Ikibazo gihuza ibibazo · Biragoye Urwego

HuBERT Yiyobora Ikibazo

Reba uburyo HuBERT ikora intego zikomoka kumajwi yipfundikiriye amajwi, igasubiramo ibirango, ikanasuzuma ibyerekanwe byize kubipimo byerekana.

Inzira zijyanye no kuyoboraHubert Yigenga Ijambo
Ikibazo 1 Bya 8

Nigute HuBERT itanga intego igerageza guhanura mugihe cy'amahugurwa?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Nigute HuBERT itanga intego igerageza guhanura mugihe cy'amahugurwa?
    • Offline k-bisobanura guhuza ibiranga amajwi mubice byihishe
    • Inyandiko-mvugo y'abantu
    • Ibirango by'urusaku bisanzwe
    • Itandukaniro rirwanya ibyitegererezo byo kurangaza
  2. Ni ubuhe bwoko bw'inyandiko buzwi bwahumekeye HuBERT gahunda yo guhugura-guhanura?
    • GPT-3
    • BERT
    • ResNet
    • Ijambo2Vec
  3. Nigute HuBERT itezimbere ibirango byayo hejuru yimyitozo ikurikirana?
    • Ikoresha ibisobanuro byinshi
    • Yongera urusaku rwinshi
    • Irongera ikusanya ikoresheje icyitegererezo cyimbere cyimbere imbere
    • Ihagarika cluster ya mbere iteka ryose
  4. Ni ibihe bintu bikunze guhurizwa hamwe muri HuBERT ya mbere cyane?
    • Amagambo yanyuma
    • Amashusho
    • Indangamuntu ya Speaker
    • Ibintu byoroshye bya MFCC acoustic
  5. Kuki HuBERT ishobora kwiga imiterere yingirakamaro nubwo ihuriro ryayo ridatunganye?
    • Birakenewe gusa guhanura indangamuntu kumwanya uhishe, igumaho umurimo uhoraho
    • Irengagiza ihuriro rwose
    • Ikoresha oracle nziza
    • Ifata mu mutwe ibizamini
  6. Which pretraining challenge does HuBERT address by using offline cluster labels?
    • Speech has no temporal structure
    • There is no lexicon of sound units available during pretraining
    • Every speaker uses the same pitch
    • Audio contains no repeated patterns
  7. In the HuBERT paper, what relative WER reductions did the 1B-parameter model report on dev-other and test-other?
    • Exactly 50% on both sets
    • No measurable change on either set
    • Up to 19% on dev-other and 13% on test-other
    • Up to 13% on dev-other and 19% on test-other
  8. Which benchmark sets did the original HuBERT paper use for speech-recognition results?
    • IshushoNet na COCO
    • GLUE and SuperGLUE
    • SQuAD and TriviaQA
    • LibriSpeech na Libri-mucyo