Highlighting of the text according to the sound file(like karaoke)

There's no total speech recognition built into Unity. All you can really do is mess around with the spectrum data, but that would be a long shot. The only thing you can do besides that is sync it all manually. That's with Unity alone. Another thing you could do is look for a Plugin, which you would only be able to use with Unity Pro or Mobile.

Just to help get started a bit, here's an example of how to use SpecData if you decide to go with it.

 using UnityEngine;
 using System.Collections;
 
 public class specData : MonoBehaviour 
 {
     public AudioListener Listener;
     public int numSamples = 128;
     public GameObject abar;
     float[] numberleft;
     float[] numberright;
     GameObject[] thebarsleft;
     GameObject[] thebarsright;
     float spacing;
     float width;
 
     void Start () 
     {
         thebarsleft = new GameObject[numSamples];
         thebarsright = new GameObject[numSamples];
         spacing = .4f - (numSamples * 0.001f);
         width = .1f - (numSamples * 0.001f);
         for(int i=0; i < numSamples; i++)
         { 
             float xpos = i*spacing -8.0f;
             Vector3 positionleft = new Vector3(xpos,3, 0);
                 thebarsleft[i] = (GameObject)Instantiate(abar, positionleft, Quaternion.identity) as GameObject;
                 thebarsleft[i].transform.localScale = new Vector3(width,1,0.2f);
             Vector3 positionright = new Vector3(xpos,-3, 0);
                 thebarsright[i] = (GameObject)Instantiate(abar, positionright, Quaternion.identity) as GameObject;  
                 thebarsright[i].transform.localScale = new Vector3(width,1,0.2f);
         }
 
     }
     
     void Update () 
     {
 
         float[] numberleft = AudioListener.GetSpectrumData (numSamples, 0,FFTWindow.BlackmanHarris);
         float[] numberright = AudioListener.GetSpectrumData (numSamples, 1,FFTWindow.BlackmanHarris);
 
         for(int i=0; i < numSamples; i++){
             if (float.IsInfinity(numberleft[i]*30) || float.IsNaN(numberleft[i]*30)){
 
             }else{
 
                 thebarsleft[i].transform.localScale = new Vector3(width, numberleft[i]*30,0.2f);
 
                 thebarsright[i].transform.localScale = new Vector3(width, numberright[i]*30,0.2f);  
 
             }
 
         }
 
     }
 
 }