how to get accurate result from webspeechapi recognition code example

Example: javascript voice reader

var msg = new SpeechSynthesisUtterance();
msg.text = "Hello World";
window.speechSynthesis.speak(msg);