Learning to generate one-sentence biographies from Wikidata
This addresses the problem of automated text generation for structured data, with incremental improvements in fact retention and quality for applications like summarization.
The paper tackles generating one-sentence Wikipedia biographies from Wikidata facts using a recurrent neural network with a novel secondary objective to ensure factual accuracy, achieving a BLEU score of 41 and performing nearly as well as human references in evaluations.
We investigate the generation of one-sentence Wikipedia biographies from facts derived from Wikidata slot-value pairs. We train a recurrent neural network sequence-to-sequence model with attention to select facts and generate textual summaries. Our model incorporates a novel secondary objective that helps ensure it generates sentences that contain the input facts. The model achieves a BLEU score of 41, improving significantly upon the vanilla sequence-to-sequence model and scoring roughly twice that of a simple template baseline. Human preference evaluation suggests the model is nearly as good as the Wikipedia reference. Manual analysis explores content selection, suggesting the model can trade the ability to infer knowledge against the risk of hallucinating incorrect information.