Punctuation_Restoration

Source attribution:

BERT punctuator Framework code borrowed from here

Modifications have been made to the notebooks, the data preprocessing, and the model architecture.

Code modified based on here by Abhinav Rao.

Model available at https://drive.google.com/file/d/1bC-a0BrbVR3wsh2JZqu9KiMBf7nOK3T-/view?usp=sharing

Pre-built:

Get the dataset here. Store the dataset in any folder and provide the necessary arguments to the model

The dataset folder structure is as follows:

dataset/
|
|-dual/
|   |
|   |-xlm-roberta-base/
|            |-- train_data.pkl
|            |-- valid_data.pkl
|            |-- test_data.pkl
|-en
| |-...
|
|-zh

Use the --data-path argument to set the data directory:

  --data-path DATA_PATH
                        path to dataset directory

Creation (WIP):

Source for the IWSLT dataset is over here (train and dev) and here(test). The notebooks folder has the necessary notebooks for the dataset creation.
WIP: converting the notebooks to python scripts.

Installation and Usage:

Virtual environment:

It is HIGHLY RECOMMENDED to use a virtualenv or a conda virtual environment for this project:

1. Setup is fairly easy, open up a terminal and do:

• python -m venv /path/to/root/dire/.../.../venv_name or - conda create --name myenv

2. Then everytime you want to run the program, just do
   • source ./venv/bin/activate or - conda activate myenv

Preprocess data

Original dataset consists of .TextGrid files. Need to convert and combine them into one .txt file.
Install the required libraries, convert the files, and then combine the txt files with combine_txts.sh shell script.
Next, we need to remove --EMPTY-- utterances. Last divide the data in to 80:10:10 with balanced distributions for train:valid:test sets.
Change path_to_files accordingly.
punc_sym_to_punc.py removes the stamp of each utterance and convert <c/> and <s/> to , and . Run if needed. Make sure the texts are in order first.

First download the dataset from google drive: after downloading, move the folder under ./sgh_scripts_datasets/ directory

$ pip install -r ./preprocess/preprocess_dataset/requirements.txt
$ cd path/to/dataset_directory
$ python3 TextGrids_to_txts.py
$ . ./preprocess/preprocess_dataset/combine_txts.sh --path path/to/sgh_txt
$ python3 ./preprocess/preprocess_dataset/txtremoveEMPTY.py --data-path path/to/directory/with/all.txt
$ python3 ./train_test_val_split.py --data-path path/to/directory/with/cleaned_all.txt
$ python3 ./punc_sym_to_punc.py --data-path path/to/directory/with/sgh_train/test/val.txt
$ python3 ./create_pkl_dataset_new.py --data-path path/to/directory/with/cleaned_sgh_train/test/val.txt

Sample code

$ pip install -r ./preprocess_TextGrid_data/requirements.txt
$ cd sgh_dataset/
$ python3 ../preprocess_TextGrid_data/TextGrids_to_txts.py --data-path ./sgh_TextGrid
$ . ../preprocess_TextGrid_data/combine_txts.sh --path ./sgh_txt
$ python3 ../../preprocess_TextGrid_data/txtremoveEMPTY.py --data-path ../
$ python3 ../../preprocess_TextGrid_data/train_val_split.py --data-path ../
$ python3 ../../preprocess_TextGrid_data/punc_sym_to_punc.py --data-path ../
$ python3 ../../preprocess_TextGrid_data/create_pkl_dataset_new.py --data-path ../

Train and Test

Run the main.py file:

usage: main.py [-h] [--save-model] [--break-train-loop] [--stage STAGE]
               [--model-path MODEL_PATH] [--data-path DATA_PATH]
               [--num-epochs NUM_EPOCHS]
               [--log-level {INFO,DEBUG,WARNING,ERROR}]
               [--save-n-steps SAVE_N_STEPS] [--force-save]

arguments for the model

optional arguments:
  -h, --help            show this help message and exit
  --save-model          save model
  --break-train-loop    prevent training for debugging purposes
  --stage STAGE         load model from checkpoint stage
  --model-path MODEL_PATH
                        path to model directory
  --train-data-path DATA_PATH
                        path to dataset directory containing train_data.pkl
  --val-data-path DATA_PATH
                        path to dataset directory containing valid_data.pkl
  --eval-type EVAL_TYPE
  			valid/tst validation dataset or test dataset for evaluation
  --num-epochs NUM_EPOCHS
                        no. of epochs to run the model
  --log-level {INFO,DEBUG,WARNING,ERROR}
                        Logging info to be displayed
  --save-n-steps SAVE_N_STEPS
                        Save after n steps, default=1 epoch
  --force-save          Force save, overriding all settings
  --config CONFIG	Path to config directory
  --action ACTION 	train/val training or testing
			

Sample code

#use GPU
$ cd /new_Multilingual-Sentence-Boundary-detection/
#training
$ python3 ./src/main.py --num-epochs 10 --train-data-path /sgh/xlm-roberta-base/ --val-data-path sgh/xlm-roberta-base/ --model-path /punctuator-model/model_10epoch- --eval-type valid/tst
#testing
$ python3 ./src/main.py --train-data-path /sgh/xlm-roberta-base/ --val-data-path /sgh/xlm-roberta-base/ --action val --model-path /path/to/model/dir/ --stage model_10epoch-xlm-roberta-base-epoch-1.pth --eval-type valid/tst