Nieto and colleagues at New York University wrote a paper in 2014 looking at how humans evaluate systems that automatically find section boundaries in songs (e.g. between verse and chorus).

They evaluated precision, recall and F-measure as a way of approximating to human judgements. They found that precision was far more important than recall in explaining people's responses.

What might this imply for the standard F-measure? Select ONE observation that best applies.

a.

Since the F-measure is based on the harmonic mean of precision and recall, it should give a sweet spot for evaluation.

b.

The standard F-measure favours high scores on one measure even if the other score is low. This will work poorly here.

c.

The standard F-measure penalises low scores on one measure even if the other score is high. This will work well here.

d.

The standard F-measure penalises low scores on one measure even if the other score is high. This will work poorly here.

e.

The standard F-measure gives equal weight to high precision and high recall, so will perform poorly if one is more important than the other.

f.

The standard F-measure favours high scores on one measure even if the other score is low. This will work well here.