Home

Awesome

DEMO and DEMO2

Punctuator

A bidirectional recurrent neural network model with attention mechanism for restoring missing inter-word punctuation in unsegmented text.

The model can be trained in two stages (second stage is optional):

  1. First stage is trained on punctuation annotated text. Here the model learns to restore puncutation based on textual features only.
  2. Optional second stage can be trained on punctuation and pause annotated text. In this stage the model learns to combine pause durations with textual features and adapts to the target domain. If pauses are omitted then only adaptation is performed. Second stage with pause durations can be used for example for restoring punctuation in automatic speech recognition system output.

How well does it work?

Remember that all the scores given below are on unsegmented text and we did not use prosodic features, so, among other things, the model has to detect sentence boundaries in addition to the boundary type (?QUESTIONMARK, .PERIOD or !EXCLAMATIONMARK) based entirely on textual features. The scores are computed on the test set.

Training speed with default settings, an optimal Theano installation and a modern GPU should be around 10000 words per second.

Pretrained models can be downloaded here (Demo + 2 models from the Interspeech paper).

English TED talks

Training set size: 2.1M words. First stage only. More details can be found in this paper. For comparison, our previous model got an overall F1-score of 50.8.

PUNCTUATIONPRECISIONRECALLF-SCORE
,COMMA64.445.253.1
?QUESTIONMARK67.558.762.8
.PERIOD72.371.571.9
Overall68.958.163.1

English Europarl v7

Training set size: 40M words. First stage only. Details in ./example.

You can try to compete with this model here.

PUNCTUATIONPRECISIONRECALLF-SCORE
?QUESTIONMARK77.773.275.4
!EXCLAMATIONMARK50.00.10.1
,COMMA68.972.070.4
-DASH55.98.815.2
:COLON60.923.834.2
;SEMICOLON44.71.12.2
.PERIOD84.784.184.4
Overall75.773.974.8

Requirements

Requirements for data:

Make sure that first words of sentences don't have capitalized first letters. This would give the model unfair hints about period locations. Also, the text files you use for training and validation must be large enough (at least minibatch_size x sequence_length of words, which is 128x50=6400 words with default settings), otherwise you might get an error.

Configuration

Vocabulary size, punctuation tokens and their mappings, and converted data location can be configured in the header of data.py. Some model hyperparameters can be configured in the headings of main.py and main2.py. Learning rate and hidden layer size can be passed as arguments.

Usage

First step is data conversion. Assuming that preprocessed and cleaned *.train.txt, *.dev.txt and *.test.txt files are located in <data_dir>, the conversion can be initiated with:

python data.py <data_dir>

If you have second stage data as well, then:

python data.py <data_dir> <second_stage_data_dir>

The first stage can be trained with:

python main.py <model_name> <hidden_layer_size> <learning_rate>

e.g python main.py <model_name> 256 0.02 works well.

Second stage can be trained with:

python main2.py <model_name> <hidden_layer_size> <learning_rate> <first_stage_model_path>

Preprocessed text can be punctuated with e.g:

cat data.dev.txt | python punctuator.py <model_path> <model_output_path>

or, if pause annotations are present in data.dev.txt and you have a second stage model trained on pause annotated data, then:

cat data.dev.txt | python punctuator.py <model_path> <model_output_path> 1

Punctuation tokens in data.dev.txt don't have to be removed - the punctuator.py script ignores them.

Error statistics in this example can be computed with:

python error_calculator.py data.dev.txt <model_output_path>

You can play with a trained model with (assumes the input text is similarly preprocessed as the training data):

python play_with_model.py <model_path>

or with:

python play_with_model.py <model_path> 1

if you want to see, which words the model sees as UNKs (OOVs).

Citing

The software is described in:

@inproceedings{tilk2016,
  author    = {Ottokar Tilk and Tanel Alum{\"a}e},
  title     = {Bidirectional Recurrent Neural Network with Attention Mechanism for Punctuation Restoration},
  booktitle = {Interspeech 2016},
  year      = {2016}
}

We used the release v1.0 in the paper.

Alternatives