2024.11.26
Overview
The present invention discloses a hallucination detection system, a hallucination detection method, and an electronic device. The electronic device comprises the hallucination detection system. The hallucination detection system includes a storage unit and a processor. The processor is configured to execute the hallucination detection method. The hallucination detection method comprises a conversion step, a veracity detection step, and a result output step. In the conversion step, multiple output word net are generated based on an output text. In the veracity detection step, multiple veracity information are generated using a pre-trained graph embedding model. Each veracity information represents whether one of the output word net is determined to be true or false by the pre-trained graph embedding model. In the result output step, hallucination detection result information is generated based on the multiple veracity information. The graph embedding model is trained using multiple positive word net and multiple negative word net from a single knowledge domain.Category
發明
Applying
18/959,776
Filing Date
2024.11.26Notification
2026.11.13