Copyright 2010 Google Inc.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

-----------------------------------------------------------------------------------
Scripts that enable third parties to replicate the results reported in the paper: 

@InProceedings{chelba:interspeech2010,
  title =        {{Study on Interaction between Entropy Pruning and Kneser-Ney Smoothing}},
  author =       {Ciprian Chelba, Thorsten Brants, Will Neveitt, Peng Xu},
  booktitle =    {Proc. Interspeech},
  year = 	 {2010},
  address = 	 {Makuhari, Japan},
  pages =	 {2242-2245},
  month = 	 {September},
  organization = {ISCA},
  url = {http://research.google.com/pubs/pub36472.html}
}

The data itself (Linguistic Data Consortium CD "CSR LM 96 CD-ROM release disk 1") 
is not part of the release. 

However, anyone that has it will be able to reproduce our experiments using the 
scripts provided here. It should be easy to generate the data, and then run the 
experiments:

0. Make sure you have perl, python and a version of the SRILM toolkit installed.

1. Copy BN data from CSR LM 96 CD-ROM release disk 1 --- "conditioned" data 
(or "vp", for "verbalized punctuation") version of the language modeling 
data for the 1996 Continuous Speech Recognition Hub 4 Evaluation --- 
to $TEXTTOP/Resources/LDC/vp_train/bn*Z.

2. Run ./Resources/bin/PrepLMData.sh to get the LM training/test data.

3. Edit the variables at the top of ./Resources/bin/Run.pl and run it to get
Table 3 in our Interspeech 2010 paper.