Multi-hop Inference for Question-driven Summarization
Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)
Question-driven summarization has been recently studied as an effective approach to summarizing the source document to produce concise but informative answers for nonfactoid questions. In this work, we propose a novel question-driven abstractive summarization method, Multi-hop Selective Generator (MSG), to incorporate multi-hop reasoning into question-driven summarization and, meanwhile, provide justifications for the generated summaries. Specifically, we jointly model the relevance to the
... levance to the question and the interrelation among different sentences via a human-like multi-hop inference module, which captures important sentences for justifying the summarized answer. A gated selective pointer generator network with a multi-view coverage mechanism is designed to integrate diverse information from different perspectives. Experimental results show that the proposed method consistently outperforms stateof-the-art methods on two non-factoid QA datasets, namely WikiHow and PubMedQA. Question: Are human coronaviruses uncommon in patients with gastrointestinal illness? Document: Coronaviruses infect numerous animal species causing a variety of illnesses including respiratory, neurologic and enteric disease. Human coronaviruses (HCoV) are mainly associated with respiratory tract disease but have been implicated in enteric disease. To investigate the frequency of coronaviruses in stool samples from children and adults with gastrointestinal illness by RT-PCR. Clinical samples submitted for infectious diarrhea testing were collected from December 2007 through March 2008. RNA extraction and RT-PCR was performed for stools negative for Clostridium difficile using primer sets against HCoV-229E, HCoV-OC43, HCoV-NL63, and HCoV-HKU1. Clinical data from samples positive for coronaviruses were reviewed and recorded. Samples from 479 patients were collected including 151 pediatric (< or = 18 years), and 328 adults (>18 years). Of these samples, 4 patients (1.3%, 2 adult; 2 pediatric) screened positive for the presence of a coronavirus. All detected coronaviruses were identified as HCoV-HKU1. No stools screened positive for either HCoV-229E, HCoV-NL63 or HCoV-OC43. All HCoV-HKU1 positive samples occurred between mid-January to mid-February. Clinical manifestations from HCoV-HKU1 positive patients included diarrhea, emesis and respiratory complaints. Three (75%) patients were admitted to the hospital with a median length of stay of 6 days. Answer: Coronaviruses as a group are not commonly identified in stool samples of patients presenting with gastrointestinal illness. HCoV-HKU1 can be identified in stool samples from children and adults with gastrointestinal disease, with most individuals having respiratory findings as well. No stool samples screened positive for HCoV-NL63, HCoV-229E, or HCoV-OC43.