https://github.com/batvoice-org/tf2-punctuator2

Tensorflow 2.0 implementation of automatic punctuation with RNN + attention
https://github.com/batvoice-org/tf2-punctuator2

Last synced: 28 days ago
JSON representation

Tensorflow 2.0 implementation of automatic punctuation with RNN + attention

• Host: GitHub
• URL: https://github.com/batvoice-org/tf2-punctuator2
• Owner: batvoice-org
• Created: 2019-10-24T14:50:00.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2022-12-08T06:46:03.000Z (about 2 years ago)
• Last Synced: 2024-08-04T00:11:00.303Z (4 months ago)
• Language: Python
• Size: 28.3 KB
• Stars: 8
• Watchers: 5
• Forks: 3
• Open Issues: 15
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## Tensorflow 2.0 implementation of RNN + attention-based automatic punctuation

Derived from [this project](https://github.com/ottokart/punctuator2) written in Theano.

At this stage this is a rough draft, tested only with a single type of "punctuation"

(actually, sentence boundaries). However, it is easily adapted to any number of punctuation markers.

Also added a little script to infer and visualize the attention weights for any sentence

fed to a trained model.

Hyperparameters and paths to data and checkpoints are written in a bash file to be sourced

before running the script, e.g.

```

source env.sh

python 01_train.python

```

See example_data to see how the data should be formatted.

Preprocessing scripts are not provided in this version.