JPH0338765A - Method and device for processing character - Google Patents

Abstract

PURPOSE:To detect the swing of the KATAKANA (square form of Japanese syllabary) description included in a compound word by dividing a KATAKANA character string consisting of plural KATAKANA words into those KATAKANA words stored in a KATAKANA description dictionary and processing each of these KATAKANA words. CONSTITUTION:A KATAKANA character string extracting part 12 extracts the KATAKANA character strings out of the sentences, and a KATAKANA compound word deciding part 14 decides whether the KATAKANA character strings are equal to the compound words or not. A KATAKANA compound word dividing part 15 divides a KATAKANA character string decides as a compound word into each KATAKANA word based on the KATAKANA word description stored in a KATAKANA description dictionary 13. Then a KATAKANA character string deforming part 16 replaces or deletes partly a KATAKANA character string and deforms the partial character string. A swing candidate detecting part 18 compares the deformed results with each other and detects the same deformed results in a group. Then a description swing deciding part 19 compares the original descriptions of each group with each other to detect the swinging descriptions. Thus it is possible to detect the description swings of KATAKANA words which are used in the compound words.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention 1 relates to a character processing device and method for document processing.

In conventional technical Japanese texts, katakana is used to represent foreign words. However, the method of writing foreign words in katakana is not fixed, and it is common for a single foreign word to be written in multiple ways in one sentence. It is said that this hinders readability.In response to this problem, a katakana notation deviation detection device that automatically detects katakana notation deviations has been devised in recent years. (Japanese Unexamined Patent Publication No. 62-290965) Figure 2 is the constituent country of the conventional katakana notation fluctuation detection device.

1+, 2114 is a text storage unit and stores the input text. i&Ic memory as the text storage unit 21;
Magnetic disk devices and the like are used.

22 (Dai) This is a katakana sequence extraction unit, which extracts katakana sequences from the sentences stored in the sentence storage unit 21. 23
(Yo is the katakana string storage section, and the katakana string extraction section 22
Along with the katakana string extracted by
Information about the position in the sentence stored in No. 1 is also stored. As the katakana string storage section 23, an IC memory, a magnetic disk device, etc. are used (similar to the sentence storage section 21). Transformation is added to the katakana string by deleting or replacing katakana or partial katakana strings. 25 is a transformation result temporary storage section, and the result transformed by the katakana string transformation section 24 is stored in the katakana string storage section 23. This is a transformation result comparison section that detects a group of katakana sequences that match the transformation results stored in the transformation result temporary storage section 25. 27 (yo) The column comparison unit compares the katakana strings stored in the katakana string storage unit 23 with respect to a group of untransformed katakana strings that match the deformation results, and compares the katakana strings that match the deformation results but differ from the katakana strings before transformation. Detects a group of columns. 284;L This is a wobbling part display section and displays a group of katakana strings detected by the katakana string comparison section 27. As the wobbling part display section 28, 4i +; 1l11
．． A cRT display, liquid crystal display, etc. is used. There are several display methods, such as highlighting the detected katakana string or displaying it in color. Force exchange in which there are other constituent elements besides these These are omitted as they are not necessary for comparison with the present invention.

In the conventional katakana notation deviation detection device configured as above, first, a continuous katakana character string is extracted from a sentence, and then the possibility of mutual deviation, such as "hue" and "he", is extracted. After deforming a certain part in one direction, the deformation results are compared to detect fluctuations. For example, when "interface" and "ink face" are mixed in a sentence, delete or replace the katakana strings such as "fehe", "heihe", r-1 (delete), and write "interface". Transform the "interface" as follows. ``Interface'' → ``Inkheath'' → ``Inkheath'', ``Inkface'' → ``Interface'' → ``Interface'' → ``Interface''. This electrical connection “interface” and “interface”
Since the transformation results match, and the katakana strings before transformation are different, it is determined that the notation is distorted.

Problem to be Solved by the Invention When detecting deviations in katakana notation using a conventional katakana notation deviation detection device, when extracting katakana literary strings, it is difficult to extract katakana character strings that are continuous in a sentence. So, as one katakana word, there is also a compound word such as ``man-machine interface'' and ``interface'' in one sentence, and ``interface'' is included in the compound word. If the notation in the ``Katakana'' part is deviated, it cannot be detected as a ``deviation in the notation''. When a character string has multiple katakana words connected to form a single katakana character string, by dividing the string and processing it as each katakana word, you can separate the katakana words used in the compound word and the katakana words used in the compound word. An object of the present invention is to provide a character processing device and method capable of detecting variations in the notation of Katakana words used alone.

Means for Solving the Problems (1) A sentence storage unit that stores sentences, a katakana string extraction unit that extracts katakana character strings from the sentences stored in the sentence storage unit, and a katakana string extraction unit that extracts at least katakana character strings from the sentences stored in the sentence storage unit. A katakana compound word determination unit that determines whether a katakana character string is a katakana compound word by comparing it with one or more katakana notation dictionaries stored in the katakana notation dictionary. and a katakana compound word division unit that divides the katakana character string determined to be a katakana compound word by the katakana compound word determination unit into words by katakana words stored in the katakana notation dictionary;
a katakana string transformation section that rewrites a partial katakana string in a katakana string; a transformation result temporary storage section that temporarily stores the transformation result transformed by the katakana string transformation section in association with the original katakana character string; A shake candidate detection unit that compares deformation results temporarily stored in a storage unit and detects a match, and some of the shake candidates detected by the shake candidate detection unit have different original katakana notation. This is a character processing device that includes a notation fluctuation determination unit that determines whether or not.

(2) Have a katakana notation dictionary that stores at least one katakana notation, and a katakana string extraction stage that extracts katakana character strings from a sentence, and whether the katakana character string is a compound word of katakana. a katakana compound word determination step in which the katakana character string determined to be a katakana compound word in the katakana compound word determination step is determined by comparing it with the katakana words stored in the katakana notation dictionary; A katakana compound word division stage in which each katakana word is divided into l words based on the katakana words stored in the notation dictionary, a katakana string transformation stage in which partial katakana strings in a katakana character string are rewritten, and the transformation results are compared to determine which ones match. A character processing method comprising: a distortion candidate detection step for detecting a distortion candidate; and a writing variation determination step for determining whether any of the distortion candidates detected in the distortion candidate detection step has a different original katakana notation. be.

Operation The present invention has the above-mentioned configuration, so that the force exchange of the single force Takana sequence extraction part is
Extracts katakana character strings from sentences and determines katakana compound words? ! Determine whether a katakana string is a compound word. The katakana compound word division unit divides the katakana character string determined to be a compound word into each katakana word using the katakana notation stored in the katakana notation dictionary. Next is the katakana sequence transformation part? L Transforms a katakana string by replacing or deleting a substring. Furthermore, the distortion candidate detection section (Y) compares the deformation results and detects the same ones as groups, and the notation variation determination section (Y) compares the original notation of each group and detects those whose notation is distorted. .

Example Embodiments of the present invention will be described below with reference to the drawings.

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

In FIG. 1, numeral 11 is a text storage section, which stores input text. 12 (Y) This is a katakana string extraction section, which sequentially extracts katakana character strings one by one from the sentences stored in the sentence storage section 11.

Reference numeral 13 is a katakana notation dictionary, which stores only katakana words that are not compound words and do not have any variations in their notation. 1
4 (Yo) is a katakana compound word determination unit that extracts katakana character strings and katakana notation dictionary 1 by katakana string extraction unit 12.
In step 3, it is determined whether the extracted katakana character string is a compound word by comparing the katakana notation stored in the memory. 1
5 i:L is a katakana compound word dividing unit, which compares the katakana string that the katakana compound word determining unit 14 has determined to be a compound word with the katakana notation stored in the katakana notation dictionary 13, and then converts it into l words. Divide into katakana words 4161; L katakana string transformation unit, which rewrites the partial katakana strings of the extracted or divided katakana words. Reference numeral 17 denotes a transformation result temporary storage section, which temporarily stores all the katakana strings rewritten by the katakana string transformation section 16 in the order in which they were extracted by the katakana string extraction section 12, in association with the original katakana character strings. 1
8 [Yo Shake candidate detection unit, Katakana sequence extraction unit 12
When an instruction to finish extracting a katakana string is given from
The transformation results temporarily stored in the transformation result temporary storage section 17 are compared, and matching results are detected as a group.
Reference numeral 19 denotes a notation deviation determination unit, and a deviation candidate detection unit 18
It is determined whether or not any of the groups of shake candidates detected in the above have different original notations. 20 (This is a wobbling part display section, and displays the katakana character strings of groups for which the notation wobbling judgment section 19 has determined that there are different notations. There is a notation method (
How to display a list of wobbling groups Negative There are methods such as placing 1 in a sentence and displaying the wobbling part in reverse video or underlining it to display it differently from other parts.

FIG. 3 is a flow diagram illustrating a character processing method according to an embodiment of the present invention.

31 & 1 Katakana string extraction stage, sentence storage section 1
Extract the katakana character string from the memorized sentence in step 1. 32 (Yo) This is the katakana compound word determination stage, and it is determined whether the katakana character string extracted in 31 is a compound word. If it is a compound word, proceed to step 33. If it is not a compound word, proceed to step 34. Shinpay 33 Cl This is the katakana compound word division stage, in which compound words are divided into words one by one according to the katakana notation stored in the katakana notation dictionary 13. Katakana notation 1 or 33 determined not to be a word
Rewrite the partial katakana string of the katakana string divided by according to certain rules. The rewritten transformation result is temporarily stored in the transformation result temporary storage section 17. 35 (above) Determine whether all the katakana character strings have been extracted from the text, and if there are still katakana character strings remaining, go back to step 31. If all the katakana strings have been extracted, proceed to step 36. This is the shaking candidate detection stage.
Compare the deformation results temporarily stored in the deformation result temporary storage section 17 and detect the same results as a group &37i
In this stage, the original notation of each group of katakana sequences detected in step 36 is compared, and if there is a different notation, it is determined that the notation is deviated.

The operation of the character processing device and method of this embodiment configured as described above will be specifically explained below using an example.

FIG. 4 is an example of a sentence stored in the text storage unit 11. In this sentence, the words ``wine maker'' and ``manufacturer'' are written interchangeably.

FIG. 5 1 is an example of katakana notation stored in the katakana notation dictionary 12.

FIG. 6 shows an example of transformation rules used when the Katakana string transformation unit 16 performs transformation.

First, the katakana character string 13- is sequentially extracted from the sentences stored in the katakana string extraction unit 12 (upper sentence storage unit 11). “Winemaker” is extracted first. Then, katakana compound word judgment is performed. Part 14 (Determine whether or not the Katakana character string extracted by the friend Katakana string extraction part 12 includes the Katakana notation stored in the Katakana notation dictionary 13. If so, perform Katakana compound word division. Part 15 (Yo) The katakana character string extracted by the katakana extraction unit 12 is divided according to the katakana notation stored in the katakana notation dictionary. In the case of "wine maker", "wine" stored in the katakana notation dictionary is Since it contains "wine", it is divided into two parts: "wine" and the remaining part "manufacturer".When the katakana character string extracted by the katakana extraction unit 12 matches any of the katakana words stored in the katakana notation dictionary , or if none of the substrings match the katakana words stored in the katakana notation dictionary, it is determined that the extracted katakana string is not a compound word.

Next, according to the transformation rules shown in FIG. Delete or replace columns.

If you transform according to this example, "Wine" → "Wai"
, "Manufacturer" → "Mecha". Sara? Q-variant transformation result occasional storage unit 1'Ni For the transformation result "Wai", "4" indicating the beginning position in the sentence and "3" indicating the length of the katakana character string are stored in association with each other and "mecha" is stored. For “
7" and "3" are stored in association with each other. When the transformation processing for "winemaker" is completed, the katakana string extraction section 12i, t extracts the next katakana character string from the sentences stored in the sentence storage section 11, and the katakana string transformation section 16 does the same. Add the transformation. "Maker" (transformed into "mecha") extracted second.Hereafter, text storage part 1
When similar processing is performed on all character strings of the sentences stored in 1, the results stored in the temporary transformation result storage section 17 will be as follows. The numbers in [ ] are the serial numbers of the katakana string, and the numbers in 0 are the position information of the katakana string, and the number of characters counted from the beginning of the sentence and the length of the katakana string are 5 - [1 Scary (4, 3) [2] Mecha
(7,3) [3] Mecha ('15.4) [4] Wai
(26, 3) For these, the shaking candidate detection unit 18 detects those whose deformation results match. In this case, +1 [1] and [4], "2" and "3" are detected. Furthermore, the notation deviation determination unit 191t compares the original notations based on the position information. Since the original notations of "2" and "3" are different, it is determined that the notations of these are mixed. "2" and "3" on the shaking part display section 20
Displays the katakana string corresponding to , as if the notation is distorted.

As described above, according to this embodiment, the katakana notation dictionary, the katakana compound word determination section, and the katakana compound word division section are used to
"Winemaker" is a combination of two katakana words that form one katakana character string, "wine" and "maker".
By dividing the word into katakana words and processing them as individual katakana words, we can distinguish between the spelling variations such as ``maker'' included in the compound word ``6-winemaker'' and ``maker j'' used alone. In addition, by limiting the katakana words stored in the katakana notation dictionary to katakana words that are not compound words and have no variations in spelling, the dictionary size can be reduced.

In other words, if you memorize all katakana words,
It is clear that if we memorize all possible notations for words that have variations in notation, the amount of katakana notations to be memorized will be enormous.

In this example, we will store all katakana strings in the katakana notation dictionary only those that do not cause any fluctuations in the orthography. The notation may also be memorized. Zara&he The header of the katakana notation dictionary and the transformation result of the process performed by the katakana sequence transformation unit on the heading are stored as a pair, and the transformation performed by the katakana sequence transformation unit is omitted. (1 7- Effect of the Invention In the present invention, it is a combination of multiple katakana words.
When a compound word has two katakana character strings, it is divided into katakana words stored in the katakana notation dictionary and processed as individual katakana words to detect variations in katakana notation contained in compound words. can be used, and its practical effect is

[Brief explanation of drawings]

Figure 1 shows the configuration of a character processing device according to an embodiment of the present invention; Figure 2 shows the configuration of a conventional character processing device @ Figure 3 shows the flowchart of a character processing method according to an embodiment of the present invention. Figure 5 is an explanation of examples of sentences stored in the sentence storage section. Figure 5 is an explanation of examples of katakana notation stored in the katakana notation dictionary. Figure 6 is an example of transformation rules used when performing transformations in the katakana string transformation section. FIG. 11... Sentence storage section 12... Katakana string extraction enemy 1
3...Katakana notation dictionary, 14...Katakana compound word judgment R, 15...Katakana compound word division enemy 16.8
- Katakana sequence transformation question 17... Temporary storage of transformation results 18... Shake candidate detection paper 19... Shake judgment of notation 20... Shake part display

Claims (2)

[Claims] (1) A sentence storage unit that stores sentences; a katakana string extraction unit that extracts katakana character strings from the sentences stored in the sentence storage unit; and at least one katakana notation. a katakana notation dictionary; a katakana compound word determination unit that determines whether a katakana character string is a katakana compound word by comparing it with katakana words stored in the katakana notation dictionary; and the katakana compound word determination unit. a katakana compound word dividing unit that divides the katakana character string determined to be a katakana compound word by the katakana word division unit into words by katakana words stored in the katakana notation dictionary;
a katakana string transformation section that rewrites a partial katakana string in a katakana string; a transformation result temporary storage section that temporarily stores the transformation result transformed by the katakana string transformation section in association with the original katakana character string; A shake candidate detection unit that compares deformation results temporarily stored in a storage unit and detects a match, and some of the shake candidates detected by the shake candidate detection unit have different original katakana notation. 1. A character processing device, comprising: a notation deviation determining unit that determines whether or not the notation is correct. (2) A katakana string extraction stage that includes a katakana notation dictionary that stores at least one katakana notation, and extracts katakana character strings from a sentence, and the katakana string is a katakana compound word. a katakana compound word determination step in which it is determined whether or not the katakana compound word is a katakana compound word by comparing it with katakana words stored in the katakana notation dictionary; The katakana compound word division stage divides the katakana words stored in the katakana notation dictionary into individual words, the katakana string transformation stage rewrites partial katakana strings in the katakana character string, and the transformation results are compared to find a match. The present invention is characterized by comprising a shaking candidate detection step of detecting an object, and a notation deviation determining step of determining whether or not any of the shaking candidates detected in the shaking candidate detection step has a different original katakana notation. Character processing method.