140
0.7
0.72
0.74
0.76
0.78
0.8
0.82
0.84
0.86
0.88
0.9
0
5
10
15
20
25
30
35
40
precision
k
Precision for Artist Categorization of 224
test set
last.fm data
dm
pm
Figure 6.12. Precision of the 224 artist categorization for k-NN using Last.fm and
the two best web-based methods.
correctly classified. For values of
k larger than 1, the performance using the track
filtered data is equal to the one using either the raw or the normalized
Last.fm data.
As the results of the genre categorization using the
Last.fm data are equally
good as those gained with the best methods using arbitrary web-data, we conclude
that
DM
and
PM
are reliable methods for this classification task. We also observe
that there is no complete overlap with the data extracted from
Last.fm and the
ground truth composed by experts in the field. This on the one hand gives confi-
dence in our methods, but on the other hand raises questions on the fact that not
all artist-genre combinations are recognized by the general public. We therefore
investigate the use of
Last.fm data as a ground truth in the last part of this section.
Categorizing Painters into Movements
For this experiment, we constructed a list of painters
I
Dostları ilə paylaş: