Question answering (QA) over text passages is a problem of long-standing\ninterest in information retrieval. Recently, the conversational setting has\nattracted attention, where a user asks a sequence of questions to satisfy her\ninformation needs around a topic. While this setup is a natural one and similar\nto humans conversing with each other, it introduces two key research\nchallenges: understanding the context left implicit by the user in follow-up\nquestions, and dealing with ad hoc question formulations. In this work, we\ndemonstrate CROWN (Conversational passage ranking by Reasoning Over Word\nNetworks): an unsupervised yet effective system for conversational QA with\npassage responses, that supports several modes of context propagation over\nmultiple turns. To this end, CROWN first builds a word proximity network (WPN)\nfrom large corpora to store statistically significant term co-occurrences. At\nanswering time, passages are ranked by a combination of their similarity to the\nquestion, and coherence of query terms within: these factors are measured by\nreading off node and edge weights from the WPN. CROWN provides an interface\nthat is both intuitive for end-users, and insightful for experts for\nreconfiguration to individual setups. CROWN was evaluated on TREC CAsT data,\nwhere it achieved above-median performance in a pool of neural methods.\n