Summary of Efficiently Trained Low-resource Mongolian Text-to-speech System Based on Fullconv-tts, by Ziqi Liang
Efficiently Trained Low-Resource Mongolian Text-to-Speech System Based On FullConv-TTS
by Ziqi Liang
First submitted to arxiv on: 24 Oct 2022
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel text-to-speech (TTS) system based on deep convolutional neural networks (CNNs), eliminating the need for recurrent neural network (RNN) components. The authors leverage CNN’s high parallelism to significantly reduce training time, achieving comparable quality to classic TTS models like Tacotron. To enhance model generality and robustness, they employ various data augmentation methods, including Time Warping, Frequency Mask, and Time Mask. Experimental results demonstrate the effectiveness of this approach in reducing training time while maintaining speech quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to make computers talk using special computer programs called convolutional neural networks (CNNs). Right now, these programs use another type of program called recurrent neural networks (RNNs) to make speech sound natural. However, this requires powerful computers and takes a long time. The authors discovered that CNNs can do the job just as well without RNNs! They also found ways to make their computer program more flexible and resistant to errors by changing some of the input data. This resulted in faster training times and better sounding speech. |
Keywords
* Artificial intelligence * Cnn * Data augmentation * Mask * Neural network * Rnn
