JPS60164864A - Device for processing data - Google Patents

Abstract

PURPOSE:To correct errors of Japanese sentences with KANA (Japanese syllabary) KANJI (Chinese character), and homonyms which are used as faulse substitute characters by providing an inner-sentence from element analysis means and inner-sentence form element analysis means by the Roman letters in a processor. CONSTITUTION:For correcting misspellings in Japanese sentences with KANA and KANJI and homonyms which are used as faulse substitute characters, a character string to be examined is analysed in its every sentence through a sentence feeding out means 21 by an inner sentence form element analysis means 23. This is conducted with a vocabulary dictionary 30 and inter-vocabulary connection string 34 in a KANJI and KANA description. When the results fail, the device decides that the sentence contains errors of the faulse substitute character, by homonyms and a form element analysis is made by an inner-sentence form element analysis means 26 by the Roman letters. The resuslts of this are converted by a Roman letters - KANJI.KANA conversion means 27 into a KANJI.KANA combined sentence and are presented to a user. In this way, the provision and analysis of the means 23 and 26 enable correction of errors of the substitute characters in homonyms as well as errors of misspellings.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a text processing system for processing Japanese sentences mixed with kana and kanji, and particularly for detecting or correcting spelling errors in such a text processing system. It pertains to equipment that performs

[Background of the invention]

With the advancement of OA, text processing by computers has been attracting attention. Particularly in Europe and America, where the number of letters in the alphabet is small, computer processing of text has become quite widespread, and it is not only possible to simply process, memorize, and output text, but also to detect errors in the processing of input text.・
Correcting systems are being commercialized. For example, UN
TYPO, a utility program of IX, has a function to detect misspellings of English words, and DEClo
A program called 5PBLL has the ability to correct some spelling errors in English words. For example, J.L.Pete
rson-Computer prQgramfi
for Detection and Correcti
ng SpellngErrors”, CACM,
Vo 123, 492 introduces such an English spelling verification system in detail.

However, Japanese has three types of characters: (1) kanji, hiragana, and kanji.

(2) Kanji has an extremely large number of characters.

(3) It is difficult to cut out words because spaces are not inserted between words.

Because of these characteristics, it is difficult to develop text input methods and spelling detection/correction systems in Japan.

Furthermore, since the spelling error patterns in Japanese are different from those in Western languages, processing methods developed in Europe and the United States cannot be applied, and a method unique to Japanese is required.

The pattern of spelling errors in Japanese is also highly dependent on the input method. At present, the following Japanese input methods are known: (1) touch-type method, (2) pen-touch method, (3) kana-kanji conversion method, and (4) handwritten character recognition method (Japanese Patent Laid-Open No. 58-106663). , Japanese Patent Publication No. 5
5-143681).

On the other hand, possible spelling error patterns include (a) user's memory error (b) typographical error (C) device recognition error. Of these, in (a), there are many spelling errors using homophones as address characters, errors in okurikana, etc., which are often seen in the input methods of (1), (2), and (4). Among methods (3), in the method that allows phrase input, when the user inputs a phrase in kana, which is a phonetic character, the system converts it into a character string that is a mixture of kana, kanji, and kanji.
This type of error is less common than other methods. (b) is mainly an error specific to (1) and (2).

(C) is an error that is a problem only in method (4).

As is clear from the above considerations, the kana-kanji conversion method (3), which is currently widely used in Japanese word processors, is considered to be an input method that is relatively unlikely to cause errors, but it is easy to use for beginners, and the character recognition technology From the application of (
The present invention provides a device suitable for detecting and correcting errors of types (a) and (C), which is applied to the Japanese input method of the handwritten character recognition method in 4).

[Purpose of the invention]

The present invention is based on the above considerations, and the first
The purpose of this system is to provide a means for leaking misspellings contained in Japanese sentences in units of phrases in a text processing system that processes nine Japanese sentences containing kanji and kana. The second
The purpose of this is to provide a means for correcting misspellings that use homophones as spellings among the above-mentioned detected misspellings.

[Summary of the invention]

The present invention provides a means for dividing a Japanese sentence containing spelling errors (a combination of kanji and kana) into words by referring to a word dictionary containing 9 combinations of kanji and kana, and a means for presenting to a user the phrases for which word division has failed. , by referring to a conversion table that converts the phrases for which word division has failed from the kana-kanji conversion method to phonograms (for example, Romaji), and a word dictionary with phonograms to convert the resulting phrases in phonograms. The present invention is characterized by having means for dividing the phrase into ``Ero'' words, and, when the division is successful, means for converting the phonetic phrase into a kana-kanji conversion method and presenting it to the user as a corrected phrase. .

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described based on the drawings. FIG. 1 shows the external appearance of the document processing workstation used as this embodiment. The workstation 1 has a display 2. It has a keyboard 3, a tablet 4 for inputting handwritten characters, and a floppy storage device 6. The user inputs the Japanese text by hand from the keyboard 3 or by hand using the stylus pen 5 on the tablet 4. The input Japanese sentence is displayed on the display 2 or stored in the floppy storage device 6. Furthermore, the Japanese text once stored in the storage device 6 can be displayed on the display 2.

FIG. 2 is a block diagram showing details of the workstation of FIG. 1. Workstation is Disfray 2
, a display buffer memory 7, a keyboard 3, a tablet 4, a floppy storage device 6, a processor 8, a handwritten character LIA & device 9, and a printer IO.

The processor 8 is a device that performs various data processing such as processing input text, and may be realized by a microprocessor or a dedicated logic circuit, or may be realized by a host computer and a terminal device. The spelling error correction process of the present invention is executed by this processor.

The handwritten character recognition device 9 recognizes kanji and kana characters input from the tablet, codes them, and sends them to the processor 8.
It is a device that sends images to the computer, and a recognition rate of about 95 to 99 inches has been obtained.

The printer 10 is a device that is processed by the processor 8 and prints nine Japanese sentences.

In addition to storing input Japanese sentences, the floppy storage device 6 also stores dictionaries and tables used for analyzing Japanese. Note that an external storage device such as a disk storage device may be used instead of the floppy storage device.

FIG. 3 is a diagram showing an example of a display screen display when spelling errors are detected and corrected according to the present invention.

Japanese sentences input from the keyboard 2, tablet 4, or floppy storage device 6 are displayed in the upper half of the screen. When a spelling line detection command is input from the keyboard 3, the system analyzes the Japanese sentence phrase by phrase from the beginning, and if an incorrect phrase is found, it underlines it. In this example, the incorrect phrase "expert" is detected. In response, the system displays three menus in the bottom half of the screen. This allows the user to select 1 if the error detected by the system is a special term that is not in the system dictionary;
If the spelling corrected by the system is correct, you can select 2, and if you want to enter a new correct spelling yourself, you can select 3. Once the user has made his selection, the system resumes parsing the sentence with the next clause.

FIG. 4 shows the procedure by which the processor 8 and the floppy storage device 6 execute the spelling error detection and correction process as described above.

The character string to be tested is first stored in the character string buffer 2o, and then divided into phrases by the phrase cutting means 21. A clause is an independent word (generally begins with a kanji or katakana).
is followed by an arbitrary number of auxiliary words (hiragana), so if the point where the character type changes from hiragana to kanji or katakana is detected, it can be easily divided into phrases. Next, intra-clause morphological analysis 21 is performed for each extracted phrase. Morphological analysis is the process of dividing phrases into words, and is a concept handled in the kana-kanji conversion process of Japanese word processors. Here, since the text is a mixed sentence of kana and kanji, the analysis proceeds using the word dictionary 30 written in kanji and kana notation and the word connection matrix 34 (9). If morphological analysis is successful,
The phrase is judged to be fat, and control is transferred to phrase cutting means 21 to cut out the next phrase. If the morphological analysis fails, it is determined that the phrase contains a homonym-based typographical error, and the phrase is transferred to the Kanji-kana → Roman character conversion means 24 and the Kanji-kana → Roman character correspondence table 33.
is used to convert it into Roman characters and store it in the Roman characters buffer 25. Next, this Roman-ji phrase is subjected to morphological analysis using the Roman-ji morphological analysis means 26. This time, the morphological analysis is based on phonetic characters, so phrases that contain spelling errors, such as spelling errors but with the correct sound, are correctly analyzed.

Morphological analysis in this case uses a word dictionary 32 written in Roman letters and a word connection matrix 34. If the morphological analysis fails here, it means that the original character string contains an error other than a certain character, and the system notifies the user that there are 6 errors in the phrase. If the morphological analysis is successful, the correctly analyzed phrase is
Romaji (10) -> The kanji/kana conversion means 27 converts it into a sentence containing kanji/kana, stores it in the corrected character string buffer 28, and then presents it to the user. Conversion is performed by referring to a pointer 31 from a word in the word dictionary 32 written in Roman letters to a word in the word dictionary 30 written in Kanji/kana. Note that if a plurality of phrases containing kanji and kana characters correspond to the same phrase written in Roman characters, a plurality of candidates are presented to the user.

FIG. 5 is a diagram showing the flow of processing when the incorrect phrase "expert" is extracted. Clause buffer 20
When we morphologically analyze the phrase "expert" stored in , the analysis fails because "expert" is not in the dictionary. Therefore, when converting from kanji kana to romaji notation r8ENMONKANOJ using the Romaji conversion table 33, this time r
Since the noun 8ENMONKAJ is in the dictionary, the morphological analysis is successful, and the noun 1'-S ENMONKAJ
and the case particle ``NO'' that follows the noun. Finally, following the pointer 31 between the two dictionaries, these words are converted (11) into the corrected character string ``expert'' and presented to the user.

FIG. 6 is a diagram showing the flow of processing when the incorrect phrase ``shamejinai'' is extracted. In this case as well, morphological analysis based on the kanji/kana mixed notation fails because the academic term ``shajiru'' is not in the dictionary.

However, when converted to the Roman alphabet, the morphological analysis was successful, and the unformed form of the verb [f(AJIRU, l) and the ``NA''
It is divided into auxiliary verbs called IJ. Finally, this is converted into the phrase ``shame'' written in a mixture of kanji and kana, and the corrected phrase is presented to the user.

Although we have shown an example of Romaji as phonetic characters, Kana characters may also be used.

〔Effect of the invention〕

As described above, according to the present invention, first, spelling errors in Japanese sentences containing kana and kanji can be detected.

Errors include typographical errors, recognition errors in handwritten character recognition, and user memory errors. Second, among the above-mentioned errors, it has the effect of correcting spelling errors in which homonyms are mistakenly used as compensation.

[Brief explanation of the drawing]

(12) The drawings show one embodiment of the present invention; FIG. 1 is an external view of the system, FIG. 2 is a block diagram, FIG. 3 is a diagram showing an example of a display display, and FIG. A functional diagram showing the processor and the data flow within the floppy storage device. Figure 5 is an explanatory diagram of the processing when the incorrect character string ``expert'' is input. Figure 6 is ``embarrassing''. It is an explanatory diagram of processing when an incorrect character string is entered. 1... Workstation, 2... Display screen, 3... Keyboard, 4... Tablet,
5... Stylus pen, 6... Floppy storage device, 7... Display buffer memory, 8-Processor, 9
...Handwritten character recognition device, lO...Printer, 20.
... Character string buffer, 21 ... Bunsetsu extraction means, 22
... Kanji/kana mixed phrase buffer, 23... Morphological analysis means for kanji/kana mixed notation, 24... Kanji/kana → Roman character conversion means, 25... Roman character buffer, 26...
・Morphological analysis means for Romaji notation, 27... Romaji → Kanji-kana conversion means, 28... Corrected character string buffer,
29... Document memo IJ, 30... Word dictionary in kanji-kana notation, (13) 31... Pointer, 32... Word dictionary in romanization, 33... Kanji-kana → romanization correspondence table, 34・
...Connection matrix between words. Agent Patent attorney Akio Takahashi (14)

Claims (1)

[Scope of Claims] 1. In a data processing device that inputs a series of sentences containing kanji and kana, divides them into clauses while referring to an interrogation dictionary, and outputs the results, comprising: or a 6 first means for converting the entire phonetic alphabet to the phonetic alphabet, a phonetic alphabet zero word dictionary, and a second means for converting the phonetic alphabet to the kanji kakuho mixed transcription; When the normal division processing of a series of sentences containing kanji and kana fails, part or all of the kanji and kana sentences related to the failure are expressed in letters in the table by the first means, and the words in the phonogram are written. A data processing device characterized by performing a second division process of dividing the phrase into phrases while referring to a dictionary, converting the phrase into a sentence containing kanji and kana by the second means, and outputting the phrase. 2. Claims No. In the data processing according to item 1, a printer or a display is provided as a means for outputting the division results, and in the second division process in item 1ff, an output instructing the selection of homophones is sent to the printer or the display. A data processing device characterized by displaying.1