Automatic content redaction
Use automatic content redaction to automatically redact sensitive
personally identifiable information (PII) from your transcription results. It replaces
each identified instance of PII with a [PII] tag in the transcript. You can
use this feature to protect privacy and comply with local laws and regulations.
Automatic content redaction enables you to easily review and share transcripts to
improve the customer service experience, coach agents, and discover new business
opportunities while protecting sensitive personal information. You can use this feature
for source audio in US English (en-US) with batch API calls.
Personally Identifiable Information includes:
| PII entity | Definition |
|---|---|
|
Bank Account Number |
A number that uniquely identifies a bank account. |
|
Bank Routing Number |
A number that identifies the location of a bank account. |
|
Credit Card Number or Debit Card Number |
A value that uniquely defines a payment card issued by a bank. |
| Credit Card or Debit Card CVV Code | A 3-digit or 4-digit security code on each credit card. |
|
Credit Card or Debit Card Expiration Date |
The month and year a card expires. |
| Debit Card PIN or Credit Card PIN | A security code issued by a bank or credit union. This number is used for bank accounts and payment cards. |
| Email address | The unique identifier of an email box where messages are delivered. |
| US Mailing address | The U.S. mailing address of an individual. |
| Name | The first name and last name of a person. |
| US phone number | A 10-digit phone number within the United States. |
| Social Security Number | A 9-digit number or the last 4 digits of that number. Issued to U.S. citizens, permanent residents, and temporary residents with employment. |
For each transcription job with automatic content redaction enabled, you generate either:
-
Only the redacted transcript.
-
Both the redacted transcript and the unredacted transcript.
Both redacted and unredacted transcripts are stored in the same output S3 bucket. Amazon Transcribe stores them in either a bucket you specify or in the default S3 bucket managed by the service.
To enable content redaction, use the console or the API. In the console, you enable Automatic content redaction in the Content removal section.
To enable content redaction using the API, complete the request parameters of the
ContentRedaction object in the
StartTranscriptionJob operation. See the request syntax for the
StartTranscriptionJob action for more information. To see if content redaction has been enabled for a
particular transcription job, use GetTranscriptionJob. To see which jobs have content redaction
enabled, use ListTranscriptionJobs.
A redacted transcript has sensitive PII replaced with the [PII] tag,
shown in the first truncated JSON output on this page. Because Amazon Transcribe has
redacted this
transcript, the isRedacted field of this JSON output is true.
Each JSON output of a transcription job has a results section that contains
the transcription results. Every word, number, punctuation mark, or redaction has
a
confidence value.
The redaction feature is designed to identify and remove sensitive data. However, due to the predictive nature of machine learning, it may not identify and remove all instances of sensitive data in a transcript generated by Amazon Transcribe. We recommend you review any redacted output to ensure it meets your needs.
The redaction feature does not meet the requirements for de-identification under medical privacy laws, such as the U.S. Health Insurance Portability and Accountability Act of 1996 (HIPAA), so we recommend you continue to treat your transciption as protected health information even after redaction.
Transcription jobs using automatic content redaction generate two types of
confidence values. The Automatic Speech Recognition (ASR) confidence
indicates the items that have the type of pronunciation or
punctuation is a specific utterance. In the transcript output below,
the word Good has a confidence of 1.0. This
confidence value indicates that Amazon Transcribe is 100 percent confident that the
word uttered in
this transcript is Good. The confidence value for a
PII tag is the confidence that the speech it flagged for redaction is
truly PII. In the transcript output below, the confidence of
0.9999 indicates that Amazon Transcribe is 99.99 percent confident that the entity
it redacted in the transcript is PII.
When redaction is enabled, Amazon Transcribe can generate the following files, depending on which options you select:
-
A redacted text file. A redacted text file displays "[PII] " in place of the redacted text; see below for example excerpts from a redacted conversation.
-
A redacted audio file (in wav format). A redacted audio file has silence in place of redacted speech. These periods of silence are not included in any non-talk-time flags you create.
-
An unredated (original) file. This file is only generated if you choose to get both the redacted and the original transcripts.
Important The original unredacted file is the only place where the complete conversation is stored. If you delete it, there will be no record of the unredacted sensitive data.
The following is the redacted JSON output:
{ "jobName":"job id", "accountId":"account id", "isRedacted": true, "results": { "transcripts": [ { "transcript": "Good morning, everybody. My name is [PII], and today I feel like sharing a whole lot of personal information with you. Let's start with my Social Security number [PII]. My credit card number is [PII] and my C V V code is [PII]. I hope that Amazon Transcribe is doing a good job at redacting that personal information away. Let's check." } ], "items": [ { "start_time": "2.86", "end_time": "3.35", "alternatives": [ { "confidence": "1.0", "content": "Good" } ], "type": "pronunciation" },Items removed for brevity{ "start_time": "5.56", "end_time": "6.25", "alternatives": [ { "content": "[PII]", "redactions": [ { "confidence": "0.9999" } ] } ], "type": "pronunciation" },Items removed for brevity], }, "status": "COMPLETED" }
If you generate an additional unredacted transcript, its JSON output looks similar
to
the output of a transcription job with content redaction disabled. The only difference
is the additional isRedacted field being set to false.
The following is the unredacted JSON output:
{ "jobName":"job id", "accountId":"account id", "isRedacted": false, "results": { "transcripts": [ { "transcript": "Good morning, everybody. My name is Mike, and today I feel like sharing a whole lot of personal information with you. Let's start with my Social Security number 000000000. My credit card number 5555555555555555 is and my C V V code is 000. I hope that Amazon Transcribe is doing a good job at redacting that personal information away. Let's check." } ], "items": [ { "start_time": "2.86", "end_time": "3.35", "alternatives": [ { "confidence": "1.0", "content": "Good" } ], "type": "pronunciation" },Items removed for brevity{ "start_time": "5.56", "end_time": "6.25", "alternatives": [ { "confidence": "1.0", "content": "Mike" } ], "type": "pronunciation" },Items removed for brevity], }, "status": "COMPLETED" }
Automatic content redaction is unavailable in the AWS GovCloud (US-East) Region. For more information on the availability of Amazon Transcribe in AWS Regions, see Amazon Transcribe Endpoints and Quotas.
