引导式测验 · 中等 等级

Kaldi 语音识别工具包测验

查看 Kaldi 的工具包和解码图设计，包括历史作者身份、声学建模、相关 FSA 工具、许可和扬声器元数据。

相关引导路径Kaldi 语音识别工具包
问题 1 的 8

Which programming language is used for Kaldi’s documented core implementation?

预览每个问题8 问题

当您参加测验时，就会出现答案和解释。

  1. Which programming language is used for Kaldi’s documented core implementation?
    • 爪哇
    • 铁锈
    • Python only
    • C++
  2. Kaldi 使用什么数学结构将其声学模型、词典和语言模型组合成一个解码图？
    • 循环神经网络
    • 支持向量机
    • 加权有限状态传感器 (WFST)
    • 隐马尔可夫树
  3. Who is listed first in the project’s recommended 2011 Kaldi toolkit paper citation?
    • 杰弗里·辛顿
    • 丹尼尔·波维
    • 严乐存
    • 吴恩达
  4. In a classic Kaldi GMM acoustic model, what do Gaussian mixtures model?
    • Acoustic likelihoods for speech states
    • Only the spelling of every output word
    • Only a file’s speaker-ID string
    • The application’s interface color
  5. Which related project describes FSA/FST algorithms with differentiation and PyTorch compatibility?
    • A universal replacement requirement for every Kaldi recipe
    • Only the original HTK text formatter
    • k2
    • An image-compression file format
  6. Kaldi 的 HCLG 解码图中的“L”代表什么？
    • 响度归一化
    • 语言模型
    • 损失函数
    • 发音词典
  7. Which license does the official Kaldi project documentation list?
    • Apache License version 2.0
    • A proprietary-only research license
    • An exclusive GPL version 3 requirement
    • No license terms of any kind
  8. What association is recorded by a Kaldi data-directory utt2spk row?
    • A word mapped to its pronunciation phones
    • An utterance identifier mapped to a speaker identifier
    • A decoding state mapped to a language-model probability
    • A feature matrix mapped only to its pixel dimensions