引導式測驗 · 中等 等級
語音測驗的文字規範化
測試語音文字規範化:擴展數字和符號、對標記進行分類、解析上下文以及使用規則或神經模型。
問題 1 的 8
語音文字規範化主要做什麼?
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。
- 語音文字規範化主要做什麼?
- 將書寫文字重寫為完全說出來的單字
- 壓縮音訊檔案
- 將文字翻譯成另一種語言
- 偵測說話者的情緒
- 一個好的系統應該如何規範語音的「5 美元」?
- ‘五塊錢’
- “五美元”
- '五'
- ‘百分之五’
- 為什麼要規範化「Dr.」這樣的標記還是「IV」棘手?
- 這些標記總是以相同的方式發音
- 它們不包含字母
- 它們的正確擴展取決於上下文
- 它們僅用於數學
- What data did the cited neural text-normalization study use to evaluate its models?
- Unlabeled phone audio only
- Medical records annotated by clinicians
- A small set of manually written code snippets
- English and Russian Wikipedia sentences verbalized by Google TTS’s Kestrel system
- What safeguard against unrecoverable errors does the cited neural text-normalization paper discuss?
- WFST-based covering grammars
- Removing the reference transcript
- Replacing all token classes with a speech waveform
- Using image segmentation at inference
- 在經典規範化中,為標記分配「符號類別」的作用是什麼?
- 它選擇說話者的聲音
- 它標記標記(例如日期、金錢、度量),以便應用正確的語言表達器
- 它壓縮文字
- 它刪除所有標點符號
- In the Kestrel TTS normalization system, what follows tokenization and classification?
- A speaker diarization pass labels voices
- A vocoder renders the waveform
- A verbalization stage applies grammars to turn classified tokens into spoken forms
- An image model recognizes the page layout
- Why can English and Russian text require different normalization grammars?
- Every language uses identical number and date formats
- Language has no effect on spoken forms
- Russian text cannot be normalized for speech
- Written conventions and verbalization rules vary by language