https://github.com/attilanagy234/neural-punctuator

Complimentary code for our paper Automatic punctuation restoration with BERT models
https://github.com/attilanagy234/neural-punctuator

bert punctuation-restoration transformer

Last synced: 2 months ago
JSON representation

Complimentary code for our paper Automatic punctuation restoration with BERT models

• Host: GitHub
• URL: https://github.com/attilanagy234/neural-punctuator
• Owner: attilanagy234
• License: mit
• Created: 2020-10-13T14:26:34.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2023-11-06T08:11:56.000Z (8 months ago)
• Last Synced: 2024-01-26T19:35:19.328Z (5 months ago)
• Topics: bert, punctuation-restoration, transformer
• Language: Jupyter Notebook
• Homepage:
• Size: 7.63 MB
• Stars: 46
• Watchers: 3
• Forks: 7
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-hungarian-nlp - neural-punctuator

README

        
# neural-punctuator

Complimentary code for our paper [**_Automatic punctuation restoration with BERT models_**](https://arxiv.org/abs/2101.07343) submitted to the XVII. Conference on Hungarian Computational Linguistics.

## Abstract

We present an approach for automatic punctuation restoration with BERT models for English and Hungarian. For English, we conduct our experiments on Ted Talks, a commonly used benchmark for punctuation restoration, while for Hungarian we evaluate our models on the Szeged Treebank dataset. Our best models achieve a macro-averaged F1-score of 79.8 in English and 82.2 in Hungarian.

## Repository Structure

```

.

|-- docs

|   └── paper       # The submitted paper

|-- notebooks       # Notebooks for data preparation/preprocessing

|-- src

    └── neural_punctuator 

        ├── base            # Base classes for training Torch models

        ├── configs         # YAMl files defining the parameters of each model

        ├── models          # Torch model definitions

        ├── preprocessors   # Preprocessor class

        ├── trainers        # Train logic

        ├── utils           # Utility scripts (logging, metrics, tensorboard etc.)

        └── wrappers        # Wrapper classes for the models containing all the components needed for training/prediction

```

## Dataset

Ted Talk dataset (English) - http://hltc.cs.ust.hk/iwslt/index.php/evaluation-campaign/ted-task.html

Szeged Treebank (Hungarian) - https://rgai.inf.u-szeged.hu/node/113

## Citation

If you use our work, please cite the following paper:

```

@article{nagy2021automatic,

  title={Automatic punctuation restoration with bert models},

  author={Nagy, Attila and Bial, Bence and {\'A}cs, Judit},

  journal={arXiv preprint arXiv:2101.07343},

  year={2021}

}

```

## Authors

Attila Nagy, Bence Bial, Judit Ács

Budapest University of Technology and Economics - Department of Automation and Applied Informatics