create-language-model¶

Description¶

Creates a new custom language model.

When creating a new language model, you must specify:

If you want a Wideband (audio sample rates over 16,000 Hz) or Narrowband (audio sample rates under 16,000 Hz) base model
The location of your training and tuning files (this must be an Amazon S3 URI)
The language of your model
A unique name for your model

For more information, see Custom language models .

Synopsis¶

  create-language-model
--language-code <value>
--base-model-name <value>
--model-name <value>
--input-data-config <value>
[--tags <value>]
[--cli-input-json | --cli-input-yaml]
[--generate-cli-skeleton <value>]

Options¶

--language-code (string)

The language code that represents the language of your model. Each language model must contain terms in only one language, and the language you select for your model must match the language of your training and tuning data.

For a list of supported languages and their associated language codes, refer to the Supported languages table. Note that U.S. English (en-US ) is the only language supported with Amazon Transcribe Medical.

A custom language model can only be used to transcribe files in the same language as the model. For example, if you create a language model using US English (en-US ), you can only apply this model to files that contain English audio.

Possible values:

en-US

hi-IN

es-US

en-GB

en-AU

--base-model-name (string)

The Amazon Transcribe standard language model, or base model, used to create your custom language model. Amazon Transcribe offers two options for base models: Wideband and Narrowband.

If the audio you want to transcribe has a sample rate of 16,000 Hz or greater, choose WideBand . To transcribe audio with a sample rate less than 16,000 Hz, choose NarrowBand .

Possible values:

NarrowBand

WideBand

--model-name (string)

A unique name, chosen by you, for your custom language model.

This name is case sensitive, cannot contain spaces, and must be unique within an Amazon Web Services account. If you try to create a new language model with the same name as an existing language model, you get a ConflictException error.

--input-data-config (structure)

Contains the Amazon S3 location of the training data you want to use to create a new custom language model, and permissions to access this location.

When using InputDataConfig , you must include these sub-parameters: S3Uri , which is the Amazon S3 location of your training data, and DataAccessRoleArn , which is the Amazon Resource Name (ARN) of the role that has permission to access your specified Amazon S3 location. You can optionally include TuningDataS3Uri , which is the Amazon S3 location of your tuning data. If you specify different Amazon S3 locations for training and tuning data, the ARN you use must have permissions to access both locations.

S3Uri -> (string)

The Amazon S3 location (URI) of the text files you want to use to train your custom language model.

Here’s an example URI path: s3://DOC-EXAMPLE-BUCKET/my-model-training-data/

TuningDataS3Uri -> (string)

The Amazon S3 location (URI) of the text files you want to use to tune your custom language model.

Here’s an example URI path: s3://DOC-EXAMPLE-BUCKET/my-model-tuning-data/

DataAccessRoleArn -> (string)

The Amazon Resource Name (ARN) of an IAM role that has permissions to access the Amazon S3 bucket that contains your input files. If the role you specify doesn’t have the appropriate permissions to access the specified Amazon S3 location, your request fails.

IAM role ARNs have the format arn:partition:iam::account:role/role-name-with-path . For example: arn:aws:iam::111122223333:role/Admin .

For more information, see IAM ARNs .

Shorthand Syntax:

S3Uri=string,TuningDataS3Uri=string,DataAccessRoleArn=string

JSON Syntax:

{
  "S3Uri": "string",
  "TuningDataS3Uri": "string",
  "DataAccessRoleArn": "string"
}

--tags (list)

Adds one or more custom tags, each in the form of a key:value pair, to a new custom language model at the time you create this new model.

To learn more about using tags with Amazon Transcribe, refer to Tagging resources .

(structure)

Adds metadata, in the form of a key:value pair, to the specified resource.

For example, you could add the tag Department:Sales to a resource to indicate that it pertains to your organization’s sales department. You can also use tags for tag-based access control.

To learn more about tagging, see Tagging resources .

Key -> (string)

The first part of a key:value pair that forms a tag associated with a given resource. For example, in the tag Department:Sales , the key is ‘Department’.

Value -> (string)

The second part of a key:value pair that forms a tag associated with a given resource. For example, in the tag Department:Sales , the value is ‘Sales’.

Note that you can set the value of a tag to an empty string, but you can’t set the value of a tag to null. Omitting the tag value is the same as using an empty string.

Shorthand Syntax:

Key=string,Value=string ...

JSON Syntax:

[
  {
    "Key": "string",
    "Value": "string"
  }
  ...
]

--cli-input-json | --cli-input-yaml (string) Reads arguments from the JSON string provided. The JSON string follows the format provided by --generate-cli-skeleton. If other arguments are provided on the command line, those values will override the JSON-provided values. It is not possible to pass arbitrary binary values using a JSON-provided value as the string will be taken literally. This may not be specified along with --cli-input-yaml.

--generate-cli-skeleton (string) Prints a JSON skeleton to standard output without sending an API request. If provided with no value or the value input, prints a sample input JSON that can be used as an argument for --cli-input-json. Similarly, if provided yaml-input it will print a sample input YAML that can be used with --cli-input-yaml. If provided with the value output, it validates the command inputs and returns a sample output JSON for that command.

See ‘aws help’ for descriptions of global parameters.

Examples¶

Example 1: To create a custom language model using both training and tuning data.

The following create-language-model example creates a custom language model. You can use a custom language model to improve transcription performance for domains such as legal, hospitality, finance, and insurance. For language-code, enter a valid language code. For base-model-name, specify a base model that is best suited for the sample rate of the audio that you want to transcribe with your custom language model. For model-name, specify the name that you want to call the custom language model.

aws transcribe create-language-model \
    --language-code language-code \
    --base-model-name base-model-name \
    --model-name cli-clm-example \
    --input-data-config S3Uri="s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix-for-training-data",TuningDataS3Uri="s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix-for-tuning-data",DataAccessRoleArn="arn:aws:iam::AWS-account-number:role/IAM-role-with-permissions-to-create-a-custom-language-model"

Output:

{
    "LanguageCode": "language-code",
    "BaseModelName": "base-model-name",
    "ModelName": "cli-clm-example",
    "InputDataConfig": {
        "S3Uri": "s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix/",
        "TuningDataS3Uri": "s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix/",
        "DataAccessRoleArn": "arn:aws:iam::AWS-account-number:role/IAM-role-with-permissions-create-a-custom-language-model"
    },
    "ModelStatus": "IN_PROGRESS"
}

For more information, see Improving Domain-Specific Transcription Accuracy with Custom Language Models in the Amazon Transcribe Developer Guide.

Example 2: To create a custom language model using only training data.

The following create-language-model example transcribes your audio file. You can use a custom language model to improve transcription performance for domains such as legal, hospitality, finance, and insurance. For language-code, enter a valid language code. For base-model-name, specify a base model that is best suited for the sample rate of the audio that you want to transcribe with your custom language model. For model-name, specify the name that you want to call the custom language model.

aws transcribe create-language-model \
    --language-code en-US \
    --base-model-name base-model-name \
    --model-name cli-clm-example \
    --input-data-config S3Uri="s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix-For-Training-Data",DataAccessRoleArn="arn:aws:iam::AWS-account-number:role/IAM-role-with-permissions-to-create-a-custom-language-model"

Output:

{
    "LanguageCode": "en-US",
    "BaseModelName": "base-model-name",
    "ModelName": "cli-clm-example",
    "InputDataConfig": {
        "S3Uri": "s3://DOC-EXAMPLE-BUCKET/Amazon-S3-Prefix-For-Training-Data/",
        "DataAccessRoleArn": "arn:aws:iam::your-AWS-account-number:role/IAM-role-with-permissions-to-create-a-custom-language-model"
    },
    "ModelStatus": "IN_PROGRESS"
}

For more information, see Improving Domain-Specific Transcription Accuracy with Custom Language Models in the Amazon Transcribe Developer Guide.

Output¶

LanguageCode -> (string)

The language code you selected for your custom language model.

BaseModelName -> (string)

The Amazon Transcribe standard language model, or base model, you specified when creating your custom language model.

ModelName -> (string)

The name of your custom language model.

InputDataConfig -> (structure)

Lists your data access role ARN (Amazon Resource Name) and the Amazon S3 locations you provided for your training (S3Uri ) and tuning (TuningDataS3Uri ) data.

S3Uri -> (string)

The Amazon S3 location (URI) of the text files you want to use to train your custom language model.

Here’s an example URI path: s3://DOC-EXAMPLE-BUCKET/my-model-training-data/

TuningDataS3Uri -> (string)

The Amazon S3 location (URI) of the text files you want to use to tune your custom language model.

Here’s an example URI path: s3://DOC-EXAMPLE-BUCKET/my-model-tuning-data/

DataAccessRoleArn -> (string)

The Amazon Resource Name (ARN) of an IAM role that has permissions to access the Amazon S3 bucket that contains your input files. If the role you specify doesn’t have the appropriate permissions to access the specified Amazon S3 location, your request fails.

IAM role ARNs have the format arn:partition:iam::account:role/role-name-with-path . For example: arn:aws:iam::111122223333:role/Admin .

For more information, see IAM ARNs .

ModelStatus -> (string)

The status of your custom language model. When the status displays as COMPLETED , your model is ready to use.

Table of Contents

Feedback

User Guide