JPH0248938B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a device for supporting detection and correction of erroneous characters included in character example data in a Japanese information processing system.

Conventionally, language processing systems commercialized as word processors, for example, use a dictionary that stores words as spelling strings, and store strings of input data (sentences, words, etc.) in the dictionary. Detects errors in the input string by comparing and collating it with the string of words in the input string, and if the input string contains an error, replaces it with the characters (sequence) of the word stored in the dictionary. It was designed to correct errors. For example, in the case of an English word processor, a character string that is inputted into words is checked against words in a dictionary character by character, and the word with the largest number of matching characters, including the order of the characters, is selected. Erroneous characters are corrected by treating them as input words and replacing them with characters from the corresponding word in the dictionary, treating unmatched characters as errors. However, when targeting Japanese sentences, it is difficult to identify incorrect characters because there are many types of kanji, and many words are composed of two kanji (in the case of two-character words, which one is wrong? Errors contained in data can be automatically removed due to various (computer processing) problems that Japanese itself has, such as the difficulty in separating words because sentences are not separated. It was difficult to detect and correct it, and we had to check it by acquiring it.

In order to eliminate the above-mentioned conventional drawbacks, the present invention detects error character candidates based on the connectivity between characters in a character string of a Japanese sentence, and connects characters similar to the error character candidate with the previous and/or subsequent characters. It is designed to output connectable characters as correction character candidates, and its purpose is to improve the efficiency of detecting and correcting incorrect characters in Japanese sentences. The drawings will be described in detail below.

FIG. 1 shows an embodiment of the present invention, in which 1 is an input section, 2 is a register that temporarily stores Japanese data for two characters, and 3 is a buffer that temporarily stores data for two characters. 3a is a control unit, 4 is a collation circuit, and 5 is a storage unit that stores a character concatenation dictionary containing information on whether or not two consecutive characters in a Japanese sentence can be connected (hereinafter simply referred to as a character concatenation dictionary for the sake of brevity). ), 6 is an output section, 7 is an error code generation circuit 7-
1. A code generation circuit consisting of a candidate start code generation circuit 7-2 and a candidate end code generation circuit 7-3; 8 a search circuit; 9 a character classification of characters similar to each character according to its characteristics; This is a storage unit that stores a classification dictionary (hereinafter simply referred to as a character classification dictionary for the sake of brevity). The register 2 is composed of a shift register having a data storage area for two characters. Under the control of the control unit 3, the register 2 temporarily stores the data of the characters input from the input unit 1, and outputs the data for each character to the output unit 6.
It's starting to look like it's being sent out.

The contents of the character concatenation dictionary 5 are, for example, as shown in FIG. A value of '1' is given to the intersection of two characters that make up a conjunction that exists as a word, and a value of '0' is given to the intersection of two letters that make up a conjunction that does not exist as a Japanese word. Here, instead of the value "1" mentioned above, it is also possible to use a value that indicates the frequency of use of the word.Although the drawing only shows combinations of kanji, in reality It also includes whether or not kana and kanji can be connected, as well as kana and kana.

The error code generation circuit 7-1, candidate start code generation circuit 7-2, and candidate end code generation circuit 7-3 are driven by the control unit 3, and generate error code "/" (slush) and candidate start code "[" (), respectively. The starting bracket) and the candidate ending code "]" (closing bracket) are output to the output unit 6. In addition, the error code is
A 2-byte code uniquely defined by the system may be used in the empty area of the JIS-C6226 code table.

The contents of the character classification dictionary 9 are, for example, as shown in Figure 3, the kanji in the heading are arranged in the order of JIS-C6226 codes, and corresponding to each kanji, the kanji that have the same reading as that kanji are classified and arranged. As shown in Figure 4, arrange the heading hiragana in alphabetical order, and write the hiragana and ``shape'' corresponding to each hiragana.
A combination of a classification and arrangement of similar hiragana and a further arrangement of similar "shapes" such as Katakana and Alphabetical (not shown), or a heading as shown in Figure 5. This is a combination of the classification and arrangement of kanji that are similar in shape to the kanji corresponding to the kanji, and the kanji shown in Figure 4 above, or a combination of all of these.

Next, the operation will be explained. First, the first and second data of the mixed Japanese character string data including kanji, kana, punctuation marks, etc. inputted from the input unit 1 are taken into the register 2 as the first and second characters, respectively. The control section 3 writes the two-character data in the register 2 as is into the buffer 3a, and further sends it to the collation circuit 4. The matching circuit 4 is the register 2.
The character data of the row and column of the character concatenation dictionary 5 are accessed for the data of the first character and the second character, and the value of the intersection is sent to the control unit 3. The matching circuit 4
If the value from is "1" (that is, when the two characters in register 2 are correctly concatenated as Japanese), the control unit 3 sends the data of the first character in register 2 to the output unit 6 as is. Then, the second character is moved to the area where the first character was stored, and the next character is taken into the register 2 from the input section 1. Thereafter, the data of the new first and second characters in the register 2 are written into the buffer 3a, and character concatenation is verified in the same manner as described above.

If the value from the matching circuit 4 is "0" (that is, when the two characters in the register 2 are not concatenated as Japanese characters), the control unit 3 outputs the error code from the error code generating circuit 7-
1 outputs the error code "/" to the output section 6, furthermore, the data of the first character in the register 2 is sent to the output section 6, and the next character is taken in from the input section 1. However, here, the contents of the buffer 3a are left unchanged. Next, the control unit 3 starts the candidate start code generation circuit 7-2.
output the candidate start code “[” to the output unit 6,
Search circuit 8 for data of the first character in buffer 3a
send to The search circuit 8 searches the character classification dictionary 9 using the data of the first character as a heading, and sends similar corresponding characters to the control section 3. The control unit 3 sequentially converts the similar characters into one character and sends data of two characters combined with the second character in the buffer 3a to the matching circuit 4. The matching circuit 4 accesses the character concatenation dictionary 5, checks whether or not each of the two characters can be connected,
A value of “0” or “1” is output to the control unit 3.
Based on the value from the matching circuit 4, the control unit 3 sequentially sends to the output unit 6 only the characters that can exist immediately before the second character in the buffer 3a among the similar characters, and then outputs the candidate end code. The candidate end code "]" is sent from the generating circuit 7-3 to the output unit 6.

Next, the control section 3 uses the candidate start code generation section circuit 7-2.
causes the output unit 6 to send out the candidate start code "[",
Search circuit 8 for data of second character in buffer 3a
, and the character classification dictionary 9 is searched using the second character as a heading. The search circuit 8 sends characters similar to the second character in the buffer 3a from the character classification dictionary 9 to the control unit 3, and the control unit 3 sequentially selects the characters as two characters and assigns them to the first character in the buffer 3a. The data of the combined two characters is sent to the matching circuit 4, and the character concatenation dictionary 5 checks whether or not the characters can be connected. According to the result of the check, the control section 3 controls the buffer 3a.
Only the characters that can be connected immediately after the first character in the characters are sequentially sent to the output unit 6, and then the candidate ending code “]” is sent to the output unit 6 from the candidate ending code generation circuit 7-3.

Next, the control section 3 newly writes data for the next two characters in the register 2 into the buffer 3a, performs character concatenation verification in the same manner as described above, and repeats this process.

FIG. 6 is a flowchart showing the process flow explained in the above embodiment.

FIG. 7 shows an example of an input character string and an output character string in the above embodiment, and in the figure ai, aij (i, j
≧1) represents each character of a Japanese sentence. Here, a case is shown in which the second character a ₂ in the input character string is an error character (that is, the character combinations a ₁ a ₂ and a ₂ a ₃ do not exist in Japanese). (The underline in a ₂ is added to indicate that it is an incorrect character, and is not written in the actual string.) In the output string, [a ₁₁ a ₁₂ ...] is similar to a ₁ . characters that can be connected with a ₂ after that, i.e. candidates for corrected characters that can be substituted for a ₁ when a ₂ is a correctly input character and a ₁ is an error character. It is a gathering. Also, [a ₂₁ a ₂₂ ...] is a character similar to a ₂ that can be connected with a ₁ before it, that is, a ₁
This is a set of candidates for corrected characters that can be substituted for a 2 when A ₂ is considered to be a correctly input character and a ₂ is an error character. Also, [a′ ₂₁ a′ ₂₂ …], [a ₃₁ a ₃₂ …]
The same applies to above, but [a′ ₂₁ a′ ₂₂ …] is a ₃
is a set of characters that can be connected, and is generally different from [a ₂₁ a ₂₂ ...]. Here, Figure 7 shows the case where the character conjunctions a ₁ a ₂ and a ₂ a ₃ do not exist in Japanese, but the character conjunctions that do not exist in Japanese are either a ₁ a ₂ or a ₂ a ₃ . It may be only one of them.

The error character candidates detected as described above and the correction character candidates to replace the error characters are generally determined to be correct or incorrect by human judgment, and the error characters are corrected. A display device, a correctness indicating device, a character designation device, a character input device, etc. (not shown) used for this purpose are arranged after the device of the present invention (in the embodiment, after the output section 6).

The contents of the character concatenation dictionary 5 are as described above, but when creating this dictionary, the target of Japanese sentences is, for example, "words" recorded in the Japanese dictionary.
By using this, it can be used as a dictionary of conditions for allowing or not to connect characters within a "word". Furthermore, in the case of creating a dictionary of conditions for connecting characters in a general sentence as a whole, the device of the present invention can be applied to create a dictionary from input character string data whose erroneous characters are to be detected and corrected. However, the range of words in the input data targeted when creating the character concatenation dictionary 5 is different from the range of words in the input data for which detection/correction support processing is actually performed (for example, legal terminology and medical terminology). (different), not all detected errors are errors.

In addition, if the part corresponding to the kanji in the character classification dictionary 9 is classified by ``yomi'' as shown in Figure 3, each kanji uses the ``yomi'' as a cue, as in the case of a tablet-type input device with a 50-syllabary arrangement. It is effective for detecting and correcting errors in character string data input using input devices such as kanji OCR. This method is effective for detecting and correcting errors in character string data input using an input device in which each kanji character is input based on its shape.

The error code, candidate start code, candidate end code, etc. are not all necessarily necessary, and they can be reduced. (For example, the error code "/" is for ease of explanation and may be omitted.) The accuracy of error character detection and correction is improved not only for 2 characters, but also for 3 characters, 4 characters, and so on. This can be done by checking for concatenation between characters over a longer range. This is effective for checking the connection between adjuncts, pronouns, formal nouns, etc. and adjuncts.

As explained above, according to the present invention, candidates for erroneous characters included in character string data of Japanese sentences are detected, and candidates for correction characters to be substituted for the characters are output. Data check of the Japanese text that I relied on (so-called verification)
It is possible to share a part of the work, increasing work efficiency and improving the detection of erroneous characters.
This has the advantage of increasing the accuracy of correction.

The present invention is a Japanese word processor, a kanji word processor, and a kanji word processor.
It can be applied to all so-called Japanese input devices, such as Japanese input devices using OCR and voice recognition, and can also be used to detect and correct erroneous characters in data that has been input once (for example, data stored in a database). Applicable.

[Brief explanation of the drawing]

The drawings serve to explain the present invention; FIG. 1 is a block diagram of an erroneous character detection/correction support device showing one embodiment of the present invention, and FIG. 2 is an explanatory diagram showing an example of the contents of a character concatenation dictionary. , FIG. 3, FIG. 4, and FIG. 5 are explanatory diagrams showing examples of the contents of a character classification dictionary,
FIG. 6 is a flowchart of processing in the apparatus of FIG. 1, and FIG. 7 is an explanatory diagram showing a line of input character strings and output character strings. DESCRIPTION OF SYMBOLS 1...Input part, 2...Register, 3...Control part, 4...Verification circuit, 5...Character concatenation dictionary, 6...
...Output section, 7... Code generation circuit, 8... Search circuit, 9... Character classification dictionary.

Claims (1)

[Claims] 1. A first means for temporarily storing two or more consecutive characters of a character string of a Japanese sentence, and a second means for storing a character concatenation dictionary whose content is whether or not at least two consecutive characters of a Japanese sentence can be connected; a third means for storing a character classification dictionary in which characters similar to each character are classified according to their characteristics;
a fourth means for comparing the contents of the means with the contents of the second means to detect an erroneous character candidate; and a fourth means for searching the third means for a character similar to the erroneous character candidate; An erroneous character detection/correction support device comprising a fifth means that selects and outputs only characters that can be connected to other characters of the first means by comparing the contents of the means.