引導式測驗 · 硬 等級

利用 NLP 進行負面新聞篩選測驗

使用 NLP 來定位不良媒體並確定其優先順序,同時在決策前驗證身分、來源品質、相關性和人工審核。

問題 1 的 8

What should an NLP adverse-media alert represent?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. What should an NLP adverse-media alert represent?
    • A lead for a reviewer to investigate
    • Proof the customer committed a crime
    • An automatic reason to close an account
    • A confirmed sanctions designation
  2. Why combine name matching with additional identifiers?
    • To make every article legally authoritative
    • To reduce false matches between people or organizations with similar names
    • To remove the need for source review
    • To ensure every alert is current
  3. What should a reviewer consider besides whether an article contains a customer name?
    • 僅記錄模型的置信分數
    • The number of words in the article
    • Source reliability, context, recency and materiality
    • Whether the article uses a negative adjective
  4. How does adverse-media screening differ from sanctions screening?
    • It is always a government-issued list
    • It proves whether a person is politically exposed
    • It is only a way to detect spelling errors
    • It examines public reporting for risk context rather than matching an official sanctions list
  5. What does a classifier score establish by itself?
    • A prioritization signal under the model’s setup
    • That an allegation is true
    • That the right person was matched
    • That the source is trustworthy
  6. Why retain the article URL, date and retrieval time?
    • To replace the original investigation
    • To preserve provenance and let a reviewer inspect the evidence
    • To guarantee the publisher will not edit it
    • To turn every article into a list match
  7. How should an institution handle an allegation in an article?
    • Treat it as a proven fact
    • Ignore it in every case
    • Assess its credibility and relevance through the established review process
    • Convert it directly into a sanctions match
  8. Which evaluation check can reveal unequal screening coverage?
    • Measure only the average score
    • Count alerts without checking results
    • Use the same English source for every market
    • Test coverage and errors across relevant languages and scripts