Spam Filter Investigation

RSS
Condividi

Mar 26 2007, 0:10

In the recent (slightly problematic) version of the submissions server, we've added logging of items which are rejected by the spam filter for content reasons. I want to try and use these to evaluate whether we should remove some rules.

Here is a little summary (URLs have been broken by me to stop them profiting off my pagerank and the forums' auto-linking).

Top Rejected Artists:

14633 various artists
6193 various
5662 va
4970 artist
3130 unknown artist
2905 no artist
1287 (+44)
1175 unknown
208 [amatory]
182 various artists dubstep allstars vol. 4 - disc 2 (mixed by dj hatcha)
133 [ingenting]
132 mosaic.wav
122 [game music]
116 v/a
106 (smog)
88 noartist
71 [spunge]
66 [o3]
58 http://spyd3 r.ru/ipbforum
58 elliot goldenthal & various artists
50 [:sitd:]
46 various artists jon brion
45 [soundtrack]
41 (la vela puerca)
32 [www.potencia dance.com]
32 (delerium)
29 [raised fist]
28 various artists - system recordings
28 [unknown]
28 (by dj morphologik)

Top Rejected Tracks:

1742 track 10
1736 track 01
1666 track 02
1632 track 03
1573 (antichrist television blues)
1528 track 11
1507 track 04
1438 track 05
1317 track 06
1311 track 12
1233 track 07
1220 track 08
1138 track 13
1043 track 09
977 track 1
952 track 14
896 track 2
780 track 3
721 track 15
710 track 4
706 track 6
696 track 5
618 track 16
584 track 7
566 audiotrack 01
562 track 8
535 track 9
507 track 17
496 audiotrack 02
494 audiotrack 03

Top tracks (ignoring "Track" and "Audiotrack" filters):

1573 (antichrist television blues)
351 (untitled)
294 (nice dream)
186 (sic)
132 (exchange)
130 (*fin)
120 (intro)
102 bohemian rhapsody.mp3
78 albumwrap album
77 (manifest)
74 (10) and counting
73 (515)
71 (interlude)
68 (c'estlavie; en)
66 (d▒but)
63 (hospital vespers)
61 (letter from home)
58 (intermission)
56 childhood (1)
54 (camel; ch)
52 (hidden track)
44 (unlisted)
43 (12) bother
42 (something)
39 (past due)
33 (a)
31 (hidden finale)
28 (subversive script)
28 (b)
27 (reverse)

Commenti

  • lotrgamemast

    (+44) is the proper name for a band and (sic) is the title of a song off of slipknots self titled album.

    Mar 26 2007, 17:46
  • closedmouth

    MOSAIC.WAV, [spunge], (Smog) All real artists

    Mar 27 2007, 6:26
  • AdriaNnLA

    Is it possible to provide the filter with exceptions to the rules? Also, this post reminds me of a question I've been meaning to ask somewhere... With re: to (possible) MB (maybe, someday) integration, how has the conversation developed in light of the newer PUID replacing the older TRM MBID? Use of Picard now exclusively tags with the PUID (when available).

    Mar 27 2007, 21:21
  • Russ

    Following this analysis I am almost certainly going to remove the artist and track rules with parentheses, enabling stuff like (Smog) and (nice dream), and that popular-seeming Arcade Fire track. I am less convinced with the square brackets as yet, but I really don't want to go the way of the whitelist (bearing in mind that when we finally get Musicbrainz rocking, MB will act as one big spam filter whitelist).

    Mar 28 2007, 0:44
  • -VVV-

    [ingenting] is a Swedish band. I got the spam filter message today when I listened their song through the flash player.

    Apr 28 2007, 9:45
  • matt85

    track 10 is my favourite too.

    Set 11 2008, 8:43
Visualizza tutti (7 commenti)
Aggiungi un commento. Accedi a Last.fm o registrati (è gratuito).