Multilingual Transliteration based Urdu Chatbot using Rasa Framework
Hira Mohiuddin, Bakhtiar Khan Kasi, Anayat Ullah · 2023
Chatbots, which emulate human conversation, are meeting the rising demand for efficient customer service by reducing the need for manual intervention and alleviating the challenges of timeconsuming human interactions. This paper introduces Ubot, a domain-specific chatbot fluent in both Urdu and English transliteration, designed to reduce the need for human intervention and lower service costs. We address the challenge of low-resource languages by creating a dataset comprising 750 utterances, 28 intents, and 45 entities, validated by 36 participants. Ubot, powered by the Rasa framework, achieves remarkable accuracy in intent detection 94.1% and entity extraction 95.4%, offering efficient multilingual conversational AI solutions. Our approach holds promise for similar script-sharing languages such as Persian and Arabic, contributing to advanced language applications and intelligent virtual assistants.