Strategies Explained
Ian Frank, David Basin
To be presented at the Fifth Game Programming Workshop in Japan, 1999
The version on these pages is a slightly updated ETL Technical Report,
ETL-TR-99-32
Abstract:
For many problem-solving tasks, it is important not only to produce
solutions, but also to be able to explain them. In this paper, we
describe how we are addressing this question within the framework of
our computer game-playing research.
We build on the computer Bridge system FINESSE, which finds optimal
lines of play for single-suit Bridge problems. To explain FINESSE's
strategies, we developed an approach based on three distinct steps.
First, we identify from a strategy the possible sequences of MAX plays
that needs to be explained. Second, we remove from consideration move
sequences or game situations that would be considered too simple to
explain to human players. Third, we produce natural English text with
the aid of both game-general and game-specific patterns and idioms
that can explain each MAX and MIN move. We explain each of these
three steps, and demonstrate the effectiveness of the overall approach
by comparing automatically generated explanations against those found
in an expert Bridge text.
|