JPH04340686A - Name dictionary for post-processing of character recognition - Google Patents

Abstract

PURPOSE:To easily obtain proper reading and Chinese characters (KANJI) by providing the dictionary equipped with a family name reading part, name reading part, family name word part and name word part and arranged while being divided for each number of characters. CONSTITUTION:A family name reading part 2 stores KANJI 'Nakamura' constituting a family name in respect to reading 'Nakamura', for example, a name reading part 3 stores KANJI 'Taro' constituting a name in respect to reading 'Taro', for example, a family name word part 4 stores 'Nakamura' constituting the reading in respect to the KANJI 'Nakamura' and a name word part 5 stores 'Taro' constituting the reading in respect to the KANJI 'Taro' respectively correspondently. For example, when the reading of 'Nakamufu' is recognized as one candidate, the part of 'reading composed of four characters' is investigated in the family name reading part 2, and it is discriminated that any family name having the reading of 'Nakamufu' does not exist but the reading of 'Nakamura' exists. Further, the KANJI corresponding to 'Nakamura' is investigated among several candidates.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a name dictionary for post-processing of character recognition used in post-processing of a character recognition device.

0002

[Prior Art] Conventionally, in character recognition devices that read and recognize names on forms, a name dictionary is indexed based on a plurality of recognized candidate names, and the name considered to be the most preferable is selected as the final name. I am trying to output it as a result.

[0003] The name dictionary used for such post-processing is
Conventionally, a device has been provided with a surname storage section in which information related to the surname is stored and a given name storage section in which information related to the given name is stored.

0004

[Problems to be Solved by the Invention] In the past, surnames were not sorted and stored by the number of characters in the surname, and given names by the number of characters in the first name. For this reason, the final result is a name whose reading matches perfectly. Therefore, it is difficult, for example, to extract several names that match only the first character and character length in terms of pronunciation. That is, for example, when only one part of the characters is misrecognized in terms of pronunciation, it is not possible to perform a process of indexing possible names.

An object of the present invention is to enable processing such as indexing several names whose first characters and character lengths match as candidates.

[0006]

[Means for Solving the Problems] FIG. 1 shows a diagram of the basic configuration of the present invention. In the figure, 1 is the name dictionary, 2 is the surname reading part, and 3 is the name dictionary.
4 represents the name reading part, 4 represents the surname word part, 5 represents the name word part, 6 represents the general reading part, and 7 represents the general word part.

[0007] In the surname reading unit 2, the kanji character ``Nakamura'' constituting the surname is stored in correspondence with the pronunciation ``Nakamura'', for example ``Nakamura-Nakamura''. If the number of kana characters that make up the reading is "2", "3", "4", "5", or other than "2 to 5" characters. For example, if it is "2", it is stored in the order of the numbers within the case "2".

[0008] In the name pronunciation section 3, the kanji character ``Taro'' constituting the given name is stored in correspondence with the pronunciation ``Taro'', for example, ``Taro-Taro''. If the number of kana characters that make up the reading is "2", then "
3", "4", "5", and cases other than "2-5" characters.
Furthermore, they are stored in the order of iueo.

[0009] In the surname word part 4, for example, ``Nakamura-Nakamura'', which constitutes the pronunciation for the kanji ``Nakamura'',
Nakamura" are stored in association with each other. and,
It is divided into cases where the character that makes up the kanji is "1", "2", "3", and characters other than "1 to 3", and furthermore, for example, "1". In terms of cases, they are arranged and stored in the order of the number of strokes within the case, which is "1".

[0010] In the famous word section 5, the kanji character "Taro" is stored in correspondence with the pronunciation "Taro", for example, "Taro-Taro". In this case too,
It is stored under the same arrangement as the last name word section 4. In the general reading section 6, company names and the like are stored in the form of "reading-kanji", for example. Further, in the general word section 7, for example, company names are stored in the form of "kanji-yomi". The fact that it is divided according to the number of characters can be considered to be the same as the surname pronunciation section 2, the surname word section 4, etc.

[0011]

[Operation] In the character recognition device, the reading of the surname is, for example, “
Suppose that the reading ``Nakamfu'' is recognized as one candidate. In this case, the ``pronunciation consisting of four characters'' part in the surname reading part 2 is checked, and it is determined that there is no name with the pronunciation of ``Nakamuf'' but that ``Nakamura'' does exist. Also, the reading ``Nakamura'' was found, and the kanji corresponding to ``Nakamura'' is ``Nakamura''.
You can also check whether it is "Nakamura" or "Nakamura". Furthermore, it becomes easy to check the pronunciation based on the characters of the surname in kanji.

Furthermore, by providing a general reading section and a general word section, it can be used to read company names, etc.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIGS. 2 and 3 are flowcharts showing post-processing aspects of dictionary use according to the present invention. Figure 2 mainly shows “
This corresponds to the processing starting from "Yomi", and FIG. 3 mainly corresponds to the processing starting from "Kanji".

(S1): In step S1, data (kana+kanji) recognized by the character recognition device is received. (S2): Based on the combination condition of the first character of the pronunciation and the number of characters of the pronunciation, the dictionary 1 is accessed and data matching the condition is extracted.

(S3): The extracted data is compared with the input data, and scores such as similarity are calculated. (S4): It is checked whether there is still dictionary data that satisfies the same conditions as in step S2, and if there is, the process returns to step S2.

(S5): It is checked whether the score of the extracted data is greater than a threshold value. As described above, some data that can be candidates are determined from the "reading" perspective.

(S6): The dictionary 1 is accessed and the kanji corresponding to the candidate data is extracted. (S7): Compare the kanji of the input data with the kanji from the dictionary and calculate the score.

(S8): It is checked whether homonyms still exist. (S9): It is checked whether the obtained score is greater than or equal to a threshold value. (S10): If it is equal to or greater than the threshold value, it is output as output data. (S11): NO in step S5 or step S9
If so, proceed to processing from "Kanji".

In the process starting from "Kanji", the process proceeds to step S12 shown in FIG. (S12): Extract from the dictionary 1 kanji and their readings that match the kanji combined with the input data. For example, regarding the kanji in the input data, for example, "komoto", "yamaki",
If "Kawago" is listed as a candidate, "Komoto", "Ogi", "Kogo", "Yamamoto", "Yamaki", "Yamago", "Kawamoto", "Kawaki" , "Kawago" and their pronunciations are extracted from the dictionary 1.

(S13): Input data (kana + kanji)
and the dictionary data (kana + kanji) to calculate the score. For example, the input data of the "reading" corresponding to the kanji characters "Komoto", "Yamaki", and "Kawaki" in the input data shown in step S12 lists "Kamamoto", "Yanuhono", and "Ametai" as candidates. In this case, "Kamamoto", "Kanumoto", "Kamemoto", "Yamamoto", "Yanumoto", "Yamemoto", "Amamoto", "Anumoto"
, "Amemoto", "Kamahot", "Kamatato"...
・ etc. are obtained as combinations, and their scores are calculated by combining them with the above Kanji example candidates.

(S14): It is checked whether there is still data to be examined in the dictionary data 1. (S15): It is checked whether there are other combinations to be examined. (S16): It is checked whether the score is equal to or greater than a threshold value.

(S17): If YES, the data is output as output data. (S18): If NO, output fails. The processing is performed as described above, but if the input data is, for example, for the pronunciation, Kamamotoyanuhonoaametai is obtained as a candidate, and for the kanji, Komotoyamakikawarai is obtained as a candidate, the present invention By post-processing using the dictionary, we were able to obtain Yamamoto-Yamamoto.

[0023]

[Effects of the Invention] As explained above, according to the present invention, appropriate "readings and kanji" are obtained through post-processing from candidate characters (readings and kanji) obtained during the recognition process in a character recognition device. It becomes easier.

[Brief explanation of drawings]

FIG. 1 shows a diagram of the principle configuration of the present invention.

FIG. 2 is a flowchart showing a post-processing aspect of dictionary use.

FIG. 3 is a flowchart showing a post-processing mode using a dictionary.

[Explanation of symbols]

1 Name dictionary 2 Surname reading part 3. Famous reading section 4 Surname word part 5. Famous words section 6 General reading section 7 General vocabulary section

Claims (2)

[Claims] [Claim 1] In a name dictionary for character recognition post-processing used in a character recognition device that reads and recognizes at least characters corresponding to a name written on a form, for a surname, a kanji is associated with a reading and a kanji is , a surname reading part (2) in which readings with the same number of characters are arranged and stored in a predetermined order for each number of characters making up the reading; For each number of characters, a name reading section (3) stores readings with the same number of characters arranged in a predetermined order, and for surnames, readings are associated with kanji, and for each number of characters constituting the kanji, A surname word section (4) in which kanji with the same number of characters are arranged and stored in a predetermined order, and a kanji with the same number of characters for each kanji that corresponds to the pronunciation of the kanji for the first name. at least a famous word section (5) in which names are arranged and stored in a predetermined order, and with respect to a group of candidate names corresponding to the names extracted by the character recognition device, information on each of the group of candidate names is provided. Based on reading and/or kanji,
A name dictionary for post-processing of character recognition, characterized by being able to extract corresponding kanji and/or pronunciations. Claim 2: The surname pronunciation part (2), the given name pronunciation part (3), the surname word part (4), and the given name part (5).
2. The name dictionary for character recognition post-processing according to claim 1, further comprising a general reading section (6) for general words other than surnames and/or given names, and a general word section (7). .