Transformer-driven chatbot for Arabic agricultural knowledge

Abstract Nowadays, chatbots have become highly valuable tools for simplifying tasks across various fields. However, there is a noticeable lack of Arabic chatbots specifically tailored for the agricultural sector. This study aims to address this gap by exploring two transformer-based models, AraBERT v2 and QA AraBERT. These models are based on BERT (Bidirectional Encoder Representations from Transformers), a widely recognized and powerful transformer-based language model in the field of natural language processing (NLP). By leveraging the power of BERT, the chatbot will gain a deep understanding of the Arabic language and the contextual relationships between words. The BERT models underwent fine-tuning on a specialized Arabic dataset tailored to the agricultural sector, particularly in the realm of ornamental trees and plants. Focusing on the second model, QA AraBERT, the fine-tuning process yielded F1 score of 81.63%. This remarkable performance highlights the model’s exceptional capability to grasp the nuances of agricultural inquiries. Empowered by this model, the chatbot achieves high performance compared to other chatbots built with similar techniques.