Strategies Explained

the 5th Game Programming Workshop in Japan, 1999 TR-99-32 Re eived 07 O t 99 Strategies Explained Ian Frank Ele trote hni al Laboratory Ma hine Inferen e Group Umezono 1-1-4, Tsukuba Ibaraki, JAPAN 305 David Basin Institut f ur Informatik Universitat Freiburg Am Flughafen 17 Freiburg, Germany Abstra t For many problem-solving tasks, it is important not only to produ e solutions, but also to justify any solutions with explanations. In this paper, we des ribe how we are addressing this question within the framework of our resear h on omputer game-playing. We build on the omputer Bridge system Finesse, whi h nds optimal strategies for single-suit Bridge problems. To explain Finesse's strategies, we have developed an approa h based on three distin t steps. First, we identify from a strategy the possible sequen es of MAX plays that need to be explained. Se ond, we remove from onsideration move sequen es or game situations that would be onsidered too simple to explain to human players. Third, we produ e natural English text with the aid of both game-general and game-spe i patterns and idioms that an explain ea h MAX and MIN move. We demonstrate the e e tiveness of this approa h by omparing automati ally generated explanations against those found in an expert Bridge text.