“Gigabye LM”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
3. word-based 3-gram
2. pruning the 4k 3gram LM.
第29行: 第29行:
 
== 2. pruning the 4k 3gram LM. ==
 
== 2. pruning the 4k 3gram LM. ==
  
Model 1gram 2gram 3gram             size       ppl
+
{|
1          1e-7  1e-7     1e-7              30M    logprob= -8.55982e+06 ppl= 102.796 ppl1= 111.532
+
! Model ||2gram ||3gram ||            size ||      ppl
2          1e-6  1e-6     1e-6              5M      logprob= -9.26982e+06 ppl= 150.96  ppl1= 164.9
+
|-
3           1e-7  1e-6.5    1e-6.5          11M      logprob= -9.09681e+06 ppl= 137.467 ppl1= 149.913
+
| 1         ||           1e-7   || 1e-7         ||    30M    logprob= -8.55982e+06 ppl= 102.796 ppl1= 111.532
 
+
|-
 +
|2           ||           1e-6   || 1e-6         ||    5M      logprob= -9.26982e+06 ppl= 150.96  ppl1= 164.9
 +
|-
 +
| 3         ||          1e-7   || 1e-6         ||    11M      logprob= -9.09681e+06 ppl= 137.467 ppl1= 149.913
 +
|-
 +
|}
  
 
== 3. word-based 3-gram ==
 
== 3. word-based 3-gram ==

2012年9月13日 (四) 05:57的版本

1. very initial, without any prunning, character based. Here is the size and perplexity.

The training is with Gigabytes except the cna data, and ppl testing is based on a sub set from the cna data (big52gb applied)

2gram:

25M 2gram.4000.gz: 0 zeroprobs, logprob= -9.39983e+06 ppl= 161.965 ppl1= 177.141


3gram:

47M 3gram.500.gz:0 zeroprobs, logprob= -6.34868e+06 ppl= 85.1361 ppl1= 94.2525

117M 3gram.1000.gz  :0 zeroprobs, logprob= -7.43809e+06 ppl= 80.6408 ppl1= 87.7439

195M 3gram.2000.gz:0 zeroprobs, logprob= -7.95872e+06 ppl= 79.9875 ppl1= 86.5196

221M 3gram.3000.gz:0 zeroprobs, logprob= -8.04799e+06 ppl= 80.2418 ppl1= 86.7277

229M 3gram.4000.gz:0 zeroprobs, logprob= -8.15697e+06 ppl= 82.6585 ppl1= 89.3392

4gram:

205M 4gram.500.gz:0 zeroprobs, logprob= -6.25395e+06 ppl= 79.6739 ppl1= 88.0716

472M 4gram.1000.gz:0 zeroprobs, logprob= -7.21607e+06 ppl= 70.737 ppl1= 76.774


2. pruning the 4k 3gram LM.

Model 2gram 3gram size ppl
1 1e-7 1e-7 30M logprob= -8.55982e+06 ppl= 102.796 ppl1= 111.532
2 1e-6 1e-6 5M logprob= -9.26982e+06 ppl= 150.96 ppl1= 164.9
3 1e-7 1e-6 11M logprob= -9.09681e+06 ppl= 137.467 ppl1= 149.913

3. word-based 3-gram

tri-gram size:

org th-7 th-7/6 th-6
10k: 52M 23M 8M 4M
20k: 57M 24M 9M 4M

final fst size:

org th-7 th-7/6 th-6
10k: - 770M 193M 135M
20k: - - - -