This is a reconstruction of the paper Free as in Free Word Order: An Energy Based Model for Word Segmentation and Morphological Tagging in Sanskrit
You can refer to the original repository here
├── dir/
├──wordsegmentation/
├── skt_dcs_DS.bz2_4K_bigram_mir_10K/
├── skt_dcs_DS.bz2_4K_bigram_mir_heldout/
Change your current directory to 'dir'
Run the file Train_clique.py by using the following command
python Train_clique.py
TRAINING OUTPUTS are already stored in dir/outputs/train_t3896665073989.
NOTE: 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 Code | bz2_input_folder Path |
|---|---|
| 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/ |
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
#Citing
@inproceedings{krishna-etal-2018-free,
title = "Free as in Free Word Order: An Energy Based Model for Word Segmentation and Morphological Tagging in {S}anskrit",
author = "Krishna, Amrith and
Santra, Bishal and
Bandaru, Sasi Prasanth and
Sahu, Gaurav and
Sharma, Vishnu Dutt and
Satuluri, Pavankumar and
Goyal, Pawan",
editor = "Riloff, Ellen and
Chiang, David and
Hockenmaier, Julia and
Tsujii, Jun{'}ichi",
booktitle = "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing",
month = oct # "-" # nov,
year = "2018",
address = "Brussels, Belgium",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/D18-1276/",
doi = "10.18653/v1/D18-1276",
pages = "2550--2561",
abstract = "The configurational information in sentences of a free word order language such as Sanskrit is of limited use. Thus, the context of the entire sentence will be desirable even for basic processing tasks such as word segmentation. We propose a structured prediction framework that jointly solves the word segmentation and morphological tagging tasks in Sanskrit. We build an energy based model where we adopt approaches generally employed in graph based parsing techniques (McDonald et al., 2005a; Carreras, 2007). Our model outperforms the state of the art with an F-Score of 96.92 (percentage improvement of 7.06{\%}) while using less than one tenth of the task-specific training data. We find that the use of a graph based approach instead of a traditional lattice-based sequential labelling approach leads to a percentage gain of 12.6{\%} in F-Score for the segmentation task."
}