Analyze and Translate Text AI-102 Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A data processing pipeline handles text that may contain personal data. The pipeline must extract main topics and identify entities, but must also ensure that any PII is redacted before the text is stored. The pipeline processes a high volume of text and must minimize the number of API calls. You need to design the pipeline. What should you do?

   • ARun PII Detection to redact PII, then run Key Phrase Extraction and Entity Recognition on the redacted text.
   • BRun Key Phrase Extraction and Entity Recognition on the original text, then run PII Detection to redact PII.
   • CRun Key Phrase Extraction on the original text, then run PII Detection to redact PII, and then run Entity Recognition on the redacted text.
   • DRun Entity Recognition to redact PII, then run Key Phrase Extraction on the redacted text.
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A healthcare organization processes patient feedback forms that contain free-text comments. The organization must redact any personally identifiable information (PII) before storing the comments for analysis. The analysis must also determine whether each comment is positive or negative and identify the medical conditions mentioned. You need to configure the processing pipeline. What should you do?

   • AUse Sentiment Analysis first to determine tone, then use PII Detection to redact PII, and then use Language Detection to identify medical conditions.
   • BUse Key Phrase Extraction to identify topics, then use Entity Recognition to redact PII, and then use Sentiment Analysis on the original text.
   • CUse Entity Recognition to find medical conditions, then use PII Detection to redact the entire text, and then use Key Phrase Extraction to determine sentiment.
   • DUse PII Detection to redact PII, then use Sentiment Analysis on the redacted text, and Entity Recognition to find medical conditions.
   Reveal Solution Discussion Select an answer first
4. 13expert · hard

   A legal document processing system must redact PII from contracts, identify the organizations involved, and determine the sentiment of each clause. The compliance team requires that no PII be stored in logs. The team is considering two pipelines: (A) redact PII first, then run entity recognition and sentiment analysis on the redacted text; (B) run entity recognition and sentiment analysis on the original text, then redact PII from the results. Which pipeline should you choose and why?

   • APipeline A, because redacting PII first ensures no PII is stored, and entity recognition and sentiment analysis work on redacted text.
   • BPipeline B, because entity recognition and sentiment analysis are more accurate on the original text, and redacting the results is sufficient.
   • CPipeline B, because sentiment analysis requires the original text to determine tone, and redaction would change the sentiment.
   • DPipeline A, but only if the contract is in English, because redaction may affect entity recognition in other languages.
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A multinational legal firm processes contracts in multiple languages. They need to identify the parties involved (people, organizations), redact any personal data, and translate the contracts to English for review. The compliance team requires that no PII be stored. The team is considering two pipelines: (A) redact PII first, then run entity recognition and translation on the redacted text; (B) run entity recognition and translation on the original text, then redact PII from the results. Which pipeline should you choose and why?

   • APipeline B, because entity recognition and translation are more accurate on the original text, and redacting the results is sufficient.
   • BPipeline A, but only if the contract is in English, because translation may not work on redacted text.
   • CPipeline A, because redacting PII first ensures no PII is stored, and entity recognition and translation work on redacted text.
   • DPipeline B, because entity recognition requires the original text to identify parties, and redaction would remove the entities.
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A global customer feedback system receives messages in multiple languages. The team wants to translate each message to English for analysis, but also wants to preserve the original sentiment for escalation. They are considering two approaches: (A) translate first, then run sentiment analysis on the translated text; (B) run sentiment analysis on the original text, then translate. Which approach should you choose and why?

   • AApproach B, because sentiment analysis on the original text preserves the true sentiment, and translation is only needed for English output.
   • BApproach A, because the Translator API returns the sentiment score along with the translation.
   • CApproach B, but only if the original language is supported by sentiment analysis; otherwise, translate first.
   • DApproach A, because sentiment analysis on translated text is more accurate because the model is optimized for English.
   Reveal Solution Discussion Select an answer first