This paper introduces S_Covid, an end-to-end unsupervised learningbased question-answering engine for exploring COVID-19 scientific literature collections. S_Covid enables documents exploration for finding relevant research literature that most possibly contains information that can answer a user query. Thus, S_Covid pinpoints sentences out of research papers that can be possible answers to complex COVID-19 related user queries. We conducted experiments on 80,000 COVID-19 related papers collection. The paper shows statistically how the model performs but also through the feedback of real users. It also compares S_Covid with existing search engines addressing information retrieval of COVID-19 scientific literature.