Few-Shot KBQA Method Based on Multi-Task Learning

Question-answering systems have become a prominent topic in the field of artificial intelligence. A crucial aspect is knowledge-based question answering (KBQA), used in search engines and intelligent customer service to enhance user experiences. However, existing methods often struggle to model complex relationships and operations in few-shot learning environments. To solve this problem, a multi-task KBQA method has been proposed. This method includes various auxiliary tasks such as relational sequence prediction, knowledge completion prediction, and query program reconstruction. A multi-task fusion training approach was adopted for model generation. Experimental results show that accuracy can be significantly improved by more than 6% in few-shot learning environments, achieving better performance with an accuracy rate of 92.45%.