Analyzing Torrents

A part of the development of Aufero is inevitably analyzing torrents and naming of them. This means being able to guess what is tagging and what is actually a relevant part of the name during runtime. For instance, common words like the, is, go, have, from should be disregarded in many cases. Then we have what I call "torrent noise" which among a lot of other words can be r00, release, version. We also have a bit more interesting things that is tagging of a Torrent: xvid, dvdrip, r5, proper, pal, ntsc... Then of course, the title -- which is essentially what we want to get at.

The hardest thing in Aufero is undoubtedly finding which word falls into what category seeing as words come and go pretty rapidly. Aufero relies on other ways of finding the one torrent after this initial analysis, but that is out of scope for this post.

To the point, I ran a little snippet of code today on a subset of some 3-4 months of torrents to see what is actually "popular" out there (popular in this case is what is "hot" and what was deemed worthy of being uploaded). The list below is kind of an inverted cloud, instead of what is being searched for, it says what is available (or maybe it falls under the category of cloud anyway -- I just couldn't be arsed to set fontsize on them)...

This is the top 100 list with a few of the filters above applied:

Pos Count Word
---- ------- -------------------------
1 1,214 pro
2 845 vista
3 755 windows
4 597 bt
5 595 colombo
6 499 xp
7 459 tntvillage
8 407 professional
9 388 games
10 355 tv
11 347 aio
12 334 office
13 321 converter
14 311 world
15 297 ultimate
16 291 live
17 286 info
18 262 speed
19 256 enterprise
20 248 man
21 247 diamond
22 243 studio
23 239 severedbytes
24 239 thepeerhub
25 237 ps2
26 227 http
27 223 antivirus
28 220 avithe
29 220 internet
30 220 player
31 214 stv
32 210 premium
33 209 black
34 208 collection
35 205 life
36 205 media
37 203 fs
38 200 int
39 197 last
40 197 war
41 196 night
42 196 royale
43 196 vs
44 195 pl
45 190 line
46 189 del
47 185 x264
48 183 pdtv
49 183 security
50 181 king
51 178 dead
52 178 deluxe
53 176 tus
54 171 online
55 166 ipodnova
56 164 best
57 164 jess
58 164 season
59 162 complete
60 161 image
61 161 platinum
62 160 es
63 159 italian
64 159 newpct
65 158 wars
66 157 dsr
67 157 good
68 157 manager
69 157 nds
70 156 md
71 155 men
72 150 microsoft
73 147 anti
74 147 museum
75 146 magic
76 144 balboa
77 144 doctor
78 144 gold
79 144 sp2
80 144 team
81 143 server
82 142 photoshop
83 142 warcraft
84 140 tc
85 138 elite
86 136 psxpsp
87 135 rmvb
88 133 casino
89 133 download
90 132 burning
91 132 suite
92 132 vegas
93 131 earth
94 131 evil
95 131 mvs
96 131 nero
97 131 virus
98 130 ipod
99 130 pepedivx
100 130 star

There are not many surprises in there to be honest, at position 245 in the list we find "Linux", which was a lot higher a few months back. Not sure why it dropped so far. Microsoft is on 150 as a comparison.

Anyway, this was just a funny little tidbit. I saw there were a few comments below, I haven't had a chance to respond to anything, but I will!


Aufero

PS. I got myself an email-address too (see profile)! Welcome me to 2007!

PPS. A late edit: Perhaps I should add this thing as a background effect on main-screen of Aufero or something? I'm not entirely sure what it shows yet, but I'm sure it shows something!