Nutri-bullets: Summarizing Health Studies by Composing Segments
This addresses the problem of summarizing health studies for researchers or practitioners, but it is incremental as it builds on existing summarization methods with a novel approach for limited data.
The paper tackles multi-document summarization for health and nutrition by introducing Nutri-bullets, using an extract-compose model to improve faithfulness and relevance, achieving over 50% improvement on these metrics in the BreastCancer dataset.
We introduce \emph{Nutri-bullets}, a multi-document summarization task for health and nutrition. First, we present two datasets of food and health summaries from multiple scientific studies. Furthermore, we propose a novel \emph{extract-compose} model to solve the problem in the regime of limited parallel data. We explicitly select key spans from several abstracts using a policy network, followed by composing the selected spans to present a summary via a task specific language model. Compared to state-of-the-art methods, our approach leads to more faithful, relevant and diverse summarization -- properties imperative to this application. For instance, on the BreastCancer dataset our approach gets a more than 50\% improvement on relevance and faithfulness.\footnote{Our code and data is available at \url{https://github.com/darsh10/Nutribullets.}}