Researchers are attempting to develop computer software which can automatically create summaries of news articles. Sometimes the purpose of such summaries is to contribute to answering a question. In order to evaluate how well the software is doing, researchers would like to compare the summaries their software systems produce with those that humans create. Your task is to create some summaries that will be used to evaluate automatic summarization systems. If you have questions about this procedure you can contact Lori Buckland... Task 5 During January you will spend some time at NIST using our search engine to explore newswire documents and come up: - 10 questions of the form "Who is X?", where X is the name of a person or group of people - sets of 10 documents each, such that each document in a given set, contributes to answering the associated question. Different documents may contribute different amounts of material. There may be some repetition within a set. It would be VERY helpful if you generated 15 or so such questions before coming to NIST in January. Try to avoid choosing too many people or groups of people who were in the news every(other) day. The sources will be the following: * AP newswire, 1998-2000 * New York Times newswire, 1998-2000 * Xinhua News Agency (English version), 1996-2000 Later, NIST will send you printed versions of 25 sets of documents, some you chose and some chosen by others, along with the question that is associated with the each set. Your task will then be to create for each set a short summary (less then 665 characters including punctuation and whitespace) which is focused by the goal of responding to the question.