Hello to all lovers to look at useless charts!
Several times I heard the opinion of people who are versed in music that the listener is biased towards the timbre of the performers.
In particular, it is alleged that the gender of the vocalist influences our preferences. If it’s very rude, then “boys ceteris paribus more like listening to girls, and vice versa.”
This seems to be one of such statements, the verification of which on real data is completely useless in life, but terribly interesting.
If you adhere to the same point of view - welcome to kat!
, , - . , . -, - - .
. ? . - — , . -, .
, , , — . "". .
, "".
, !
- -
- ,
API
API . .
, :
- vk
- https://vk.com/apps?act=manage
- ( Standalone-, , )
- ID
- , ID
- acces token,
- ,
- , "", .
— API , , - .
, . , .
, . .
.
"". 1000 (-) .
"". 80-90 . . API .
, . "": 1000 . , .
1000 19 4 , .
, 500 .
. " ", , .
-, — 55% . 50 50 .
, bar chart .
, , .
, , confidence
h = std_err * scipy.stats.t.interval(confidence, nn-1, loc=0, scale=1)[1]/math.sqrt(nn)
,
std_err — , np.std(data)
nn —
.. , 90% ( 100 , 10 ) .
"" 1000 .
:
—
—
— "50 50"
( "", "", -).
, :
group_gender: 1 — , 2 —
mean: — , —
?
-, . , "" . , .
-, . ("", "" ""). — ("Nautilus Pompilius", "" "" ).
-, . , . (5 5) . - - . "" , , , ("", "", ""), ("-2", " " , , ""). .
?
" " — , . , .
The basic hypothesis "the name of the group in the top should be" is generally confirmed. If you work a little with tokens, you can highlight the "distance" between related groups. But the floor just does not reach.
So just admire the top in the form of word clouds:
A lot of pictures
, .
""
"" . "" "", .
. . "", .
"" .
"" . "", "" "" .
: "", "", "", .
"-2" . "".
.
"" "" , "".
.
"-" -
"Radiohead" .
"" , "", "" "" .
"" .
"" - .
From the constructive part for today. The source code is available to everyone .
If you want to see something else on this dataset - write in the comments.
I ask you not to take this "study" more seriously than it's worth it. This is nothing more than satisfying curiosity and experimenting with new tools for me.