Published August 23, 2018 | Version v1
Software Open

Word Segmentation in Sanskrit Using Energy Based Models

  • 1. IIT Kharagpur
  • 2. American Express
  • 3. Chinmya Visvavidyapeeth

Description

This is the repository for word segmentation in sanskrit using energy based models.

 

# Word Segmentation in Sanskrit Using Energy Based Models

 
## Getting Started
 
Please download the 2 compressed files 'dir.zip' and 'wordsegmentation.rar' to your working directory and extract them into folders named 'dir' and 'wordsegmentation' respectively.
 
Your working directory should be as follows
* Working Directory
  * wordsegmentation
    * skt_dcs_DS.bz2_4K_bigram_mir_10K
    * skt_dcs_DS.bz2_4K_bigram_mir_heldout
  * dir
 
## Prerequisites
* Python3
  * scipy
  * numpy
  * csv
  * pickle
  * multiprocessing
  * bz2
## Instructions for Training
Change your current directory to 'dir'
 
Run the file Train_clique.py by using the following command
 
* python Train_clique.py
 
To train on different input features like BM2,BM3,BR2,BR3,PM2,PM3,PR,PR3 please modify the bz2_input_folder value in the main function before beginning the training.
 
Feature  | bz2_input_folder
------------- | -------------
BM2 | wordsegmentation/skt_dcs_DS.bz2_4K_bigram_mir_10K/
BM3 | wordsegmentation/skt_dcs_DS.bz2_1L_bigram_mir_10K
BR2 | wordsegmentation/skt_dcs_DS.bz2_4K_bigram_rfe_10K/
BR3 | wordsegmentation/skt_dcs_DS.bz2_1L_bigram_rfe_10K/
PM2 | wordsegmentation/skt_dcs_DS.bz2_4K_pmi_mir_10K/
PM3 | wordsegmentation/skt_dcs_DS.bz2_1L_pmi_mir_10K2/
PR2 | wordsegmentation/skt_dcs_DS.bz2_4K_pmi_rfe_10K/
PR3 | wordsegmentation/skt_dcs_DS.bz2_1L_pmi_rfe_10K/
 
## Instructions for Testing
 
After training, please modify the 'modelList' dictionary  in 'test_clique.py' with the name of the neural network that has been saved during training. While testing for a feature, please provide the name of the neural net which was trained for the same feature.
 
We only provide the trained model for the feature BM2 which was our best performing feature. If the name of the neural net is not changed, then the testing will be performed on the pre-trained model for BM2 provided in outputs/train_t7978754709018
 
To test with a particular feature vector use the tag of the feature while execution
 
* python test_clique.py -t <tag>
 
For example:  
  * python test_clique.py -t BM2
 
After finishing the testing please run the following command to see the precision and recall values for both the word and word++ prediction tasks
 
* python evaluate.py <tag>
 
For example:  
  * python evaluate.py BM2

Files

dir.zip

Files (42.2 GB)

Name Size Download all
md5:016462cbd311404a6c9fb9af950d38a5
453.2 MB Preview Download
md5:c0163b57ec0ab0013603d017556e2f2b
2.4 kB Preview Download
md5:6339b68e76df5aab37d2850fccf68c98
41.7 GB Download