Chatbot Intentions and Training Phrases - Arabic

---

During my work with the Jordanian Ministry in partnership with Infobip, I designed and built a multilingual Arabic chatbot system that automated major government-service interactions. I developed the full intent taxonomy, generated training phrases using LLM-assisted augmentation, and engineered a transformer-based classifier to understand both MSA and dialectal Arabic.

For this project I:

• Created a structured intent model and conversational flow
• Generated and cleaned Arabic training data using Python + LLMs
• Applied text normalization to handle morphology and diacritic issues
• Trained and evaluated transformer-based intent classifiers
• Integrated the bot into Infobip’s messaging platform
• Wrote documentation and collaborated closely with ministry and Infobip technical teams
   

This project combined my skills in LLMs, NLP, multilingual model training, data engineering, and cross-team collaboration while delivering a production-ready automation system for a real government organization.