Semantic Noise Matters for Neural Natural Language Generation
This addresses the issue of unreliable text generation for users of NNLG systems, but it is incremental as it focuses on data cleaning rather than a new method.
The paper tackled the problem of neural natural language generation systems producing unrelated outputs by investigating the impact of semantic noise, finding that cleaned data improved semantic correctness by up to 97% while maintaining fluency, with omission being the most common error.
Neural natural language generation (NNLG) systems are known for their pathological outputs, i.e. generating text which is unrelated to the input specification. In this paper, we show the impact of semantic noise on state-of-the-art NNLG models which implement different semantic control mechanisms. We find that cleaned data can improve semantic correctness by up to 97%, while maintaining fluency. We also find that the most common error is omitting information, rather than hallucination.