首页> 外国专利> Language model specific device, language model specific method, acoustic model specific device, and acoustic model specific method

Language model specific device, language model specific method, acoustic model specific device, and acoustic model specific method

机译:语言模型专用设备,语言模型专用方法,声学模型专用设备和声学模型专用方法

摘要

PPROBLEM TO BE SOLVED: To determine an optimal language model in a short period of time. PSOLUTION: The language model-specifying device 100 includes: a voice recognition section 102 for performing voice recognition for an input voice by utilizing an integrated LM integrated from a plurality of language models (LM) discriminated by an ID; and a determination section 103 for determining which LM is used among the plurality of LMs for the voice recognition based on the ID attached to the character string of a voice recognition result. Preferably, in a "reading" field included in each LM, a first ID-imparting section 107 and second ID-imparting section 109 for imparting an ID for specifying the LM, and an LM integration section 106 for integrating each LM including the "reading" field to which the ID is imparted, to create an integrated LM, are also provided to the device 100. The voice recognition section 102 performs voice recognition to the input voice by utilizing the integrated LM, and outputs a character string to which the ID for specifying any LM among the plurality of LMs is imparted, as a voice recognition result, to the determination section 103. PCOPYRIGHT: (C)2011,JPO&INPIT
机译:

要解决的问题:在短时间内确定最佳语言模型。

解决方案:语言模型指定设备100包括:语音识别部分102,用于通过利用从ID区分的多个语言模型(LM)集成的集成LM对输入语音进行语音识别;确定部分103,用于基于附加在语音识别结果的字符串上的ID,确定在用于语音识别的多个LM中使用哪个LM。优选地,在每个LM中包括的“读取”字段中,第一ID赋予部分107和第二ID赋予部分109用于赋予用于指定LM的ID,并且LM整合部分106用于对包括“读取”的每个LM进行整合。被赋予ID以创建集成LM的字段也被提供给设备100。语音识别部分102通过使用集成LM对输入的语音执行语音识别,并且输出ID被输入到的字符串。

版权:(C)2011,JPO&INPIT,将用于指定多个LM中的任何一个LM的语音识别结果作为语音识别结果。

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