Year: 2019
Type: preprint
Abstract: We address the issue of hallucination in data-to-text generation, i.e., reducing the generation of text that is unsupported by the source. We conjecture that hallucination can be caused by an encoder-decoder model generating content phrases without attending to the source; so we propose a confidence score to ensure that the model attends to the source whenever necessary, as well as a variational Bayes training framework that can learn the score from data. Experiments on the WikiBio (Lebretet al., 2016) dataset show that our approach is more faithful to the source than existing state-of-the-art approaches, according to both PARENT score (Dhingra et al., 2019) and human evaluation. We also report strong results on the WebNLG (Gardent et al., 2017) dataset.
Source: arXiv (Cornell University)
Institution: Google (United States)
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2981259322