引導式測驗 · 中等 等級

檔案管理員人工智慧和查找輔助測驗

了解人工智慧如何轉錄和描述檔案記錄,為什麼出處和來源圖像很重要,以及檔案管理員如何審查機器生成的文字。

問題 1 的 8

What should an archivist verify in a generated finding-aid draft?

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 archivist verify in a generated finding-aid draft?
    • That the description matches the actual records and preserves collection context
    • That the summary is more detailed than the records
    • That every uncertain name is guessed
    • That folder hierarchy is removed
  2. Why label machine-generated transcription?
    • Labels prove every word is correct
    • OCR output cannot be edited
    • Researchers can distinguish OCR from verified human text
    • Attribution removes the need for source images
  3. Which elements give archival description context about a collection?
    • A guarantee that every folder was digitized
    • Provenance, scope, arrangement, and access information
    • An interpretation of every historical event
    • A list of model parameters
  4. How can uncertain handwriting be represented responsibly?
    • Replace it with the most likely name without notation
    • Delete the page from the collection
    • Treat low confidence as proof the document is blank
    • Mark uncertainty and retain access to the original image
  5. What does semantic search return?
    • Candidate records that still need source inspection
    • Every relevant record in the archive with certainty
    • A verified historical conclusion
    • A replacement for catalog records
  6. Which transcription metric may fail to show archival usefulness?
    • The date of processing
    • The number of pages in a box
    • Character error alone without considering discovery and uncertainty
    • The model’s file size
  7. Why should a finding aid preserve folder relationships?
    • Folders are always identical
    • Hierarchy can convey provenance and context
    • Relationships prevent researchers from searching
    • AI cannot process text without folder names
  8. What records help trace AI-assisted description?
    • Only a final summary
    • Only the archivist’s initials
    • A confidence score without source text
    • Model version, processing date, attribution, and corrections