensional, context-sensitive Roman Urdu text. Several machine learning classifiers were evaluated, with Random Forest achieving the highest accuracy of 95%. Additionally, we fine-tuned the transformer-based RoBERTa model, which outperformed traditional methods with 97% accuracy. Our results demonstrate the potential of combining machine learning and deep learning for stance detection in low-resource languages. This study not only introduces a novel dataset but also provides a robust evaluation of methods, highlighting the importance of modern AI techniques in processing informal and multilingual text data.