A Simple Recipe Towards Reducing Hallucination In Neural Surface Realisation
57TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2019)(2019)
摘要
Recent neural language generation systems often hallucinate contents (i.e., producing irrelevant or contradicted facts), especially when trained on loosely corresponding pairs of the input structure and text. To mitigate this issue, we propose to integrate a language understanding module for data refinement with self-training iterations to effectively induce strong equivalence between the input data and the paired text. Experiments on the E2E challenge dataset show that our proposed framework can reduce more than 50% relative unaligned noise from the original data-text pairs. A vanilla sequence-to-sequence neural NLG model trained on the refined data has improved on content correctness compared with the current state-of-the-art ensemble generator.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络