Quiz lié à un guide · Moyen Niveau

Quiz sur l'alignement forcé

Passez en revue la synchronisation de la transcription vers l'audio, les lexiques de prononciation, l'alignement HMM, la segmentation longue et les limites de la sortie d'alignement forcé.

Chemins de guidage associésAlignement forcé
Question 1 de 8

Étant donné l’audio et sa transcription, quel résultat l’alignement forcé crée-t-il ?

Prévisualisez chaque question8 des questions

Les réponses et les explications apparaissent au fur et à mesure que vous répondez au quiz.

  1. Étant donné l’audio et sa transcription, quel résultat l’alignement forcé crée-t-il ?
    • Horodatages précis de début et de fin pour les mots ou les sons dans une transcription connue
    • Une traduction dans une autre langue
    • A new transcript created without using supplied words
    • A noise-reduced audio file
  2. Why is alignment called “forced”?
    • The model must increase recording volume
    • The model is constrained to fit the supplied transcript
    • A user must set each boundary by hand
    • The system translates the transcript
  3. What does a pronunciation dictionary provide in a classic aligner?
    • The recording’s sample rate
    • The final timestamps for every word
    • A mapping from written words to expected phone sequences
    • A translation for each language
  4. What search method does the classic Kaldi-based Montreal Forced Aligner use to find a likely sequence through speech?
    • Bubble sort over word strings
    • Randomly selecting timestamps
    • Linear regression on sentence length
    • Viterbi decoding over an HMM-based acoustic model
  5. In WhisperX, what helps handle long-form audio before word-level forced alignment?
    • Voice-activity segmentation into speech chunks
    • Deleting the supplied transcript
    • Changing every word into a phoneme by hand
    • Skipping audio and estimating timestamps from punctuation
  6. Which task can use the word boundaries produced by an aligner?
    • Determining the speaker’s identity with certainty
    • Synchronizing captions or karaoke text to speech
    • Translating speech without a transcript
    • Removing every background sound
  7. Can forced alignment establish that the supplied transcript is factually correct?
    • Yes, it always corrects every transcript error
    • Only when the audio is stereo
    • No; it estimates timing for the provided text and does not verify the words
    • Yes, if a pronunciation dictionary is supplied
  8. What can happen when a word is missing from an aligner’s pronunciation dictionary?
    • The audio automatically becomes translated
    • All following timestamps become legally certified
    • The word is guaranteed to align correctly without a model
    • The tool may need a pronunciation entry or a supported unknown-word strategy