Share via


使用語音轉換文字顯示文字格式設定

語音轉換文字提供格式化功能的數位,以確保轉譯的文字清晰易讀。 如需如何使用每項功能來改善最終文字輸出的整體清晰性,請參閱下列各節。

ITN

反向文字正規化 (ITN) 是將口語文字轉換成其書面形式的程式。 例如,口語 「four」 字會轉換成寫入格式 「4」。 語音轉換文字服務會完成此程式,且無法設定。 部分支援的文字格式包括日期、時間、小數點、貨幣、地址、電子郵件和電話號碼。 您可以自然說話,服務會如預期般格式化文字。 下表顯示套用至文字輸出的ITN規則。

辨識的語音 顯示文字
that will cost nine hundred dollars That will cost $900.
my phone number is one eight hundred, four five six, eight nine ten My phone number is 1-800-456-8910.
the time is six forty five p m The time is 6:45 PM.
I live on thirty five lexington avenue I live on 35 Lexington Ave.
the answer is six point five The answer is 6.5.
send it to support at help dot com Send it to support@help.com.

大寫

語音轉換文字模型可辨識應大寫的文字,以改善可讀性、精確度和文法。 例如,語音服務會自動將句子開頭的適當名詞和單字大寫。 下表顯示一些範例。

辨識的語音 顯示文字
i got an x l t shirt I got an XL t-shirt.
my name is jennifer smith My name is Jennifer Smith.
i want to visit new york city I want to visit New York City.

拿掉不流暢

說話時,人們常會說出口語、重複字組,以及說“uhm” 或 “uh” 等填充字。 語音轉換文字可以辨識這類不多,並從顯示文字中移除它們。 不流暢的移除非常適合轉譯即時未標語的語音,以便稍後再閱讀。 下表顯示一些範例。

辨識的語音 顯示文字
i uh said that we can go to the uhmm movies I said that we can go to the movies.
its its not that big of uhm a deal It's not that big of a deal.
umm i think tomorrow should work I think tomorrow should work.

標點符號

語音轉換文字會自動標點您的文字,以改善清晰度。 標點符號有助於讀取回呼或交談轉譯。 下表顯示一些範例。

辨識的語音 顯示文字
how are you How are you?
we can go to the mall park or beach We can go to the mall, park, or beach.

當您使用語音轉換文字進行連續辨識時,您可以設定語音服務來辨識明確的標點符號。 然後,您可以大聲說話標點符號,讓您的文字更清晰。 在您想要使用複雜標點符號而不需稍後合併的情況下,這特別有用。 下表顯示一些範例。

辨識的語音 顯示文字
they entered the room dot dot dot They entered the room...
i heart emoji you period I <3 you.
the options are apple forward slash banana forward slash orange period The options are apple/banana/orange.
are you sure question mark Are you sure?

當您使用語音轉換文字進行連續辨識時,請使用語音 SDK 來啟用聽寫模式。 此模式會導致語音設定實例解譯句子結構的文字描述,例如標點符號。

speechConfig.EnableDictation();
speechConfig->EnableDictation();
speechConfig.EnableDictation()
speechConfig.enableDictation();
speechConfig.enableDictation();
[self.speechConfig enableDictation];
self.speechConfig!.enableDictation()
speech_config.enable_dictation()

不雅內容篩選條件

您可以指定是否要在最後的轉譯文字中遮罩、移除或顯示粗話。 遮罩會將粗話字取代為星號 — 字元,讓您可以保留文字的原始情感,同時使其更適合某些情況

注意

Microsoft 也保留遮罩或移除任何被視為不適當的字詞的權利。 不論您是否啟用不雅的篩選,語音服務都不會傳回這類單字。

粗話篩選選項如下:

  • Masked:以星號 \ 字元取代粗話字中的字母。 遮罩是預設選項。
  • Raw:包含粗話字。
  • Removed:移除粗話字。

例如,若要從語音辨識結果中移除粗話字,請將粗話篩選設定為 Removed ,如下所示:

speechConfig.SetProfanity(ProfanityOption.Removed);
speechConfig->SetProfanity(ProfanityOption::Removed);
speechConfig.SetProfanity(common.Removed)
speechConfig.setProfanity(ProfanityOption.Removed);
speechConfig.setProfanity(sdk.ProfanityOption.Removed);
[self.speechConfig setProfanityOptionTo:SPXSpeechConfigProfanityOption.SPXSpeechConfigProfanityOption_ProfanityRemoved];
self.speechConfig!.setProfanityOptionTo(SPXSpeechConfigProfanityOption_ProfanityRemoved)
speech_config.set_profanity(speechsdk.ProfanityOption.Removed)
spx recognize --file caption.this.mp4 --format any --profanity masked --output vtt file - --output srt file -

不雅內容篩選會套用至結果 TextMaskedNormalizedForm 屬性。 不雅內容篩選不會套用至結果 LexicalFormNormalizedForm 屬性。 這兩個篩選都不會套用至文字層級結果。

下一步