CaGOP v.s. GOP (Examples for Interspeech 2020)

This page is presenting examples of the paper.

Examples with Explanation

• Reference Text: “HAVE YOU EVER SEEN AN ANT WORK”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  • GOP:
    • Overall: 58.4698
    • Details: Txt in Json Format
  • CaGOP:
    • Overall: 100
    • Details: Txt in Json Format
  • Explanation: This example mainly shows the effect of the transition factor. As the GOP focuses on the whole aligned segments, it may include some phonetic transition that lower the phoneme-level score. For CaGOP, the transition factor helps to remove those transitions, thus, improve the scoring process.

 

• Reference Text: “PLEASE COME QUICKLY”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  •  
    • GOP:
      • Overall: 74
      • Details: Txt in Json Format
    • CaGOP:
      • Overall: 64
      • Details: Txt in Json Format
    • Explanation: This example mainly shows the effect of the duration factor. As the GOP does not pay attention to the prosody of the utterance, it cannot detect the non-native prosody. The CaGOP adopts the duration factor to combine the prosody naturalness with the pronunciation scoring, which makes it more consistent with the human.

 

Other Examples:

• Reference Text: “ANTS”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  • GOP:
    • Overall: 26
    • Details: Txt in Json Format
  • CaGOP:
    • Overall: 75
    • Details: Txt in Json Format

 

• Reference Text: “LIKE YOUR FATHER”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  • 
  • GOP:
    • Overall: 51
    • Details: Txt in Json Format
  • CaGOP:
    • Overall: 76
    • Details: Txt in Json Format

 

• Reference Text: “WHAT IS THE BEST WAY TO BE, THE LITTLE PRINCESS ASKED ”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  • 
  • GOP:
    • Overall: 44
    • Details: Txt in Json Format
  • CaGOP:
    • Overall: 100
    • Details: Txt in Json Format

 

• Reference Text: “HOW WINGS LOOK AT THE BEE”S WINGS”
  • Audio:
  • Visualization using Audacity in the format of  Phone (GOP Score, CaGOP Score):
  • 
  • GOP:
    • Overall: 73
    • Details: Txt in Json Format
  • CaGOP:
    • Overall: 80
    • Details: Txt in Json Format