Microsoft AI-102 (Designing and Implementing a Microsoft Azure AI Solution) Practice Test 39

Question 380

A customer mentions delivery and packaging in one review; the manager needs distinct scores by topic. What aggregation practice fits?

  1. Use confidence and a manual-review threshold for ambiguous sentiment cases
  2. Validate coverage and supplement detection with access controls and manual review where appropriate
  3. Run named entity recognition when typed organization identification is required
  4. Group opinion targets by normalized aspects before summarizing their sentiments

Explanation

Correct Answer

D:

Targets are extracted at the text level; cross-review category summaries require normalization to shared aspects. In the customer-experience feedback processor, confirm the fix by retesting the affected behavior after grouping opinion targets by normalized aspects before summarizing their sentiments.

Incorrect Answers

A:

Uncertain sentiment should not trigger an irreversible process when the decision has meaningful consequences in its own workflow, but that is a different correction from the present requirement to group opinion targets by normalized aspects before summarizing their sentiments. The reported issue would remain.

B:

Although useful elsewhere in the customer-experience feedback processor, pII models cover supported entity categories with imperfect recall; they are not a comprehensive security guarantee; the present case is different and calls for the team to group opinion targets by normalized aspects before summarizing their sentiments.

C:

That change addresses another failure mode: salient phrases are not equivalent to classified organization entities; it does not address this case, for which the relevant intervention is to group opinion targets by normalized aspects before summarizing their sentiments.

Question 381

Users submit messages in unknown languages and an app hard-codes English as the source. Which change is needed?

  1. Use language detection or supported automatic source-language identification
  2. Read the selected translated text from the response's translation results
  3. Provide approved parallel sentence pairs or dictionaries for supported custom translation
  4. Use supported asynchronous document translation and retrieve the translated file artifact

Explanation

Correct Answer

A:

Hard-coded source language can misinterpret multilingual input and degrade translation results. In the international customer-portal localization, confirm the fix by retesting the affected behavior after using language detection or supported automatic source-language identification.

Incorrect Answers

B:

Although useful elsewhere in the international customer-portal localization, translator responses can carry metadata and multiple targets; the client must select the intended output content; the present case is different and calls for the team to use language detection or supported automatic source-language identification.

C:

That change addresses another failure mode: custom translation learns domain terminology from curated bilingual evidence rather than ad hoc prompt phrasing; it does not address this case, for which the relevant intervention is to use language detection or supported automatic source-language identification.

D:

Document translation handles complete supported files, including layout preservation, rather than only raw string snippets in its own workflow, but that is a different correction from the present requirement to use language detection or supported automatic source-language identification. The reported issue would remain.

Question 382

A two-word phrase produces an uncertain language prediction. What should the system do?

  1. Inspect the target storage location and output naming for the document-translation operation
  2. Use translation for semantic language conversion, not character-script transliteration
  3. Use additional context or a supported fallback instead of trusting a low-confidence short-text result
  4. Specify the intended target languages and map each translation separately

Explanation

Correct Answer

C:

Short ambiguous samples often lack enough linguistic evidence for strong identification. In the international customer-portal localization, confirm the fix by retesting the affected behavior after using additional context or a supported fallback instead of trusting a low-confidence short-text result.

Incorrect Answers

A:

Although useful elsewhere in the international customer-portal localization, document translation writes output artifacts to the configured target; the source path is not necessarily overwritten; the present case is different and calls for the team to use additional context or a supported fallback instead of trusting a low-confidence short-text result.

B:

Changing scripts does not translate syntax or meaning between languages in its own workflow, but that is a different correction from the present requirement to use additional context or a supported fallback instead of trusting a low-confidence short-text result. The reported issue would remain.

D:

That change addresses another failure mode: translation APIs may return multiple target-language results; a single text field cannot be assumed to represent all versions; it does not address this case, for which the relevant intervention is to use additional context or a supported fallback instead of trusting a low-confidence short-text result.

Question 383

A support reply needs both German and Spanish versions. What should the request handle?

  1. Inspect the target storage location and output naming for the document-translation operation
  2. Filter and align bilingual training data to the intended terminology and subject domain
  3. Specify the intended target languages and map each translation separately
  4. Use additional context or a supported fallback instead of trusting a low-confidence short-text result

Explanation

Correct Answer

C:

Translation APIs may return multiple target-language results; a single text field cannot be assumed to represent all versions. In the international customer-portal localization, confirm the fix by retesting the affected behavior after specifying the intended target languages and map each translation separately.

Incorrect Answers

A:

Although useful elsewhere in the international customer-portal localization, document translation writes output artifacts to the configured target; the source path is not necessarily overwritten; the present case is different and calls for the team to specify the intended target languages and map each translation separately.

B:

That change addresses another failure mode: noisy or mismatched parallel data teaches contradictory translations and can degrade adaptation; it does not address this case, for which the relevant intervention is to specify the intended target languages and map each translation separately.

D:

Short ambiguous samples often lack enough linguistic evidence for strong identification in its own workflow, but that is a different correction from the present requirement to specify the intended target languages and map each translation separately. The reported issue would remain.

Question 384

Automated translation repeatedly mishandles a proprietary chemical term despite generic vocabulary being correct. What training data helps?

  1. Use language detection or supported automatic source-language identification
  2. Provide approved parallel sentence pairs or dictionaries for supported custom translation
  3. Use supported transliteration between the requested writing systems
  4. Use supported asynchronous document translation and retrieve the translated file artifact

Explanation

Correct Answer

B:

Custom translation learns domain terminology from curated bilingual evidence rather than ad hoc prompt phrasing. In the international customer-portal localization, confirm the fix by retesting the affected behavior after providing approved parallel sentence pairs or dictionaries for supported custom translation.

Incorrect Answers

A:

That change addresses another failure mode: hard-coded source language can misinterpret multilingual input and degrade translation results; it does not address this case, for which the relevant intervention is to provide approved parallel sentence pairs or dictionaries for supported custom translation.

C:

Although useful elsewhere in the international customer-portal localization, transliteration maps script representation; ordinary translation attempts to convey meaning, not just spelling; the present case is different and calls for the team to provide approved parallel sentence pairs or dictionaries for supported custom translation.

D:

Document translation handles complete supported files, including layout preservation, rather than only raw string snippets in its own workflow, but that is a different correction from the present requirement to provide approved parallel sentence pairs or dictionaries for supported custom translation. The reported issue would remain.

Question 385

An international customer-portal localization must translate uploaded documents while maintaining their document-oriented processing workflow. Reviewers will accept the implementation when the translated document remains an accessible file artifact. Which approach best fulfills that requirement?

  1. Use automatic language detection
  2. Use Azure Translator document translation
  3. Train a custom Translator model
  4. Use transliteration

Explanation

Correct Answer

B:

Document translation handles stored document files and can preserve supported document structure or formatting. For this international customer-portal localization, the intended evidence is that the translated document remains an accessible file artifact. It is not intended to recognize a live speaker's microphone audio.

Incorrect Answers

A:

Language detection predicts the source language from the text so the workflow can select downstream processing; it may be used to determine which supported language an incoming message is written in in a international customer-portal localization, but the current requirement is instead to translate uploaded documents while maintaining their document-oriented processing workflow. That separate capability does not prove that the translated document remains an accessible file artifact; Detection does not itself translate the content into another language, so selecting that alternative would not meet the stated acceptance check.

C:

Custom Translator uses suitable domain translation examples to adapt translation behavior for supported language pairs; it may be used to improve translation of specialized terminology using approved parallel or dictionary data in a international customer-portal localization, but the current requirement is instead to translate uploaded documents while maintaining their document-oriented processing workflow. That separate capability does not prove that the translated document remains an accessible file artifact; It does not correct microphone acoustic recognition errors upstream, so selecting that alternative would not meet the stated acceptance check.

D:

Transliteration changes the script representation of source terms, useful for names or pronunciation hints; it may be used to convert words from one writing system into another script without translating their meaning in a international customer-portal localization, but the current requirement is instead to translate uploaded documents while maintaining their document-oriented processing workflow. That separate capability does not prove that the translated document remains an accessible file artifact; It does not supply a semantic translation of an entire passage, so selecting that alternative would not meet the stated acceptance check.

Question 386

A training corpus mixes translations from unrelated industries and domain quality worsens. What dataset change is appropriate?

  1. Use translation for semantic language conversion, not character-script transliteration
  2. Specify the intended target languages and map each translation separately
  3. Use additional context or a supported fallback instead of trusting a low-confidence short-text result
  4. Filter and align bilingual training data to the intended terminology and subject domain

Explanation

Correct Answer

D:

Noisy or mismatched parallel data teaches contradictory translations and can degrade adaptation. In the international customer-portal localization, confirm the fix by retesting the affected behavior after filtering and align bilingual training data to the intended terminology and subject domain.

Incorrect Answers

A:

Changing scripts does not translate syntax or meaning between languages in its own workflow, but that is a different correction from the present requirement to filter and align bilingual training data to the intended terminology and subject domain. The reported issue would remain.

B:

That change addresses another failure mode: translation APIs may return multiple target-language results; a single text field cannot be assumed to represent all versions; it does not address this case, for which the relevant intervention is to filter and align bilingual training data to the intended terminology and subject domain.

C:

Although useful elsewhere in the international customer-portal localization, short ambiguous samples often lack enough linguistic evidence for strong identification; the present case is different and calls for the team to filter and align bilingual training data to the intended terminology and subject domain.

Question 387

The app assumes every detected language is supported by the chosen downstream translation model. What guard is missing?

  1. Review critical terminology using approved glossaries and bilingual validation
  2. Check downstream language support before dispatching detected-language text
  3. Use the published supported custom model or category identifier for translation
  4. Validate the rendered spelling against authoritative identity records

Explanation

Correct Answer

B:

Recognition of a language code and availability of a target translation pair are separate constraints. In the international customer-portal localization, confirm the fix by retesting the affected behavior after checking downstream language support before dispatching detected-language text.

Incorrect Answers

A:

Although useful elsewhere in the international customer-portal localization, machine translation is probabilistic, and high-impact text should not be released solely on automated fluency; the present case is different and calls for the team to check downstream language support before dispatching detected-language text.

C:

That change addresses another failure mode: training a customization is not sufficient when runtime calls continue targeting a generic configuration; it does not address this case, for which the relevant intervention is to check downstream language support before dispatching detected-language text.

D:

Transliteration can vary across conventions and should not be treated as verified legal spelling in its own workflow, but that is a different correction from the present requirement to check downstream language support before dispatching detected-language text. The reported issue would remain.

Question 388

A localization service must translate a 40-page Word file while retaining its layout. Which processing style is suitable?

  1. Use language detection or supported automatic source-language identification
  2. Provide approved parallel sentence pairs or dictionaries for supported custom translation
  3. Use supported asynchronous document translation and retrieve the translated file artifact
  4. Read the selected translated text from the response's translation results

Explanation

Correct Answer

C:

Document translation handles complete supported files, including layout preservation, rather than only raw string snippets. In the international customer-portal localization, confirm the fix by retesting the affected behavior after using supported asynchronous document translation and retrieve the translated file artifact.

Incorrect Answers

A:

That change addresses another failure mode: hard-coded source language can misinterpret multilingual input and degrade translation results; it does not address this case, for which the relevant intervention is to use supported asynchronous document translation and retrieve the translated file artifact.

B:

Although useful elsewhere in the international customer-portal localization, custom translation learns domain terminology from curated bilingual evidence rather than ad hoc prompt phrasing; the present case is different and calls for the team to use supported asynchronous document translation and retrieve the translated file artifact.

D:

Translator responses can carry metadata and multiple targets; the client must select the intended output content in its own workflow, but that is a different correction from the present requirement to use supported asynchronous document translation and retrieve the translated file artifact. The reported issue would remain.

Question 389

A document translation service has access to the input file but fails writing results. Which permission could be missing?

  1. Use the published supported custom model or category identifier for translation
  2. Grant the service authorized write access to the output storage destination
  3. Review critical terminology using approved glossaries and bilingual validation
  4. Check downstream language support before dispatching detected-language text

Explanation

Correct Answer

B:

Read permission on source content does not automatically authorize output-blob creation. In the international customer-portal localization, confirm the fix by retesting the affected behavior after granting the service authorized write access to the output storage destination.

Incorrect Answers

A:

Although useful elsewhere in the international customer-portal localization, training a customization is not sufficient when runtime calls continue targeting a generic configuration; the present case is different and calls for the team to grant the service authorized write access to the output storage destination.

C:

That change addresses another failure mode: machine translation is probabilistic, and high-impact text should not be released solely on automated fluency; it does not address this case, for which the relevant intervention is to grant the service authorized write access to the output storage destination.

D:

Recognition of a language code and availability of a target translation pair are separate constraints in its own workflow, but that is a different correction from the present requirement to grant the service authorized write access to the output storage destination. The reported issue would remain.

Question 390

A user name has several valid Latin spellings. How should a legal onboarding flow treat generated transliteration?

  1. Use the published supported custom model or category identifier for translation
  2. Grant the service authorized write access to the output storage destination
  3. Review critical terminology using approved glossaries and bilingual validation
  4. Validate the rendered spelling against authoritative identity records

Explanation

Correct Answer

D:

Transliteration can vary across conventions and should not be treated as verified legal spelling. In the international customer-portal localization, confirm the fix by retesting the affected behavior after validating the rendered spelling against authoritative identity records.

Incorrect Answers

A:

Although useful elsewhere in the international customer-portal localization, training a customization is not sufficient when runtime calls continue targeting a generic configuration; the present case is different and calls for the team to validate the rendered spelling against authoritative identity records.

B:

Read permission on source content does not automatically authorize output-blob creation in its own workflow, but that is a different correction from the present requirement to validate the rendered spelling against authoritative identity records. The reported issue would remain.

C:

That change addresses another failure mode: machine translation is probabilistic, and high-impact text should not be released solely on automated fluency; it does not address this case, for which the relevant intervention is to validate the rendered spelling against authoritative identity records.

Designing and Implementing a Microsoft Azure AI Solution (AI-102) Practice Tests