A machine learning system trained on scholarly journals could correct Wikipedia's gendered under-representation problem

boing
11

So most of the data it’s trained on is written by men and most of the data it’s processing is written by men. The algorithm could work perfectly, but it’s built on sand.