呼叫中心语音分析中的人工智能测验

当您参加测验时，就会出现答案和解释。

1. What output can automatic speech recognition provide for speech analytics?
   • A machine-generated transcript for downstream text analysis
   • A guaranteed speaker identity
   • A final diagnosis of the caller’s emotional state
   • A payment authorization
2. What does speaker diarization do in a call recording?
   • 将通话翻译成另一种语言
   • Segments the recording into speaker turns and assigns speaker labels
   • 测量通话的音量
   • 加密录音
3. In IBM’s two-stream call-center emotion-recognition study, which information streams are combined?
   • Video and keystroke timing
   • Customer purchase history and account balance
   • Acoustic and lexical information
   • Only transcript punctuation
4. 除了所使用的词语之外，哪种声学特征有助于检测客户的情绪状态？
   • 录音文件大小
   • 代理人的员工 ID
   • 电话运营商
   • 音调、节奏和音量（韵律）
5. In a two-stream call-center emotion system, what do acoustic features contribute?
   • Evidence from vocal delivery, such as how speech sounds
   • The caller’s verified identity
   • The account’s payment history
   • A guaranteed diagnosis of the caller’s emotion
6. What task did Vidrascu and Devillers study for customer-service phone calls?
   • Predicting credit scores from call recordings
   • Speaker segmentation and emotion-related analysis of conversational turns
   • Generating live agent sales recommendations
   • Automating legal decisions about customer accounts
7. Why might a call-center sentiment system combine lexical and acoustic features?
   • Acoustic features replace the transcript in every task
   • Lexical features reveal the phone carrier
   • The words and the way they are spoken provide complementary signals
   • Combining them guarantees the caller’s true emotion is known
8. What does word-error rate compare in speech recognition evaluation?
   • Speaker identity against a customer account
   • Emotion labels against call duration
   • Call volume against agent schedule
   • Substitutions, deletions and insertions against the words in a reference transcript