You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@opennlp.apache.org by William Li <wp...@MIT.EDU> on 2013/07/05 16:26:54 UTC

MaxEnt model file interpretation

Hi,

I have a question similar to the one below:
http://sourceforge.net/p/maxent/discussion/18385/thread/aa4c0e30/

I have a maxent model for multi-class classification working well, but I
would like to understand the MaxEnt model file better. Is there
documentation, or can someone explain what each line of the trained model
file means? I'm especially unclear on what terms like "8 0 1 2 3" mean.
Also, my model has 6462 features and 54374 model weights. If there were one
parameter for each feature and class, there would be 6462*9 = 58158
weights. I'm guessing that the model doesn't assign weights to certain
classes and features, but how I can tell from the file?

Thanks,
William


Specifically, my model consists of something that looks like this:

GIS
1
0.0
9
class1
class2
class3
class4
class5
class6
class7
class8
class9
375
8 0 1 2 3
14 0 1 2 3 4
15 0 1 2 3 4 5
93 0 1 2 3 4 5 6
706 0 1 2 3 4 5 6 7
2372 0 1 2 3 4 5 6 7 8
87 0 1 2 3 4 5 6 8
32 0 1 2 3 4 5 7
51 0 1 2 3 4 5 7 8
14 0 1 2 3 4 5 8
40 0 1 2 3 4 6
116 0 1 2 3 4 6 7
159 0 1 2 3 4 6 7 8
32 0 1 2 3 4 6 8
18 0 1 2 3 4 7
20 0 1 2 3 4 7 8
12 0 1 2 3 4 8
7 0 1 2 3 5
28 0 1 2 3 5 6
<snip>
6462 // number of features
feature1
feature2
...
feature6462
0.1231
0.123213
<snip>





William Li