Abstract

E-mail is a widely used mode of communication in both professional and personal contexts. However, automating customer service responses in the Arabic language remains a challenging task. This study aims to develop a machine learning–based framework for classifying Arabic e-mails and generating appropriate automated responses. Two key innovations are introduced: (i) the release of the first well-described Arabic institutional e-mail dataset, and (ii) the design of a hybrid processing pipeline that integrates supervised classification with term frequency–inverse document frequency (TF-IDF) and cosine similarity–based retrieval. This dual approach directly addresses the underexplored domain of Arabic-language e-mail automation. A design science research (DSR) framework was employed to analyze e-mail records from 9,511 employees at King Faisal University. The e-mails were labeled as either inquiries or complaints using supervised learning methods, with particular emphasis on a support vector machine (SVM) classifier. TF-IDF was used for text feature extraction, while cosine similarity measured lexical similarity for response retrieval. Based on the predicted category, the system generated appropriate automated replies. The SVM trained on TF-IDF features achieved an accuracy of 90.4% on the held-out test set, with a precision of 93.5%, a recall of 94.1%, and an F1-score of 93.8%. During five-fold cross-validation, it obtained an average accuracy of 91.8% and an average F1-score of 91.4%. The SVM consistently outperformed logistic regression, naïve Bayes, and random forest classifiers. This study addresses an important research gap in the humanization of Arabic e-mail responses by proposing an effective supervised classification and response generation framework. The high accuracy, precision, and recall demonstrate that SVM is well suited for categorizing Arabic e-mails. Overall, the proposed system offers practical tools for Arabic language technology and contributes to the development of automated customer support solutions.

Keywords

Arabic e-mail classification, Machine learning, Support vector machine, TF-IDF, Cosine similarity, Automated customer support.

Cite this article

Almuqati A, Albarrak A, Mudasser Q, Wahsheh H. A data-driven approach to answering customer service e-mail inquiries in the Arabic language. International Journal of Advanced Technology and Engineering Exploration. 2026;13(134):123-143. DOI : 10.19101/IJATEE.2025.121220769

References

[1] Bouchareb N, Morad I. Analyzing the impact of ai-generated email marketing content on email deliverability in spam folder placement. HOLISTICA Journal of Business and Public Administration. 2024; 15(1):96-106.

[2] Kshetri N, Sharma RS. Development, diffusion, and impact of generative AI in the gulf cooperation council economies. Journal of Global Information Technology Management. 2025; 28(1):1-5.

[3] Al-khalifa S, Durrani N, Al-khalifa H, Alam F. The landscape of Arabic large language models. Communications of the ACM. 2025; 68(10):54-61.

[4] Darwish K, Habash N, Abbas M, Al-khalifa H, Al-natsheh HT, Bouamor H, et al. A panoramic survey of natural language processing in the Arab world. Communications of the ACM. 2021; 64(4):72-81.

[5] Abdelaziz ME, Deif MA, Algamdi SA, Elgohary R. A benchmark Arabic dataset for Arabic question classification using AAFAQ framework. Scientific Data. 2025; 12(1):1-11.

[6] Al-mutawa RF, Al-aama AY. Arabic opinion classification of customer service conversations using data augmentation and artificial intelligence. Big Data and Cognitive Computing. 2024; 8(12):1-21.

[7] Hossain MM, Hossain MS, Safran M, Alfarhood S, Alfarhood M, Mridha MF. A hybrid attention-based transformer model for Arabic news classification using text embedding and deep learning. IEEE Access. 2024; 12:198046-66.

[8] Kaddoura S, Alex SA, Itani M, Henno S, Alnashash A, Hemanth DJ. Arabic spam tweets classification using deep learning. Neural Computing and Applications. 2023; 35(23):17233-46.

[9] Alsuwaylimi AA. Enhancing arabic phishing email detection: a hybrid machine learning based on genetic algorithm feature selection. International Journal of Advanced Computer Science & Applications. 2024; 15(8):194-211.

[10] Wahdan A, Al-emran M, Shaalan K. A systematic review of Arabic text classification: areas, applications, and future directions. Soft Computing-A Fusion of Foundations, Methodologies & Applications. 2024; 28(2):1545-66.

[11] Alammary AS. BERT models for Arabic text classification: a systematic review. Applied Sciences. 2022; 12(11):1-20.

[12] Daraghmi EY, Qadan S, Daraghmi YA, Yousuf R, Cheikhrouhou O, Baz M. From text to insight: an integrated CNN-BILSTM-GRU model for Arabic cyberbullying detection. IEEE Access. 2024; 12:103504-19.

[13] Masri A, Al-jabi M. A novel approach for Arabic business email classification based on deep learning machines. Peer J Computer Science. 2023; 9:1-21.

[14] Hantom WH, Rahman A. Arabic spam tweets classification: a comprehensive machine learning approach. AI. 2024; 5(3):1049-65.

[15] Ghanim TM, Khalil MI, Abbas HM. Comparative study on deep convolution neural networks DCNN-based offline Arabic handwriting recognition. IEEE Access. 2020; 8:95465-82.

[16] Almanea M. Deep learning in written Arabic linguistic studies: a comprehensive survey. IEEE Access. 2024; 12:172196-233.

[17] Al-anzi FS, Abuzeina D. Toward an enhanced Arabic text classification using cosine similarity and latent semantic indexing. Journal of King Saud University-Computer and Information Sciences. 2017; 29(2):189-95.

[18] Jurn S, Kim W. Improving text classification of imbalanced call center conversations through data cleansing, augmentation, and NER Metadata. Electronics. 2025; 14(11):1-23.

[19] Kannan A, Kurach K, Ravi S, Kaufmann T, Tomkins A, Miklos B, et al. Smart reply: automated response suggestion for email. In proceedings of the 22nd ACM international conference on knowledge discovery and data mining 2016 (pp. 955-64). ACM.

[20] Li Y, Shi C, Duan Z, Liu F, Yang M. Fine-tuning BERT for intelligent software system fault classification. In 24th international conference on software quality, reliability, and security companion (QRS-C) 2024 (pp. 872-9). IEEE.

[21] Kazanci N. Extended topic classification utilizing LDA and BERTopic: a call center case study on robot agents and human agents. Applied Intelligence. 2025; 55(5):1-22.

[22] Alnagi E, Ghnemat R, Abu AQ. Boosting Arabic text classification using hybrid deep learning approach. Discover Applied Sciences. 2025; 7(6):1-20.

[23] Wang P, Fang J, Reinspach J. CS-BERT: a pretrained model for customer service dialogues. In proceedings of the 3rd workshop on natural language processing for conversational AI 2021 (pp. 130-42). Association for Computational Linguistics.

[24] Al-harbi NK, Alghieth M. Assessing BERT-based models for Arabic and low-resource languages in crime text classification. Peer J Computer Science. 2025; 11:1-42.

[25] Alabbas A, Alomar K. Tayseer: a novel ai-powered Arabic chatbot framework for technical and vocational student helpdesk services and enhancing student interactions. Applied Sciences. 2024; 14(6):1-28.

[26] Khaled S, Mohamed EH, Medhat W. Evaluating large language models for Arabic sentiment analysis: a comparative study using retrieval-augmented generation. Procedia Computer Science. 2024; 244:363-70.

[27] Alotaibi T, Al-dossari H. A review of fake news detection techniques for Arabic language. International Journal of Advanced Computer Science & Applications. 2024; 15(1):1-16.

[28] Saleh H, Almohimeed A, Hassan R, Ibrahim MM, Alsamhi SH, Hassan MR, et al. Advancing Arabic dialect detection with hybrid stacked transformer models. Frontiers in Human Neuroscience. 2025; 19:1-14.

[29] Mutawa AM, Sruthi S. A comparative evaluation of transformers and deep learning models for Arabic meter classification. Applied Sciences. 2025; 15(9):1-22.

[30] Lan F. Research on text similarity measurement hybrid algorithm with term semantic information and TF‐IDF method. Advances in Multimedia. 2022; 2022(1):1-11.

[31] Peffers K, Tuunanen T, Rothenberger MA, Chatterjee S. A design science research methodology for information systems research. Journal of Management Information Systems. 2007; 24(3):45-77.

[32] Hevner AR, March ST, Park J, Ram S. Design science in information systems research. MIS Quarterly. 2004; 28(1):75-105.

[33] Zhang Y. Support vector machine classification algorithm and its application. In international conference on information computing and applications 2012 (pp. 179-86). Berlin, Heidelberg: Springer Berlin Heidelberg.

[34] Ekbal A, Bandyopadhyay S. Named entity recognition in Bengali and Hindi using support vector machine. Lingvisticæ Investigationes. 2011; 34(1):35-67.

[35] Borg A, Boldt M. Using VADER sentiment and SVM for predicting customer response sentiment. Expert Systems with Applications. 2020; 162:113746.

[36] Vishwanathan SV, Murty MN. SSVM: a simple SVM algorithm. In proceedings of the international joint conference on neural networks (Cat. No. 02CH37290) 2002 (pp. 2393-8). IEEE.

[37] Vo NN, Liu S, Li X, Xu G. Leveraging unstructured call log data for customer churn prediction. Knowledge-Based Systems. 2021; 212:106586.