JPH0721177A - Character input device - Google Patents

Abstract

PURPOSE:To smoothly input a KANJI (Chinese character) notation word string by automating the selection of a homonym by applying character recognition technology when there is a handwritten or type-printed document to be inputted. CONSTITUTION:A character recognizing means 101 convert entry character string image data into >=1 group of character codes and a KANA (Japanese syllabary)-KANJI (Chinese character) converting means 102 converts a character code string of HIRAGANA (cursive form of KANA), etc., inputted corresponding to them into KANJI notation word string candidates. A collating means 103 performs collation to decide the order of the KANJI notation word string candidates according to the degree of matching with character code candidates, and an input/output control means 104 displays the character image data and the KANJI notation word string whose order is decided and a selects and a determine a KANJI notation word string according to an external indication.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for inputting a printed or handwritten character string to a computer system or the like, and more particularly to converting a character code string input by hiragana, katakana or romaji into a handwritten character shape. The present invention relates to a character input device for converting and inputting into a corresponding kanji character string.

[0002]

2. Description of the Related Art With the spread of word processors, Japanese front-end programs (FEP), etc., it has become common to input Japanese character strings into computer systems and the like.
In order to input Japanese characters in Kanji, these Japanese input means must have a so-called Kana-Kanji conversion function that converts a character string in Hiragana or Katakana input from a keyboard into a word string in Kanji notation. This kana-kanji conversion function eliminates the need for many keys, such as the Japanese type, and makes it easier to enter kanji words.
In this kana-kanji conversion, the main function is to search the word dictionary by reading the words entered in hiragana or katakana and display the results on the display. In addition, in the recent kana-kanji conversion function, the function of learning the kanji notation word selected by the input person has become popular.

[0003]

When a kana-kanji conversion function is used to input a word string in kanji notation from reading a word, the input efficiency depends on how to select homonyms. The conventional kana-kanji conversion function is based on the method in which the input person selects a desired word from multiple homonyms, and the time required to select this homonym is the input time for reading hiragana or katakana. There was a problem that the working time was not negligible. Furthermore, the task of selecting homonyms requires knowledge of the correct kanji notation, and also requires a great deal of nerve to select confusing kanji notations, which hinders smooth Japanese input. there were.

As described above, it is generally known that the conventional kana-kanji conversion function has a word learning function. However, in this learning function, the homonyms selected immediately before by the input person are selected. It is not enough to correctly select the homonyms in various scenes of Japanese input, because it memorizes mechanically. Attempts have also been made to automate the determination of homonyms using Japanese analysis techniques that use grammatical information, etc. However, regarding the selection of homonyms such as noun words that are synonymous in Japanese analysis, There was a problem that it was insufficient and had to depend on the input person's selection.

The present invention has been made in view of the above circumstances, and it is an object of the present invention to solve the above-mentioned problems in the prior art when there is a manuscript to be input which is handwritten or printed. By applying character recognition technology and automating the selection of homonyms, it is to provide a character input device that enables smooth input of kanji word strings.

[0006]

The above-mentioned objects of the present invention are as follows.
A character recognition unit converts the sequentially input character string image data into one or more sets of character codes, and the character codes of hiragana, katakana, or romaji are input corresponding to the entered character string image data. For the columns, the kana-kanji conversion means performs conversion processing into one or more sets of kanji notation word string candidates, and the matching means compares the kanji notation word string candidates with the character code candidates to obtain kanji notation word string candidates. The characters are ordered according to the degree of matching with the character code candidates, the character image data is displayed by the input / output control means, and the ordered kanji written word string is selected and the kanji written word string is selected and confirmed by an external instruction. Achieved by a character input device. In addition,
The kanji notation word string referred to here may be a character string containing only one word or a character string containing a plurality of words.

[0007]

With the character input device according to the present invention, in key input of an original handwritten or printed, it is possible to automatically select a homonym that matches the character shape on the original, and the key input person can use a homonym. It is possible to realize smooth Japanese input without being bothered by selecting.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a functional block configuration diagram for realizing the character input device of the present invention. In FIG. 1, 101 is a character recognition means for converting character image data into a character code, and 102.
Is a kana-kanji conversion means for converting a hiragana, katakana, or roman character string into a kanji notation word string. Reference numeral 103 is a collating means for collating a character code with a Kanji written word string, 104
Is an input / output control means for inputting a display of a processing result, selection of a Kanji written word string, and a confirmation instruction. Reference numeral 105 is character string image data on the manuscript, 106 is character code data input by a key, 107 is a character code candidate as a character recognition result, 108 is a kanji written word string candidate, 109 is a collation result character string image on the manuscript Kanji writing word string candidates ranked in order of closeness to the shape, 110 is display data (above Kanji writing word string candidates and character image data), 11
Reference numeral 1 indicates key input data from the operator, and reference numeral 112 indicates a kanji written word string which is the final output.

The operation of the functional block configuration shown in FIG. 1 will be briefly described below. First, in the character recognition means 101,
Character string image data on the manuscript input from outside 10
Character code is executed for 5 and character code (JIS or shift JIS code) corresponding to image data of each character
Is output together with the character score (the degree indicating the character code likelihood). Here, various means can be used to acquire the character string image data, such as a scanner and a fax.
Is generally used as an input device. Specifically, character recognition is performed by preprocessing for normalizing the character size, processing for extracting character features, and input character patterns and registered character patterns in the character dictionary using the extracted character features. It is performed by four steps of a process of calculating a distance value (corresponding to the character score) and finally a process of converting to a corresponding character code based on the distance value. The operation of the character recognition means 101 will be described later in detail with reference to FIG.

On the other hand, the kana-kanji conversion means 102 converts a character string consisting of character code data 106 of hiragana, katakana or romaji input by a key operation of the input person into a word string in kanji notation by searching a word dictionary. And output it. In this conversion process, all homonyms with the same reading but different Kanji notation are output as a plurality of candidates. The operation of the kana-kanji conversion means 102 will be described later in detail with reference to FIG.

Next, the collating means 103 matches the character code candidate 107, which is the character recognition result, with the kanji notation word string candidate 108, which is the kana-kanji conversion result, and determines the degree of matching between both results. According to the above, the kanji written word strings are ranked, and these are output as candidates 109. The operation of the matching means 103 will be described in detail later with reference to FIGS. 4 and 5.

Finally, in the input / output control means 104, first, as the display data 110, the above-mentioned kanji word sequence candidate 1
09 is output to the outside (display) together with the character image data according to the order. The operator (operator) of the hiragana, katakana, or romaji character code confirms whether or not the displayed character string image data is correctly converted into the kanji notation word string. In some cases, the correct word is selected and the result is transmitted to the input / output control means 104. Input / output control means 1
04 outputs the Chinese character writing word string 112 confirmed by the instruction from the operator with the key input data 111.

Next, the operation example of the character recognition means 101 which can be adopted will be described in detail with reference to FIG. 201 in FIG.
Is a preprocessing unit for normalizing the size of character image data, 202 is a feature extraction unit for extracting features used for character recognition from the character image data, and 203 is a standard character pattern registered in advance in the above features and character information. , A distance calculation unit for calculating the distance to the feature
A code conversion unit 5 obtains a character code corresponding to a standard pattern having a relatively small value as a result of distance calculation. Reference numeral 206 denotes input character image data after preprocessing, 20
Reference numeral 7 represents the extracted character feature, 208 represents the character feature of the standard pattern from the character dictionary, and 209 represents the distance calculation result between the character feature of the input character image data and the character feature of the standard pattern.

The preprocessing unit 201 mainly performs noise removal and normalization processing on the input character image data. Here, the noise removal is a process of removing fine black pixels added due to stains on the paper surface. In addition, the normalization process is a process aimed at matching the conditions of subsequent feature extraction for the input image data with the extraction conditions of the character features of the standard pattern to be compared. It is a process of enlarging or reducing to.

The feature extraction unit 202 extracts the features required for character recognition from the preprocessed input character image data 206. Various types of character features have been proposed here, but for example, Hagita et al., “Identification of Kanji Characters by Contribution Features in the Outer Direction” (IEICE Transactions, Vol.J66-D, No.
It is possible to use the contour direction contribution feature shown in 10, 1983).

The distance calculation unit 203 includes character features 207 of the input character image data obtained by the above operation,
The distance calculation is performed with the character feature 208 of the standard pattern registered in advance as the character dictionary 204, and which standard pattern in the character dictionary 204 the shape of the input character image data is close to is identified.

The code conversion unit 205 calculates the distance calculation result 20.
Based on 9, the character code corresponding to the standard pattern identified as having a relatively close shape is obtained and output. Here, a JIS code, a shift JIS code, or the like can be used as the character code.

Next, referring to FIG. 3, kana-kanji conversion means 1
The operation of No. 02 will be described in detail. FIG. 3 shows an operation example in the case where the reading "shiri" is given as an input.
In FIG. 3, 301 is a search key generation unit that generates a search key for a later word dictionary search, 302 is a word search unit that executes a word dictionary search, 303 is a word dictionary, and 304 and 305 are search keys.

The search key generation unit 301 generates a search key for a word dictionary that is extended by one character from the beginning to the end of the input hiragana or katakana character code string. For example, when the hiragana character code string “shiri” is input, “shi”, “shiri”, “shiri”, and “shiri” are generated. In FIG. 3, for simplification of description, a case where only “shiri” is generated as the search key is shown.

Next, the word search unit 302 searches the word dictionary 303 using the search key. In the example of Figure 3,
From the search key "shiryo", a total of 6 kanji words (same synonyms) of "historical material", "dead spirit", "thinking", "sample", "material", and "feed" can be obtained as search results. . The kana-kanji conversion means 102 described with reference to FIG. 3 is a method that is generally adopted in a Japanese word processor and the like, and various methods other than those shown here are conceivable. Can be used for the character input device.

FIG. 4 is a functional block configuration diagram showing an embodiment of the collating means 103. In FIG. 4, 401 is a character candidate table that temporarily stores the character code that is the result of character recognition together with the character score, 402 is a word score calculation unit that calculates the score of the Kanji notation word (word string) candidate based on the character score, 40
Reference numeral 3 is a word candidate table for temporarily storing the kanji written word candidates as the kana-kanji conversion result and the word score thereof, and 404 is an alignment unit for performing sorting based on the score of the kanji written word candidate. The detailed operation of the matching unit 103 having the functional block configuration shown in FIG. 4 will be described with reference to FIG.

FIG. 5 is a drawing for explaining the operation of the collating means 103. 5, 401 and 402 are the character candidate table and the word candidate table shown in FIG. The character candidate table 401 stores character code candidates that are the result of character recognition and the character score of each character code candidate. In FIG. 5, as the character recognition result of the character image data “material”, the first character is “wise (9 points)”, “material (8 points)”, “relative (6 points)”, and the second character is "Fee (10
“Points” ”,“ Departments (8 points) ”, and“ Villages (5 points) ”are stored. Further, the word candidate table 403 stores kanji notation word candidates that are kana-kanji conversion results. FIG. 5 shows an example in which a total of six word candidates of “historical material”, “dead spirit”, “thought”, “sample”, “material”, and “feed” are stored.

In the word score calculation, each word candidate in the word candidate table 403 is converted into a character candidate table 401.
When a character code candidate in the middle is included, it is added as a word score of a word candidate including the character score, and stored in the word score area of the word candidate table 403. In the example of FIG. 5, the word score of each word candidate is only 10 points in which “historical material” is “material”,
0 points because "dead spirit" and "thought" do not include character code candidates at all, "sample" only has 10 points of "charge", "data" has 8 points of "material" and 10 points of "charge" 18 points, "feed"
Is only 10 points of "fee". In the sorting processing by the sorting unit 404, the word candidates in the word candidate table 403 are sorted in order from the word candidate with the highest score. As a result, in the example of FIG. 5, the first place is “data”, the second place is “historical material”, “sample”, and “feed”.

In order to simplify the explanation, the example in which the conversion target is one word has been described. However, even in the case of a word string including a plurality of words, ranking is performed by the score obtained by adding the word scores, and similarly. It goes without saying that the present invention can be implemented.

[0025]

As described above, according to the present invention,
Input of handwritten or printed characters When inputting a character string on the manuscript,
By collating the hiragana, katakana, or romaji entered by the key with the character recognition result for the character string image on the manuscript, it is possible to automatically select the homonyms that match the character shape on the manuscript, and the key input person This has the remarkable effect of being able to realize smooth Japanese input without being bothered by the selection of homonyms.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating an embodiment of a character input device according to the present invention.

FIG. 2 is a diagram illustrating an operation of a character recognition unit.

FIG. 3 is a diagram illustrating an operation of a kana-kanji conversion unit.

FIG. 4 is a functional block configuration diagram showing an embodiment of a matching unit.

FIG. 5 is a diagram illustrating an operation of a matching unit.

[Explanation of symbols]

 101 Character Recognition Means 102 Kana-Kanji Conversion Means 103 Collating Means 104 Input / Output Control Means 105 Character String Image Data 106 Character Code Data 107 Character Code Candidates 108 Kanji Notation Word String Candidates 109 Ranked Kanji Notation Word String Candidates 110 Display Data 111 Key input data 112 Kanji notation word string

Claims (1)

[Claims] 1. A character recognizing means for converting character image data into character code candidates, and a kana kanji character for converting a character code of hiragana, katakana or romaji inputted corresponding to the character image data into kanji notation word string candidates. A degree of matching between the conversion means, the character code candidates obtained by the character recognition means and the kanji notation word string candidates obtained by the kana-kanji conversion means, and matching the kanji notation word string candidates with the character code candidates The collating means for performing ordering according to the above, the character image data and the ordered kanji notation word string candidates are displayed, and a kanji notation word string is selected from the kanji notation word string candidates by an external instruction to confirm. A character input device comprising: output control means.