Comparision of POS Tagging using different GRU cell mutations

Code submitted as mini project for partial fullfilment of M.Tech degree.

Setup

Ensure that you have python2/3.

This project is organized into 3 folders literature, logs, report_ppt and 5 files gru.py, main.py, pos_tagger.py, train.py, TreeBankDataSet.py

PosTagging
├── gru.py
├── literature
│   └── Empirical Analysis of RNN.pdf
├── logs
│   └── logs_400_400_0.1_10_40_s_Adam_treebank_nll_0.8_file.log
├── main.py
├── pos_tagger.py
├── README.md
├── report_ppt
│   ├── POS Tagging using GRU.pdf
│   ├── ppt.pdf
│   └── report_sc18m002_pos_tagging.lyx~
├── train.py
└── TreeBankDataSet.py

literature folder contains the base research paper from which the idea is inspired. logs folder contains generated logs for experiment. report_ppt folder contains final report and presentation.

Hardware requirements

This project requires an NVIDIA-GPU with CUDA 9.0 to run PyTorch with GPU computing capabilities.

Dataset

We have used TreeBankDataSet for the expermients.

Code

• main.py is starting script for running the experiment
• train.py is training script
• pos_tagger.py architecture of generic pos_tagger
• gru.py, contains architecture of GRU cell

Training the model (train_code)

Start the training process by calling main.py