Combined Application of Speech Recognition and Natural Language Processing Technologies in the Electric Power Industry
, , , and
Mar 19, 2025
About this article
Published Online: Mar 19, 2025
Received: Nov 02, 2024
Accepted: Feb 02, 2025
DOI: https://doi.org/10.2478/amns-2025-0530
Keywords
© 2025 Tao Xu et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
The system response time varies according to different types of voice duration
| Speech duration/s | Response time/s | ||
|---|---|---|---|
| Real-time voice | Voice file | Telephone terminal | |
| 5 | 2.6 | 1.6 | 3.3 |
| 10 | 3.0 | 1.9 | 3.8 |
| 20 | 4.5 | 3.0 | 4.9 |
| 30 | 5.3 | 4.4 | 5.8 |
| 60 | 8.5 | 7.1 | 9.7 |
Comparison of different input features
| Data set | Input feature | CER% |
|---|---|---|
| THCHS-30 | Spectrogram | 15.72 |
| Fbank | 17.28 | |
| MFCC | 18.68 | |
| AISHELL-1 | Spectrogram | 15.54 |
| Fbank | 16.77 | |
| MFCC | 19.30 |
Comparison of experimental results on Power voice data set
| Method | Parameter quantity | WER(%) | RTF |
|---|---|---|---|
| Discriminative | 20 | 11.42 | 0.040 |
| Pseudo Visual | 23 | 13.34 | 0.047 |
| Vanilla Transformer | 25 | 10.62 | 0.031 |
| Speech-Transformer | 26 | 10.06 | 0.026 |
| Open-Transformer | 36 | 9.64 | 0.029 |
| Ours | 25 | 8.21 | 0.017 |