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
    • 33. 发明专利
    • HIERARCHIZATION TERM COLLATING SYSTEM
    • JPH04171591A
    • 1992-06-18
    • JP29888890
    • 1990-11-06
    • NIPPON TELEGRAPH & TELEPHONE
    • SAKAMOTO TOMIYOSHIOGURO MASAMINAKAMURA OSAMU
    • G06K9/72G06F17/22G06F17/28
    • PURPOSE:To output a term candidate with high reliability by outputting the whole character string term candidate when the whole character string term candidate includes a partial character string term candidate and connecting and outputting a residual character candidates except the partial character string term candidate among the character candidates and the partial character string term candidate when it does not include. CONSTITUTION:A recognized character candidate is collated with a whole character string candidate dictionary 105 and a partial character string candidate dictionary 104, the whole character string term candidate and a partial character string term candidate are outputted respectively and the outputted whole character string term candidate and partial character string candidates are compared. As the result of the comparison, when the whole character string term candidate includes the partial character string term candidate, the whole character string term candidate is outputted and when it does not include, the residual character candidates except the partial character string term candidate among the character candidates and the partial character string term candidate are connected and are outputted. Thus, a character recognizing result is collated with a term dictionary and the precise term candidate can be obtained with high reliability.
    • 34. 发明专利
    • ANALOGOUS CHARACTER DISCRIMINATING SYSTEM
    • JPH0496885A
    • 1992-03-30
    • JP21418690
    • 1990-08-13
    • NIPPON TELEGRAPH & TELEPHONE
    • TANAKA AKIMICHINAKAMURA OSAMUHOSOKAWA MASAYOSHI
    • G06K9/68G06K9/62
    • PURPOSE:To enable an stable application to characteristic vectors whose variance is almost 0 dimension or more than 1,000 dimensions by obtaining a differential vector set after extracting the characteristic vectors from character images, and moreover obtaining a transformation matrix for analogous characters. CONSTITUTION:When an inputted character image is applied, the characteristic vector is extracted by a characteristic vector extracting means 101, and converted by a converting means 102 for the analogous characters. An average in a character dictionary 108 in a category included in an analogous character table is converted by the converting means 102 for the analogous characters, and a distance calculation is operated between it and the character vector obtained from the inputted character image, by a distance calculating means 105. Then, a character code 106 in the category whose distance is minimum is outputted. Thus, the analogous character discrimination transformation matrix can be stably obtained towards the character vectors whose distributions are almost 0 dimension or more than 1,000 dimensions, such as a PDC characteristic.
    • 37. 发明专利
    • VOCABULARY DICTIONARY RETRIEVING DEVICE
    • JPH02121078A
    • 1990-05-08
    • JP27522488
    • 1988-10-31
    • NIPPON TELEGRAPH & TELEPHONE
    • NAKAMURA OSAMUKITAMURA TADASHIOGURO MASAMI
    • G06K9/72G06F17/30
    • PURPOSE:To highly speedily retrieve a vocabulary dictionary and to improve efficiency for identifying an accurate vocabulary by removing a high-frequency character to be used, which has high frequency and is used in many vocabularies, from the selection of the vocabularies. CONSTITUTION:The title device is composed of a processor 9, a program memory 11, a working memory 13, and a table memory 15. Further, for respective candidate characters obtained by removing the prescribed high-frequency character to be used from input character candidates, vocabularies having the same character positions and the same number of the characters of a character string are selected, scores included in these respective vocabularies are added, the total scores of the respective vocabularies are calculated, and when the respective vocabularies include the high-frequency character to be used, prescribed scores corresponding to the high-frequency character to be used are added to the total scores, and the scores are corrected. Thus, the vocabulary dictionary can be retrieved at a high speed, and the vocabulary dictionary can be highly identified.