Web archives constitute an increasingly important source of information for computer scientists, humanities researchers and journalists interested in studying past events.
However, currently there are no access methods that help Web archive users to efficiently access event-centric information in large-scale archives that go beyond the retrieval of individual disconnected documents. In this article, we tackle the novel problem of extracting interlinked event-centric document collections from large-scale Web archives to facilitate an efficient and intuitive access to information regarding past events. We address this problem by: 1) facilitating users to define event-centric document collections in an intuitive way through a Collection Specification; 2) development of a specialised extraction method that adapts focused crawling techniques to the Web archive settings; and 3) definition of a function to judge the relevance of the archived documents with respect to the Collection Specification taking into account the topical and temporal relevance of the documents. Our extended experiments on the German Web archive (covering a time period of 19 years) demonstrate that our method enables efficient extraction of event-centric collections for different event types.