other,23-2-N03-2003,bq | match the <term> style </term> and/or <term> | topic | </term> of the target <term> recognition task | #3051 In this paper, we show how training data can be supplemented with text from the web filtered to match the style and/or topic of the target recognition task, but also that it is possible to get bigger performance gains from the data by using class-dependent interpolation of N-grams. |