other,10-6-H92-1045,bq be used as an additional source of <term> constraint </term> for improving the performance of
other,12-3-H92-1045,bq completion , we observed a very strong <term> discourse </term> effect . That is , if a <term> polysemous
other,20-5-H92-1045,bq share <term> sense </term> in the same <term> discourse </term> is extremely strong ( 98 % ) . This
other,11-1-H92-1045,bq </term> like <term> sentence </term> whose <term> meaning </term> or <term> sense </term> depends on the
other,13-1-H92-1045,bq </term> whose <term> meaning </term> or <term> sense </term> depends on the context of use .
other,31-4-H92-1045,bq that they will all share the same <term> sense </term> . This paper describes an experiment
other,16-5-H92-1045,bq found that the tendency to share <term> sense </term> in the same <term> discourse </term>
other,9-1-H92-1045,bq <term> polysemous words </term> like <term> sentence </term> whose <term> meaning </term> or <term>
other,9-4-H92-1045,bq <term> polysemous word </term> such as <term> sentence </term> appears two or more times in a <term>
hide detail