会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 5. 发明专利
    • Learning data selection device, identifiable speech recognition precision estimation device, learning data selection method, identifiable speech recognition precision estimation method and program
    • 学习数据选择装置,识别语音识别精度估计装置,学习数据选择方法,可识别语音识别精度估计方法和程序
    • JP2014160168A
    • 2014-09-04
    • JP2013030772
    • 2013-02-20
    • Nippon Telegr & Teleph Corp 日本電信電話株式会社
    • OGAWA ATSUNORIHORI TAKAAKINAKAMURA ATSUSHI
    • G10L15/06G10L15/01
    • PROBLEM TO BE SOLVED: To provide a learning data selection device capable of acquiring learning data for efficiently learning an identification model in an identifiable speech recognition precision estimation device applicable to an N-best recognition result.SOLUTION: The learning data selection device includes: an N-best recognition result storage part which stores N-best recognition result word strings; a learning data storage part which stores a first-place recognition result word string; an i-place recognition result extraction part which extracts an i-place recognition result word string from the N-best recognition result storage part; a k-place recognition result extraction part which extracts a k-place recognition result word string which is most similar to the i-place recognition result word string from each recognition result word string in the learning data storage part; a different word rate calculation part which calculates a different word rate R from the number of words different between each other in the i-place recognition result word string and k-place recognition result word string; and a comparison part which causes the learning data storage part to store the i-place recognition result word string when the different word rate R is over a predetermined threshold T.
    • 要解决的问题:提供一种学习数据选择装置,其能够获取用于有效学习可应用于N最佳识别结果的可识别语音识别精度估计装置中的识别模型的学习数据。解决方案:学习数据选择装置包括: 存储N最佳识别结果字串的N最佳识别结果存储部分; 存储第一位识别结果字串的学习数据存储部; i位置识别结果提取部分,从N最佳识别结果存储部分提取i位置识别结果字串; k位识别结果提取部分,从学习数据存储部分中的每个识别结果字串提取与i位置识别结果字串最相似的k位识别结果字串; 不同的字速率计算部分,其在i位置识别结果字串和k位识别结果字串中从彼此不同的字数计算不同的字速率R; 以及比较部,其使得当不同的字率R超过预定阈值T时,学习数据存储部存储i位置识别结果字串。
    • 6. 发明专利
    • Discriminative speech recognition accuracy estimating device, discriminative speech recognition accuracy estimating method and program
    • 歧视性语音识别精度估计设备,辨别语音识别精度评估方法和程序
    • JP2013171244A
    • 2013-09-02
    • JP2012036448
    • 2012-02-22
    • Nippon Telegr & Teleph Corp 日本電信電話株式会社
    • OGAWA ATSUNORIHORI TAKAAKINAKAMURA ATSUSHI
    • G10L15/01G10L15/18
    • PROBLEM TO BE SOLVED: To provide a discriminative speech recognition accuracy estimating device capable of obtaining an estimated value of recognition accuracy with high accuracy.SOLUTION: The discriminative speech recognition accuracy estimating device comprises: a speech recognition part for recognizing an inputted speech to output the speech recognition result; a word feature quantity vector forming part for forming a word feature quantity vector on the basis of information obtained from the speech recognition result; a recognition result word random classification part for forming a random word classification result which indicates correct answer probability of each word except ε in a 1-best recognition result word string, substitution error probability and insertion error probability for each segment, by using a word feature quantity vector of each word except ε in the 1-best recognition result word string formed by coupling words having maximum existence probability in a segment, and a recognition result word random classification model previously learned; and a random recognition accuracy calculating part for calculating speech recognition accuracy using the random word classification result.
    • 要解决的问题:提供能够以高精度获得识别精度的估计值的鉴别语音识别精度估计装置。解决方案:鉴别语音识别精度估计装置包括:语音识别部分,用于识别输入的语音以输出 语音识别结果; 基于从语音识别结果获得的信息,形成词特征量向量的单词特征量向量形成部; 用于形成随机字分类结果的识别结果字随机分类部分,其指示除egr之外的每个单词的正确答案概率; 通过使用除egr之外的每个单词的单词特征量向量,在1个最佳识别结果字串中,每个段的替换误差概率和插入误差概率; 在通过在段中具有最大存在概率的耦合词形成的1最佳识别结果字串中,以及先前学习的识别结果字随机分类模型; 以及随机识别精度计算部分,用于使用随机字分类结果来计算语音识别精度。
    • 7. 发明专利
    • Speech recognition accuracy estimating device, speech recognition precision estimating method and program
    • 语音识别精度估计装置,语音识别精度估计方法和程序
    • JP2013171243A
    • 2013-09-02
    • JP2012036447
    • 2012-02-22
    • Nippon Telegr & Teleph Corp 日本電信電話株式会社
    • OGAWA ATSUNORIHORI TAKAAKINAKAMURA ATSUSHI
    • G10L15/01G10L15/18
    • PROBLEM TO BE SOLVED: To provide a speech recognition accuracy estimating device capable of calculating recognition accuracy as a detailed numerical value after estimating the number of correct answers/substitution errors/insertion errors/deletion errors.SOLUTION: The speech recognition accuracy estimating device comprises: a speech recognition part for creating a word confusion network representing probability that any recognition result word exists for each segment from an inputted speech and probability (existence probability of ε) that any recognition result word does not exist; a word alignment network acquisition part for acquiring a word alignment network representing existence probability of a word with maximum existence probability as correct answer probability, total existence probability of words without maximum existence probability except ε as substitution error probability and the existence probability of ε as insertion error probability when ε has not the maximum existence probability in any segment, and representing total existence probability of the words without maximum existence probability as deletion error probability when ε has the maximum existence probability in any segment; and a random recognition accuracy calculating part for calculating speech recognition accuracy.
    • 要解决的问题:提供一种在估计正确答案/替换错误/插入错误/删除错误的数量之后能够计算识别精度作为详细数值的语音识别精度估计装置。解决方案:语音识别精度估计装置包括: 语音识别部分,用于创建表示任何识别结果字从输入的语音存在于每个段的概率和概率(存在概率)的单词混淆网络,所述识别结果字不存在任何识别结果字; 用于获取表示具有最大存在概率的单词的存在概率的单词对齐网络作为正确答案概率的单词对齐网络获取部分,除“&egr”之外的没有最大存在概率的单词的总存在概率; 作为替代误差概率和&egr的存在概率; 作为插入误差概率 在任何段中没有最大存在概率,并且表示没有最大存在概率的单词的总存在概率作为&egr的删除错误概率; 在任何段中都具有最大存在概率; 以及用于计算语音识别精度的随机识别精度计算部分。