Re:Science, Now Under Scrutiny Itself
by Al-Khwarizmi on Thursday June 18, 2015 @04:07AM
(#9737329)
In computer science, the case where the easier thing to do is to replicate something does happen sometimes.
For example if you create a natural language parser that you claim is more accurate than another, you may need to replicate the other in order to compare accuracies. Sometimes you can just take the authors' reported accuracy, but often it's not that easy (maybe they just tested on English and you're interested in performance in other languages, maybe you want a different metric, maybe you want to compare on some standard set of features, etc.)
For example if you create a natural language parser that you claim is more accurate than another, you may need to replicate the other in order to compare accuracies. Sometimes you can just take the authors' reported accuracy, but often it's not that easy (maybe they just tested on English and you're interested in performance in other languages, maybe you want a different metric, maybe you want to compare on some standard set of features, etc.)