In this paper, we discussthe differentstrategiesusedin COMET (COordinated Multimedia Explanation Testbed) for selecting words withwhichthe user is familiar. Whenpicturescannotbe used to disambiguate a word or phrase, COMET has four strategies for avoiding unknown words. We give examplesfor each of these strategies and showhow they are implementedin COMET.
Kathleen McKeown, Jacques Robin, Michael A. Tanenb