Easy ReviewAIF-C01
Fundamentals of AI and ML9

Transcribe

Convert speech in audio and video to text.

Amazon Transcribe is an automatic speech recognition service that converts audio to text.

Speech to text

Amazon Transcribe turns speech into text. It works on live audio streams and on recorded files.

How it works

  1. Stream audio, or point Transcribe at a media file in Amazon S3.
  2. Transcribe returns the text, with options such as custom vocabulary.
  3. Feed the text to other services, such as Comprehend for analysis or Translate for other languages.

Exam cues

  • If the question says speech to text, transcripts or captions, think Amazon Transcribe.
  • If the question says individual words such as acronyms, think custom vocabulary; for the context of specialized speech, think custom language models. Custom vocabularies and custom language models

Remember

  • Real time (streaming) or batch (S3 files).

Trap: Transcribe is speech to text; Polly is the reverse.

From the AWS docs (7)
  • Amazon Transcribe is an automatic speech recognition service that uses machine learning to convert audio to text, standalone or inside applications.
  • Amazon Transcribe transcribes media in real time (streaming) or files stored in Amazon S3 (batch).
  • Amazon Transcribe Medical is automatic speech recognition for medical speech, such as physician-dictated notes, telemedicine appointments and physician-patient conversations.
  • Amazon Transcribe can identify the languages spoken in media without a language code, including all languages in multilingual media.
  • Custom vocabularies improve recognition of specific words such as brand names, acronyms and proper nouns, using hints such as pronunciations.
  • Custom language models are trained on domain-specific text so Transcribe learns the context in which words are used; they improve accuracy for specialized speech beyond what custom vocabularies do.
  • AWS HealthScribe is a HIPAA-eligible capability that combines speech recognition and generative AI to transcribe patient-clinician conversations and generate clinical notes for review.

Speakers, privacy and toxicity

Amazon Transcribe can identify individual speakers and help handle sensitive or toxic speech in transcripts.

How it works

  1. Use speaker partitioning to distinguish individual speakers in the audio.
  2. Use toxicity detection to classify speech using text and audio cues such as tone and pitch.
  3. Use personally identifiable information (PII) redaction to remove PII from the transcript. Choose a redacted transcript alone or both redacted and unredacted versions.

Exam cues

  • If a meeting transcript needs to show who spoke, think speaker partitioning, because it identifies individual speakers.
  • If a game needs to flag abusive voice chat, think toxicity detection, because it uses both audio and text cues.
  • If a call transcript must omit customer names or card numbers, think PII redaction, because it produces a redacted transcript.

Remember

  • Toxicity detection classifies speech into seven categories, including abuse, threats and profanity.
  • PII redaction can produce a redacted transcript alone or alongside an unredacted transcript.

Trap: Toxicity detection uses tone and pitch as well as text; it does not rely on text cues alone.

From the AWS docs (3)
  • Amazon Transcribe supports language customization, content filtering for privacy, multi-channel audio and identifying individual speakers.
  • Transcribe toxicity detection uses audio cues such as tone and pitch together with text cues to classify toxic speech into seven categories: sexual harassment, hate speech, threat, abuse, profanity, insult and graphic.
  • Amazon Transcribe can redact PII from transcripts, producing only a redacted transcript or both redacted and unredacted versions.

Flashcards

tap to flip
Answer

·

All 9 cards as a list

A call center wants written transcripts of recorded customer calls. Which AWS service fits?

Amazon Transcribe, which converts speech to text.

Can Amazon Transcribe produce captions for a live stream as it happens?

Yes. It supports real-time streaming transcription as well as batch files in Amazon S3.

Does Amazon Transcribe translate the transcript into another language?

No. It produces text in the spoken language; translation is Amazon Translate.

Which AWS service turns physician dictation and doctor-patient conversations into text?

Amazon Transcribe Medical, speech recognition built for medical vocabulary.

A company wants to convert recorded speech into text. Which AWS AI service fits?

Amazon Transcribe uses machine learning to convert audio to text.

A healthcare software vendor wants clinical notes drafted automatically from recorded doctor-patient visits. Which AWS capability fits?

AWS HealthScribe, which transcribes the conversation and generates clinical notes with generative AI, and is HIPAA eligible.

A meeting recording has three speakers. Which Amazon Transcribe feature labels who said what?

Speaker partitioning, which separates the speech of individual speakers.

An online game must flag abusive voice chat. Which AWS feature analyzes both how something is said and what is said?

Amazon Transcribe toxicity detection, which combines tone and pitch with the text to classify seven categories of toxic speech.

Call recordings must be transcribed without customer names or card numbers in the text. What does Transcribe offer?

PII redaction, which outputs a redacted transcript (optionally also an unredacted one).

You finished every topic of AIF-C01!

Your badge turned gold. Next step: book the exam on AWS. Once you pass, add yourself to Who passed.

Share

Share that you finished this certification

I finished every topic of AWS Certified AI Practitioner (AIF-C01) on Easy Review. Come study too!