引導式測驗 · 中等 等級

即時通話中 AI 客戶情緒偵測測驗

測試情緒評分不確定性、語音管道錯誤、文化差異、人工審核、隱私、代理追索權和客戶結果。

問題 1 的 8

A dashboard labels a caller “angry.” What can the agent infer from that score?

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. A dashboard labels a caller “angry.” What can the agent infer from that score?
    • That the caller intends to cancel
    • That the customer is dissatisfied with every service
    • That the agent caused the issue
    • Only that the system detected cues worth checking, not the caller’s true emotional state
  2. Why can a transcription error change a sentiment alert?
    • A missed word such as negation can change the meaning analyzed by later stages
    • Sentiment systems do not use text
    • Transcripts always remove punctuation only
    • Negation has no meaning in a call
  3. What did the cited cross-cultural voice study find?
    • Voice cues were perfectly universal
    • Emotion judgments varied across groups and accuracy declined with greater language dissimilarity from the stimulus voices
    • Every language produced identical scores
    • The study evaluated hotel booking behavior
  4. What should an agent do when the score conflicts with the call?
    • Follow the score without review
    • End the call to prevent another score
    • Listen to the source segment and ask a clarifying question if useful
    • Change the customer’s account status automatically
  5. Why should an employer avoid using a momentary score as an automatic performance penalty?
    • A single score always measures effort
    • Agents cannot challenge recorded calls
    • Sentiment is a certified performance measure
    • The score may reflect noise, language, or speaking style rather than agent conduct
  6. What monitoring should precede using alerts across language groups?
    • Check error patterns, calibration, and sample coverage by language and conditions
    • Use one global threshold without tests
    • Assume all accents are identical
    • Compare only call duration
  7. What data safeguards are useful for live-call scoring?
    • Share every recording with the full team
    • Restrict access and retention for recordings and derived scores
    • Keep scores forever
    • Use call audio for unrelated profiling by default
  8. Which measure best indicates whether a sentiment alert helps support?
    • Number of color changes on the dashboard
    • How often the model runs
    • Verified issue resolution alongside customer feedback and agent overrides
    • A higher average sentiment score alone