Towards multi-document summarization in the open-domain

Multi-document summarization (MDS) traditionally assumes a set of topic-related documents are provided. However, this document set is often an artifact of the dataset curation process; in practice, it is not necessarily available and would need to be retrieved given an information need, i.e. a question or topic statement. We study this more challenging "open-domain" setting by formalizing the task and bootstrapping it using existing datasets, retrievers and summarizers. Via extensive experimentation, we determine that: (1) state-of-the-art summarizers suffer large reductions in performance when applied to the open-domain, even when retrieval performance is high, (2) additional training in the open-domain setting can reduce this sensitivity to imperfect retrieval, and (3) summarizers are insensitive to the retrieval of duplicate documents and the order of retrieved documents, but highly sensitive to other errors, like the retrieval of irrelevant documents. Based on our results, we provide practical guidelines to enable future work on open-domain MDS, e.g. how to choose the number of retrieved documents to summarize. Our results suggest that new methods for retrieval and summarization, as well as annotated resources for training and evaluation, will be necessary for further progress in the open-domain.