Home Previous - The alphabet.txt file Next - Acoustic Model and Language Model

Scorer - language model for determining which words occur together


What is a scorer?

A scorer is a language model and it is used by DeepSpeech to improve the accuracy of transcription. A language model predicts which words are more likely to follow each other. For example, the word chicken might be frequently followed by the words nuggets, soup or rissoles, but is unlikely to be followed by the word purple. The scorer identifies probabilities of words occurring together.

The default scorer used by DeepSpeech is trained on the LibriSpeech dataset. The LibriSpeech dataset is based on LibriVox - an open collection of out-of-copyright and public domain works.

You may need to build your own scorer - your own language model if:

DeepSpeech supports the optional use of an external scorer - if you’re not sure if you need to build your own scorer, stick with the built-in one to begin with.

Building your own scorer

This section assumes that you are using a Docker image and container for training, as outlined in the environment section. If you are not using the Docker image, then some of the scripts such as will not be available in your environment.

This section assumes that you have already trained a model and have a set of checkpoints for that model. See the section on training for more information on checkpoints.

DeepSpeech uses an algorithm called connectionist temporal classification or CTC for short, to map between input sequences of audio and output sequences of characters. The mapping between inputs and outputs is called an alignment. The alignment between inputs and outputs is not one-to-one; many inputs may make up an output. CTC is therefore a probabilistic algorithm. This means that for each input there are many possible outputs that can be selected. A process call beam search is used to identify the possible outputs and select the one with the highest probability. A language model or scorer helps the beam search algorithm select the most optimal output value. This is why building your own scorer is necessary for training a model on a narrow domain - otherwise the beam search algorithm would probably select the wrong output.

The default scorer used with DeepSpeech is trained on Librivox. It’s a general model. But let’s say that you want to train a speech recognition model for agriculture. If you have the phrase tomatoes are ..., a general scorer might identify red as the most likely next word - but an agricultural model might identify ready as the most likely next word.

The scorer is only used during the test stage of training (rather than at the train or validate stages) because this is where the beam search decdoder determines which words are formed from the identified characters.

The process for building your own scorer has the following steps:

  1. Having, or preparing, a text file (in .txt or .txt.gz format), with one phrase or word on each line. If you are training a speech recognition model for a particular domain - such as technical words, medical transcription, agricultural terms etc, then they should appear in the text file. The text file is used by the script.

  2. Using the with your dataset (your .csv files) and a set of checkpoints to find optimal values of --default_alpha and --default_beta. The --default_alpha and --default_beta parameters are used by the generate_scorer_package script to assign initial weights to sequences of words.

  3. Using the script which is distributed with DeepSpeech, along with the text file, to create two files, called lm.binary and vocab-500000.txt.

  4. Downloading the prebuilt native_client from the DeepSpeech repository on GitHub, and using the generate_scorer_package to create a kenlm.scorer file.

  5. Using the kenlm.scorer file as the external_scorer passed to, and used for the test phase. The scorer does not impact training; it is used for calculating word error rate (covered more in testing).

In the following example we will create a custom external scorer file for Bahasa Indonesia (BCP47: id-ID).

Preparing the text file

This is straightforward. In this example, we will use a file called indonesian-sentences.txt. This file should contain phrases that you wish to prioritize recognising. For example, you may want to recognise place names, digits or medical phrases - and you will include these phrases in the .txt file.

These phrases should not be copied from test.tsv, train.tsv or validated.tsv as you will bias the resultant model.

~/deepspeech-data$ ls cv-corpus-6.1-2020-12-11/id
total 6288
   4 drwxr-xr-x 3 root root    4096 Feb 24 19:01 ./
   4 drwxr-xr-x 4 root root    4096 Feb 11 07:09 ../
1600 drwxr-xr-x 2 root root 1638400 Feb  9 10:43 clips/
 396 -rwxr-xr-x 1 root root  401601 Feb  9 10:43 dev.tsv
 104 -rwxr-xr-x 1 root root  103332 Feb  9 10:43 invalidated.tsv
1448 -rwxr-xr-x 1 root root 1481571 Feb  9 10:43 other.tsv
  28 -rwxr-xr-x 1 root root   26394 Feb  9 10:43 reported.tsv
 392 -rwxr-xr-x 1 root root  399790 Feb  9 10:43 test.tsv
 456 -rwxr-xr-x 1 root root  465258 Feb  9 10:43 train.tsv
1848 -rwxr-xr-x 1 root root 1889606 Feb  9 10:43 validated.tsv

The indonesian-sentences.txt file is stored on the local filesystem in the deepspeech-data directory so that the Docker container can access it.

~/deepspeech-data$ ls | grep indonesian-sentences
 476 -rw-rw-r--  1 root root  483481 Feb 24 19:02 indonesian-sentences.txt

The indonesian-sentences.txt file is formatted with one phrase per line, eg:

Kamar adik laki-laki saya lebih sempit daripada kamar saya.
Ayah akan membunuhku.
Ini pulpen.
Akira pandai bermain tenis.
Dia keluar dari ruangan tanpa mengatakan sepatah kata pun.
Besok Anda akan bertemu dengan siapa.
Aku mengerti maksudmu.
Tolong lepas jasmu.

Using to generate values for the parameters --default_alpha and --default_beta that are used by the generate_scorer_package script

The script is located in the DeepSpeech directory if you have set up your [environment][] as outlined in the PlayBook.

root@57e6bf4eeb1c:/DeepSpeech# ls | grep

This script takes a set of test data (--test_files), and a --checkpoint_dir parameter and determines the optimal --default_alpha and --default_beta values.

Call and pass it the --test_files and a --checkpoint_dir directory.

root@57e6bf4eeb1c:/DeepSpeech# python3 \
     --test_files deepspeech-data/cv-corpus-6.1-2020-12-11/id/clips/test.csv \
     --checkpoint_dir deepspeech-data/checkpoints

In general, any change to geometry - the shape of the neural network - needs to be reflected here, otherwise the checkpoint will fail to load. It’s always a good idea to record the parameters you used to train a model. For example, if you trained your model with a --n_hidden value that is different to the default (1024), you should pass the same --n_hidden value to, i.e:

root@57e6bf4eeb1c:/DeepSpeech# python3 \
     --test_files deepspeech-data/cv-corpus-6.1-2020-12-11/id/clips/test.csv \
     --checkpoint_dir deepspeech-data/checkpoints \
     --n_hidden 4 will create a new study.

[I 2021-03-05 02:04:23,041] A new study created in memory with name: no-name-38c8e8cb-0cc2-4f53-af0e-7a7bd3bc5159

It will then run testing and output a trial score.

[I 2021-03-02 12:48:15,336] Trial 0 finished with value: 1.0 and parameters: {'lm_alpha': 1.0381777700987271, 'lm_beta': 0.02094605391055826}. Best is trial 0 with value: 1.0.

By default, will run 6 trials, and identify the trial with the most optimal parameters.

[I 2021-03-02 17:50:00,662] Trial 6 finished with value: 1.0 and parameters: {'lm_alpha': 3.1660260368070423, 'lm_beta': 4.7438794403688735}. Best is trial 0 with value: 1.0.

The optimal parameters --default_alpha and --default_beta are now known, and can be used with generate_scorer_package. In this case, the optimal settings are:

--default_alpha 1.0381777700987271
--default_beta 0.02094605391055826

because Trial 0 was the best trial.

Additional parameters for

There are additional parameters that may be useful.

Please be aware that these parameters may increase processing time significantly - even to a few days - depending on your hardware.

For example:

root@57e6bf4eeb1c:/DeepSpeech# python3 \
     --test_files deepspeech-data/cv-corpus-6.1-2020-12-11/id/clips/test.csv \
     --checkpoint_dir deepspeech-data/checkpoints \
     --n_hidden 4 \
     --n_trials 3 \
     --lm_alpha_max 0.92 \
     --lm_beta_max 1.05

Using to create lm.binary and vocab-500000.txt files

We then use script that comes with DeepSpeech to create a trie file. The trie file represents associations between words, so that during training, words that are more closely associated together are more likely to be transcribed by DeepSpeech.

The trie file is produced using a software package called KenLM. KenLM is designed to create large language models that are able to be filtered and queried easily.

First, create a directory in deepspeech-data directory to store your lm.binary and vocab-500000.txt files:

deepspeech-data$ mkdir indonesian-scorer

Then, use the script as follows:

cd data/lm
python3 \
  --input_txt /DeepSpeech/deepspeech-data/indonesian-sentences.txt \
  --output_dir /DeepSpeech/deepspeech-data/indonesian-scorer \
  --top_k 500000 --kenlm_bins /DeepSpeech/native_client/kenlm/build/bin/ \
  --arpa_order 5 --max_arpa_memory "85%" --arpa_prune "0|0|1" \
  --binary_a_bits 255 --binary_q_bits 8 --binary_type trie

Note: the /DeepSpeech/native_client/kenlm/build/bin/ is the path to the binary files for kenlm. If you are using the Docker image and container (explained on the environment page of the PlayBook), then /DeepSpeech/native_client/kenlm/build/bin/ is the correct path to use. If you are not using the Docker environment, your path may vary.

You should now have a lm.binary and vocab-500000.txt file in your indonesian-scorer directory:

deepspeech-data$ ls indonesian-scorer/
total 1184
  4 drwxrwxr-x 2 root   root   4096 Feb 25 23:13 ./
  4 drwxrwxr-x 5 root   root   4096 Feb 26 09:24 ../
488 -rw-r--r-- 1 root   root      499594 Feb 24 19:05 lm.binary
 52 -rw-r--r-- 1 root   root       51178 Feb 24 19:05 vocab-500000.txt

Generating a kenlm.scorer file from generate_scorer_package

Next, we need to install the native_client package, which contains the generate_scorer_package. This is not pre-built into the DeepSpeech Docker image.

The generate_scorer_package, once installed via the native client package, is usable on all platforms supported by DeepSpeech. This is so that developers can generate scorers on-device, such as on an Android device, or Raspberry Pi 3.

To install generate_scorer_package, first download the relevant native client package from the DeepSpeech GitHub releases page into the data/lm directory. The Docker image uses Ubuntu Linux, so you should use either the native_client.amd64.cuda.linux.tar.xz package if you are using cuda or the native_client.amd64.cpu.linux.tar.xz package if not.

The easiest way to download the package and extract it is using curl [URL] | tar -Jxvf [FILENAME]:

root@dcb62aada58b:/DeepSpeech/data/lm# curl -L -o native_client.amd64.cuda.linux.tar.xz && tar -Jxvf native_client.amd64.cuda.linux.tar.xz

_NOTE: curl -L is used here to follow any 302 redirect headers that are served.

You can now generate a ken.lm scorer file.

root@dcb62aada58b:/DeepSpeech/data/lm# ./generate_scorer_package \
  --alphabet ../alphabet.txt  \
  --lm ../../deepspeech-data/indonesian-scorer/lm.binary
  --vocab ../../deepspeech-data/indonesian-scorer/vocab-500000.txt \
  --package kenlm-indonesian.scorer \
  --default_alpha 0.931289039105002 \
  --default_beta 1.1834137581510284
6021 unique words read from vocabulary file.
Doesn't look like a character based (Bytes Are All You Need) model.
--force_bytes_output_mode was not specified, using value infered from vocabulary contents: false
Package created in kenlm-indonesian.scorer.

The message Doesn't look like a character based (Bytes Are All You Need) model. is not an error.

If you receive the error message:

--force_bytes_output_mode was not specified, using value infered from vocabulary contents: false
Error: Can’t parse scorer file, invalid header. Try updating your scorer file.
Error loading language model file: Invalid magic in trie header.

then you should add the parameter --force_bytes_output_mode when calling generate_scorer_package. This error most usually occurs when training languages that use alphabets that contain a large number of characters, such as Mandarin. --force_bytes_output_mode forces the decoder to predict UTF-8 bytes instead of characters. For more information, please see the DeepSpeech documentation. For example:

root@dcb62aada58b:/DeepSpeech/data/lm# ./generate_scorer_package \
  --alphabet ../alphabet.txt  \
  --lm ../../deepspeech-data/indonesian-scorer/lm.binary
  --vocab ../../deepspeech-data/indonesian-scorer/vocab-500000.txt \
  --package kenlm-indonesian.scorer \
  --default_alpha 0.931289039105002 \
  --default_beta 1.1834137581510284 \
  --force_bytes_output_mode True

The kenlm-indonesian.scorer file is stored in the /DeepSpeech/data/lm directory within the Docker container. Copy it to the deepspeech-data directory.

root@dcb62aada58b:/DeepSpeech/data/lm# cp kenlm-indonesian.scorer ../../deepspeech-data/indonesian-scorer/
root@dcb62aada58b:/DeepSpeech/deepspeech-data/indonesian-scorer# ls -las
total 1820
  4 drwxrwxr-x 2 1000 1000   4096 Feb 26 21:56 .
  4 drwxrwxr-x 5 1000 1000   4096 Feb 25 22:24 ..
636 -rw-r--r-- 1 root root 648000 Feb 26 21:56 kenlm-indonesian.scorer
488 -rw-r--r-- 1 root root 499594 Feb 24 08:05 lm.binary
 52 -rw-r--r-- 1 root root  51178 Feb 24 08:05 vocab-500000.txt

Using the scorer file during the test phase of training

You now have your own scorer file that can be used during the test phase of model training process using the --scorer parameter.

For example:

python3 \
  --test_files deepspeech-data/cv-corpus-6.1-2020-12-11/id/clips/test.csv \
  --checkpoint_dir deepspeech-data/checkpoints-newscorer-id \
  --export_dir deepspeech-data/exported-model-newscorer-id \
  --n_hidden 2048 \
  --scorer deepspeech-data/indonesian-scorer/kenlm.scorer

For more information on scorer files, refer to the DeepSpeech documentation.

Home Previous - The alphabet.txt file Next - Acoustic Model and Language Model