Nutri-bullets Hybrid: Consensual Multi-document Summarization

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5213–5222 June 6–11, 2021. ©2021 Association for Computational Linguistics 5213 Nutri-bullets Hybrid: Consensual Multi-document Summarization Darsh J Shah1 Lili Yu2 Tao Lei2 Regina Barzilay1 1Computer Science and Artificial Intelligence Lab Massachusetts Institute of Technology 2ASAPP, Inc. darsh@csail.mit.edu liliyu@asapp.com tao@asapp.com regina@csail.mit.edu Abstract We present a method for generating compara- tive summaries that highlights similarities and contradictions in input documents. The key challenge in creating such summaries is the lack of large parallel training data required for training typical summarization systems. To this end, we introduce a hybrid generation ap- proach inspired by traditional concept-to-text systems. To enable accurate comparison be- tween different sources, the model first learns to extract pertinent relations from input docu- ments. The content planning component uses deterministic operators to aggregate these re- lations after identifying a subset for inclu- sion into a summary. The surface realization component lexicalizes this information using a text-infilling language model. By separately modeling content selection and realization, we can effectively train them with limited annota- tions. We implemented and tested the model in the domain of nutrition and health – rife with inconsistencies. Compared to conventional methods, our framework leads to more faithful, relevant and aggregation-sensitive summariza- tion – while being equally fluent.1 1 Introduction Articles written about the same topic rarely exhibit full agreement. To present an unbiased overview of such material, a summary has to identify points of consensus and highlight contradictions. For in- stance, in the healthcare domain, where studies often exhibit wide divergence of findings, such comparative summaries are generated by human experts for the benefit of the general public.2 Ide- ally, this capacity will be automated given a large number of relevant articles and continuous influx of new ones that require a summary update to keep 1Our code and data is available at https://github.c om/darsh10/Nutribullets 2Examples include https://www.healthline.c om and https://foodforbreastcancer.com. Figure 1: We consider the database extracted from four Pubmed studies on Pears and Cancer. The key facts (bold) and consensus (contradiction) are realized in the text generated by our model. it current. However, standard summarization archi- tectures cannot be utilized for this task since the amount of comparative summaries is not sufficient for their training. In this paper, we propose a novel approach to multi-document summarization based on a neural interpretation of traditional concept-to-text gener- ation systems. Specifically, our work is inspired by the symbolic multi-document summarization system of (Radev and McKeown, 1998) which pro- duces summaries that explicitly highlight agree- ments, contradictions and other relations across input documents. While their system was based on human-crafted templates and thus limited to a narrow domain, our approach learns different com- ponents of the generation pipeline from data. To fully control generated content, we frame the task of comparative summarization as concept-to- text generation. As a pre-processing step, we ex- 5214 tract pertinent entity pairs and relations (see Figure 1) from input documents. The Content Selection component …

Content truncated to 500 words