Prepstellar
Get the app

AIF-C01 · AI Use Cases and Services

21 cards

AWS Managed AI/ML Services

Swipe, scroll or use ← →

  1. One service builds models, the rest transform data

    This group of services splits cleanly in two. One of them hands you the machinery to make your own model; the others hand you a finished capability that turns one kind of data into another. Knowing which half a scenario is asking for settles most questions before the service names even matter.

    Amazon SageMaker AI provides fully managed infrastructure, tools, and workflows for building, training, and deploying ML models. It brings the components used for ML into one toolset so that models can reach production with less effort and lower cost.

    1 / 21

  2. One service builds models, the rest transform data

    SageMaker AI is the choice in this group when the requirement is the broader model lifecycle rather than a prebuilt language or speech transformation.

    That phrase is the discriminator. If a scenario talks about training on the company's own data, tuning, deploying an endpoint or managing the pipeline that produces a model, it belongs to SageMaker AI. If it describes a transformation that already exists off the shelf — audio to text, text to another language, text to speech — one of the other services is meant.

    The requirement is about… The answer is
    Building, training and deploying a model SageMaker AI
    A ready-made language or speech transformation One of the prebuilt services
    2 / 21

  3. Quick check

    Which capability identifies Amazon SageMaker AI within this group of services?

    1. AManaged infrastructure and workflows to build, train and deploy models

      Right. SageMaker AI covers the model lifecycle with managed infrastructure, tools and workflows.

    2. BAutomatic speech recognition for recorded and streaming audio

      Recognizing speech is a single prebuilt transformation offered by a different service, not the model lifecycle.

    3. CNeural machine translation for websites and large text volumes

      Machine translation is likewise one finished capability rather than the toolset for producing models.

    3 / 21

  4. Turn speech into text

    Amazon Transcribe is an automatic speech recognition service that converts speech to text. It can process audio files in common formats and attach timestamps to words.

    Those timestamps matter more than they look: they are what makes a transcript searchable and what lets a caption line up with the moment it belongs to.

    4 / 21

  5. Turn speech into text

    Transcribe is not limited to files that already exist. It can also accept a live audio stream and return transcripts in real time, which is what a contact center needs when an agent is still on the call.

    Typical applications include transcribing customer-service calls and generating subtitles for audio or video.

    Input What Transcribe returns
    An audio file in a common format Text, with timestamps on the words
    A live audio stream Transcripts in real time
    5 / 21

  6. Quick check

    A contact center must turn a live audio stream into a real-time text transcript. Which service fits?

    1. AAmazon Polly, which returns audio streams from written text

      Polly produces audio from text, which is the opposite of what a transcript requires.

    2. BAmazon Comprehend, which analyzes text that already exists

      Comprehend analyzes text that already exists and cannot take audio as its input.

    3. CAmazon Transcribe, which streams text from live audio

      Right. Transcribe accepts a live audio stream and returns transcripts in real time.

    6 / 21

  7. Change the language, keep the text

    Amazon Translate is a neural machine translation service for translating text between languages. It can localize websites and applications, translate large volumes of text for analysis, and support communication between users who use different languages.

    One detail decides every scenario about it: the input and output remain text; this distinguishes translation from speech recognition and speech synthesis. Nothing is spoken and nothing is listened to. Text goes in, text comes out, and only the language has changed.

    7 / 21

  8. Quick check

    An application must render its help center in several languages, working only with written content. Which service provides that?

    1. AAmazon Lex, for intent-driven conversations by voice or text

      Lex builds conversational interfaces and does not localize a body of existing written content.

    2. BAmazon Translate, for neural translation between languages

      Right. Translate performs neural machine translation and keeps the content in text form.

    3. CAmazon Transcribe, for turning spoken audio into text

      Transcribe needs audio input, and the help center content is already written.

    8 / 21

  9. Keep your progress in the app

    That’s 3 of 8 quick checks. In the app they stay answered, and every lesson remembers where you left off.

    Get the app Keep reading here

  10. Read what the text is saying

    Amazon Comprehend uses ML and natural language processing to uncover insights and relationships in unstructured text. It works on text you already have and returns what that text contains, without changing its language or turning it into sound.

    It can identify a text's language, extract key phrases and entities, determine positive or negative sentiment, analyze parts of speech, and organize text files by topic. Read that list once as a whole: it is the standard answer to "we have documents and we want to know what is in them".

    9 / 21

  11. Read what the text is saying

    Comprehend is not limited to what ships in the box. Its AutoML capabilities can also create custom entity recognition or text classification models for organization-specific needs — a product code, an internal category, a document type that only your company uses.

    That is worth separating from SageMaker AI. Comprehend's custom models stay inside a text-analysis service and target text-analysis outputs; SageMaker AI is the general toolset for any model lifecycle.

    10 / 21

  12. Quick check

    A company has written reviews and needs sentiment, key phrases and named entities, without translating or speaking the text. Which service fits?

    1. AAmazon Translate, which only changes the language of the text

      Translation changes the language but returns none of the requested insights about the content.

    2. BAmazon Polly, which reads the reviews aloud as an audio stream

      Polly turns the text into speech, which the requirement explicitly rules out.

    3. CAmazon Comprehend, which extracts entities and sentiment

      Right. Comprehend applies NLP to unstructured text and returns language, key phrases, entities, sentiment and topics.

    11 / 21

  13. Build the conversation

    Amazon Lex is a fully managed service for designing, building, testing, and deploying conversational interfaces that use voice or text.

    What makes it more than a transcriber is that it combines two capabilities. Lex combines automatic speech recognition, which converts speech to text, with natural language understanding, which recognizes intent. Hearing the words is only half the job; deciding what the caller is trying to achieve is the other half.

    12 / 21

  14. Build the conversation

    Lex also holds the thread of the exchange. It manages dialogue and dynamically adjusts responses, making it suitable for text or voice chatbots and interactive voice response systems.

    Requirement in a scenario Service
    "Produce a transcript of this audio" Transcribe
    "Understand what the user wants and reply in turn" Lex

    So a scenario that mentions intent, slots to fill, or a conversation that continues over several turns is a Lex scenario, whether the channel is voice or text.

    13 / 21

  15. Quick check

    A team needs a chatbot that accepts voice or text, recognizes user intent and manages the dialogue. Which service fits?

    1. AAmazon Lex, which recognizes intent and manages the dialogue

      Right. Lex combines speech recognition with natural language understanding and manages the conversation.

    2. BAmazon Transcribe, which only turns spoken audio into written text

      Transcribe produces a transcript and stops there; it neither recognizes intent nor holds a dialogue.

    3. CAmazon Polly, which only turns written text into spoken audio

      Polly speaks a supplied text and has no understanding of what the user is asking for.

    14 / 21

  16. Turn text into speech

    Amazon Polly turns text into lifelike speech. An application sends text to the Amazon Polly API and receives an audio stream.

    From there the application decides what to do with the sound: it can play the stream directly or store it in a standard audio format. Polly supports real-time interactive dialogue, and generated speech can be cached and replayed — useful when the same announcement is spoken thousands of times and should not be synthesized again each time.

    Polly and Transcribe are the mirror pair of this group, and scenarios test exactly that:

    Service In Out
    Transcribe Spoken audio Text
    Polly Text Spoken audio
    15 / 21

  17. Quick check

    An application must produce spoken audio from a written passage. Which service provides that?

    1. AAmazon Lex, because a chatbot is required here

      No conversation or intent recognition is described, so a chatbot service is not what the requirement asks for.

    2. BAmazon Polly, because it synthesizes speech from text

      Right. Polly returns an audio stream from the text that the application sends to its API.

    3. CAmazon Translate, because the passage changes language

      Translation would return text in another language and would still leave nothing to play.

    16 / 21

  18. Choose by the transformation

    There is one test that resolves this whole group, and it does not require remembering marketing lines. Ask what goes in and what must come out.

    Choose by the required transformation: SageMaker AI for the managed ML model lifecycle, Transcribe for speech to text, Translate for text between languages, Comprehend for insights from unstructured text, Lex for intent-driven voice or text conversations, and Polly for text to speech.

    17 / 21

  19. Choose by the transformation

    This input-to-output test prevents common mix-ups, and the mix-ups are predictable:

    • Transcribe and Polly reverse the direction of the speech transformation.
    • Translate changes language while retaining text.
    • Comprehend analyzes text rather than translating or speaking it.
    • Lex adds dialogue and intent recognition to a conversational interface.

    Whenever an option offers the right service with the wrong direction — or the wrong service with a description borrowed from the right one — this test is what catches it.

    18 / 21

  20. Quick check

    Which comparison correctly describes the direction of the two speech transformations?

    1. ATranscribe maps text to speech, while Polly maps speech to text

      The two services have been swapped: this describes Polly's job as Transcribe's and the other way round.

    2. BTranscribe maps speech to text, while Polly maps text into speech

      Right. Transcribe recognizes speech and returns text, while Polly synthesizes speech from text.

    3. CComprehend maps text to audio, while SageMaker AI maps audio to text

      Comprehend produces insights rather than audio, and SageMaker AI is a model toolset rather than a transcriber.

    19 / 21

  21. Key takeaways

    • SageMaker AI supports building, training, and deploying ML models with managed infrastructure and workflows.
    • Transcribe converts speech to text, while Polly converts text to speech — the two are mirror images and are tested that way.
    • Translate translates text between languages, while Comprehend extracts insights from unstructured text.
    • Lex builds voice or text conversational interfaces and recognizes user intent, adding dialogue management on top of speech recognition.
    • Match the service to the transformation, not to the industry in the scenario: the data going in and the result required decide it.
    20 / 21

  22. Quick check

    Which mapping of services to transformations is entirely correct?

    1. ATranscribe speech to text, Translate text to text, and Polly text to speech

      Right. Transcribe recognizes speech, Translate changes the language of text, and Polly synthesizes speech.

    2. BTranscribe text to speech, Comprehend text to audio, Lex text only

      Transcribe is reversed, Comprehend returns insights rather than audio, and Lex handles voice as well as text.

    3. CPolly speech to text, Translate audio to text, Comprehend audio to text

      Polly is reversed, and neither Translate nor Comprehend accepts audio as its input.

    21 / 21

  23. 8 quick checks · then the test

    In the app, finishing the quick checks opens this lesson’s 10-question test, and the ones you miss come back exactly when you’re about to forget them.

    Take this lesson’s test → Next lesson: Traditional ML vs Foundation ModelsGet the app
← Previous Keep your progress · Get the app Next card →

The whole course, on your phone

Lessons you can read, audio you can listen to on the way to work, and practice that remembers what you got wrong.

Get the app