首页> 外国专利> COMMON AUTOMATIC LANGUAGE RECOGNITION AND TEXT-TO-LANGUAGE CHANGE FOR THE USE OF GEOGRAPHICAL NEURAL NETWORKS

COMMON AUTOMATIC LANGUAGE RECOGNITION AND TEXT-TO-LANGUAGE CHANGE FOR THE USE OF GEOGRAPHICAL NEURAL NETWORKS

机译:用于使用地理神经网络的常见自动语言识别和文本语言变更

摘要

This revelation provides a common automatic language recognition and text-to-language conversion using enemy neural networks. This script reveals a comprehensive Deep Learning-based system that can solve both ASR and TTS problems together using uncoupled text and audio samples. An adversely trained approach is used to generate a more robust independent neuronal TTS network and a neuronal ASR network,which can be used individually or simultaneously. The process of training the neural networks involves generating an audio probe using a text sample using the neural TTS network and then feeding the generated audio probe into the neural ASR network to regenerate the text. The difference between the regenerated text and theRecent text is used as a first loss to train the neural networks. A similar process is used for an audio call. The difference between the regenerated audio and the original audio is used as a second loss. A text and an audio critic are applied in a similar way to the output of the neural networkto generate additional losses for training.
机译:这种启示提供了使用敌方神经网络的常见自动语言识别和文本语言转换。此脚本显示了一个全面的基于深度学习的系统,可以使用未耦合的文本和音频样本在一起解决ASR和TTS问题。不利训练的方法用于生成更强大的独立神经元TTS网络和神经元ASR网络,其可以单独地或同时使用。训练神经网络的过程涉及使用神经TTS网络使用文本样本生成音频探测,然后将生成的音频探头馈送到神经ASR网络中以重新生成文本。再生文本与第一个文本之间的差异被用作培训神经网络的第一个损失。类似的过程用于音频呼叫。再生音频和原始音频之间的差异用作第二损耗。文本和音频评论仪以类似的方式应用于神经网络的输出,为培训产生额外的损失。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号