How to segment Speech-To-Text recognized result text ?

Question

Hello,

i'm using Speech SDK. How can i segment Speech-To-Text recognized result text ?

Answer 1

To segment Speech-To-Text recognized result text, you can use the getOffset() method of the RecognitionResult class. This method returns the start and end positions of the recognized text in the original audio stream. You can use these positions to segment the recognized text as needed. Here's an example in Java:

String recognizedText = result.getText();
int startOffset = result.getOffset();
int endOffset = startOffset + recognizedText.length();
String originalAudio = // get original audio stream
String segmentedAudio = originalAudio.substring(startOffset, endOffset);

In this example, result is an instance of the RecognitionResult class, which is obtained from the SpeechRecognizer object. recognizedText is the text that was recognized from the audio stream, and startOffset and endOffset are the start and end positions of the recognized text in the original audio stream. originalAudio is the original audio stream, and segmentedAudio is the portion of the audio stream that corresponds to the recognized text.

---

References:

• Get speech recognition results (programming-language-cli)
• How to recognize speech (programming-language-java)