The SENSEI Annotated Corpus: Human Summaries of Reader Comment Conversations in On-line News.

Researchers are beginning to explore howto generate summaries of extended argumentativeconversations in social media,such as those found in reader comments inon-line news. To date, however, there hasbeen little discussion of what these summariesshould be like and a lack of humanauthoredexemplars, quite likely becausewriting summaries of this kind of interchangeis so difficult. In this paper wepropose one type of reader comment summary– the conversation overview summary– that aims to capture the key argumentativecontent of a reader commentconversation. We describe a method wehave developed to support humans in authoringconversation overview summariesand present a publicly available corpus –the first of its kind – of news articles pluscomment sets, each multiply annotated,according to our method, with conversationoverview summaries.