Arabic chatbot technologies: A scoping review

Background: Chatbots have been widely used in many spheres of life from customer services to mental health companions. Despite the breakthroughs in achieving human-like conversations, Arabic language chatbots driven by AI and NLP are relatively scarce due to the complex nature of the Arabic language. Objective: We aim to review published literature on Arabic chatbots to gain insight into the technologies used highlighting the gap in this emerging field. Methods: To identify relevant studies, we searched eight bibliographic databases and conducted backward and forward reference checking. Two reviewers independently performed study selection and data extraction. The extracted data was synthesized using a narrative approach. Results: We included 18 of 1755 retrieved publications. Thirteen unique chatbots were identified from the 18 studies. ArabChat was the most common chatbot in the included studies (n = 5). The type of Arabic language in most chatbots (n = 13) was Modern Standard Arabic. The input and output modalities used in 17 chatbots were only text. Most chatbots (n = 14) were able to have long conversations. The majority of the chatbots (n = 14) were developed to serve a specific purpose (Closed domain). A retrieval-based model was used for developing most chatbots (n = 17). Conclusion: Despite a large number of chatbots worldwide, there is relatively a small number of Arabic language chatbots. Furthermore, the available Arabic language chatbots are less advanced than other language chatbots. Researchers should develop more Arabic language chatbots that are based on more advanced input and output modalities, generative-based models, and natural language processing (NLP).