JPH04195253A - Continuous clause japanese-syllabary chinese-character conversion method and device - Google Patents

Abstract

PURPOSE: To efficiently segment a clause by using any arbitrary independent word indication when an independent word contained in the clause at the time of clause segmentation error including a proper clause boundary is equal with an independent word contained in the clause at the time of proper description just before the proper clause boundary. CONSTITUTION: While referring to a dictionary 2, a Japanese syllabary(KANA)/Chinese character(KANJI) converting part 1 finds converted result candidates more than one. A clause resegmenting part 6 preserves the converted result OUT before clause resegmentation due to an operator in a non-corrected converted result buffer 7, clause segment designation data are prepared and stored in a buffer 5, the converting part 1 is called, and the converted result is corrected. Further, the converted result OUT after clause resegmentation is preserved in a corrected converted result buffer 8 and a learning part 9 is called. The learning part 9 compares the buffer 7 with the buffer 8, decides the kind of clause resegmentation and updates a 1st type resegmentation learning dictionary 3 or a 2nd type resegmentation learning dictionary 4. Namely, when the independent words contained in clauses before and after correction are equal, the clause is registered in the dictionary 3 but when they are not equal, it is registered in the dictionary 4.

Description

[Detailed Description of the Invention] A. Field of Industrial Application This invention relates to a method and device for converting kana to kanji in connected clauses;
In particular, it allows for efficient learning of bunsetsu segmentation.

B. Conventional technology In the continuous clause kana-kanji conversion method, which converts an input kana character string into a sentence containing kanji and kana for multiple clauses, the input kana character string is divided into clause candidates using a predetermined algorithm. A string of characters mixed with kana and kanji corresponding to each is generated and concatenated to form the conversion result. If this bunsetsu cut is incorrect, the correct bunsetsu cut is learned, so that the same bunsetsu cut error will not occur in the future.

For example, in contrast to the official text that says ``Saitama-kei will rebuild its prefectural finances,'' it says ``Saitama Kenzai will rebuild its prefectural finances.''
If you enter the kana character string, it may be incorrectly converted to ``Rebuild the dispatch finances in the Saitama system.'' this is"
This is the result of incorrectly cutting off the sentence as ``Saitama Prefecture Te/Dispatch'' when the phrase should have been ``7 prefectures in Saitama.'' When performing learning to deal with such bunsetsu cutting errors, when correcting to ``7 prefectures in the Saitama system,'' the bunsetsu cutting information is stored so that the bunsetsu cutting will be performed correctly thereafter.

By the way, conventional bunsetsu cutting information for learning is specific to the input character string in which the bunsetsu cutting error occurred, for example, "
``Saitama Ken de Haken'' is pronounced ``Saitama Prefecture J.
It was against the official text. Therefore, it was necessary to learn about similar errors individually. After studying the above-mentioned ``Saitama-kei prefecture'', ``Chibaken dehaken...''
”, there is still a risk that a phrase error will occur, such as ``Dispatched in Chiba,'' and in that case, it will be necessary to study it individually.

The prior art related to this invention is JP-A-61-
There are those published in Japanese Patent Application Laid-open No. 147365 and Japanese Patent Application Laid-open No. 147368/1983.

The prior art disclosed in Japanese Unexamined Patent Publication No. 147365/1981 stores an adjunct word including the correct boundary before bunsetsu cutting, and a kana character string from the beginning of the adjunct word to the correct boundary, and learns bunsetsu cutting. For example, when the input ``Ima no Nihanba'' for ``Today's Japan'' (overwritten) is mistakenly converted to ``Today's hon wa'', learning can be done using ``noni'' and ``no''. From then on, when ``no'' is input, a clause boundary is placed after ``no''. However, this prior art can only deal with cases in which segmentation occurs within a character string of an adjunct word. For example, when "I'm going to the doctor today" (overwritten) is incorrectly converted to "I'm going to the dentist today", the incorrect translation "I'm going to the doctor today" (
This prior art cannot be applied because there is no attached word that includes a correct clause boundary (- indicates a correct clause boundary).

Additionally, there are many negative effects (side effects) to learning J Una Bunseki Kiri.

If the adjunct word ``nonij'' were to be used to cut the sentence with ``no'', there would be no meaning in registering the adjunct word ``noni'' in the adjunct word dictionary.

The prior art disclosed in Japanese Unexamined Patent Application Publication No. 147366/1983 learns bunsetsu cutting by storing information on sentences including the adjunct word of the clause before bunsetsu cutting, including the correct boundary, and the independent word following the clause. For example, when the input ``Please be careful not to display the content'' (overwrite) is incorrectly converted to ``Please be careful not to display the contents'', the input ``Please be careful not to display the content'' (overwrite) is incorrectly converted to ``Please be careful not to display the contents''. It memorizes ``nai ni ni uni chui'' and ``shi'', and from now on, when ``nai ni ni ni chui'' is input, a clause boundary is placed after ``shi''. This prior art also causes the same problem as that of Japanese Patent Application Laid-Open No. 61-147365. However, the negative effects can be improved slightly, as can be understood by considering the ``chiyuui'' in the previous example.

However, while the negative effects have been improved, the range of phrase-cutting errors for which learning is effective is narrowing. In the previous example, if ``note'', ``consider'', ``reflect'', etc. come in place of ``Caution J'', it cannot be applied.

Problems to be solved by the C0 invention This invention was made in consideration of the above circumstances,
- It is possible to learn bunsetsu cut information of times, and not only the bunsetsu cut specific to the input character, but also the typical bunsetsu cut represented by that input character, and efficient bunsetsu cut learning is possible. Furthermore, the purpose of this invention is to provide a continuous phrase phrase conversion method that can minimize the negative effects of such typical learning.

91 Means for Solving the Problems In order to achieve the above object, in this invention, when learning bunsetsu cutting, an independent word included in the bunsetsu at the time of bunsetsu cutting error that includes the correct bunsetsu boundary is When the independent word is the same as the independent word included in the orthographic clause immediately before , learning is performed using an arbitrary independent word display instead of the independent word of the clause. In this way, comprehensive learning of bunsetsu cutting becomes possible. On the other hand, if the independent word included in the clause at the time of bunsetsu cut error that includes the correct clause boundary is different from the independent word included in the orthographic clause immediately before the correct clause boundary, the independent word itself of the clause is used. , we are trying to avoid comprehensive learning and avoid the negative effects of comprehensive learning.

Here, I would like to touch on the definition of independent words. As mentioned above, in the continuous clause kana-kanji conversion, the first step is to cut the clauses. This bunseki-kiri is the edict registered in the first dictionary (or part of the dictionary), and the word (including empty strings, which will be explained later) registered in the second dictionary (filter or part of the dictionary). Example) are connected to form a clause, and this is matched against the input character string.Here, the words registered in the first dictionary are called independent words, and the words registered in the second dictionary are called independent words. are called attached words.For example, the previous word ``Saitama prefecture'' means ``Saitama (independent word)''.
−Empty string (adjunct word) −Prefecture (independent word) −De (adjunct word)
-Prefecture (independent word) It is composed of J.

E, l! ii! Example An embodiment of this invention will be described below.

FIG. 1 shows this embodiment as a whole. Although FIG. 1 shows a hardware implementation using functional blocks, it is of course possible to implement the system using software. In FIG. 1, a kana-kanji conversion unit 1 receives a kana input IN and performs continuous clause kana-kanji conversion while referring to a dictionary 2. The kana-kanji conversion section 1 first refers to a dictionary (including an independent word dictionary section and an attached word dictionary section) 2 to obtain one or more conversion result candidates. Since this process itself is normal, a detailed explanation of this point will be omitted (an example will be given later regarding the movement). Next, the kana-kanji conversion unit 1 determines the conversion result according to predetermined rules (for example, depending on the size of the sum of the score or penalty of each word and the score or penalty of the connection of the word). The conversion result is determined by referring to the recutting learning dictionary 3, the second type recutting learning dictionary 4, and the bunsetsu cutting specification data buffer 5.

The phrase re-cutting section 6 is a section that performs processing when the operator instructs an error in phrase cutting. That is, the conversion result OUT before the bunsetsu re-cutting by the operator is saved in the uncorrected conversion result buffer 7, the bunsetsu-cutting specification data is created (stored in the buffer 5), and the kana-kanji conversion unit 1 is called to correct the conversion result. . Furthermore, the conversion result OUT after the phrase re-cutting is stored in the modified conversion result buffer 8, and the learning section 9 is called. Note that phrase re-cutting can be performed by, for example, converting the erroneous conversion result into the original kana characters, and then marking only the desired phrase as a conversion area by highlighting it, etc., and converting it again.

The learning unit 9 compares the pre-correction conversion result buffer 7 and the post-correction conversion result buffer 8 to determine the type of phrase re-cutting, and uses the first type re-cut learning dictionary 3 or the second type re-cut learning dictionary 4. Update.

In other words, as will be shown later with an example, when the independent word included in the uncorrected clause that includes the correct boundary is the same as the independent word included in the corrected clause immediately before the correct boundary, the Type 1 reshuffle dictionary is used. 3, and if they are not the same, they are registered in the second re-cutting dictionary 4.

Next, the process of kana-kanji conversion, phrase re-cutting, and re-conversion will be explained using an input kana character string such as ``Shitsuken no Naiyo'' as an example.

A detailed explanation of the conversion before learning bunsetsu segmentation.

In Figure 2, in the independent word search (step Sl), the fifth
An independent word as shown in the figure is stored in an independent word buffer JB. In the adjunct word search (step S2>, the adjunct words shown in FIG. 6 are stored in the adjunct word buffer FB. In the clause composition process (step S3), the above two buffers are referred to and the town name shown in FIG. 7 is stored. A grammatically correct clause is stored in the clause buffer BB. In the candidate composition process (step S4), the clause buffer BB is referred to and candidates as shown in FIG. 8 are stored in the candidate buffer CB. search,
All the processes of adjunct word search, clause construction, and candidate construction are collectively referred to as "candidate generation."

Conversion result determination (step S5) refers to the pre-correction conversion result buffer 7, the post-correction conversion result buffer 8, and the bunsetsu cut designation data buffer 5, selects 1 from the candidates stored in the candidate buffer CB, and selects the conversion result OUT. Update. In this case, there is no data in the above-mentioned two types of re-cutting learning dictionaries 3.4, and there is no phrase-cutting specification data, so
``There is no experiment'' is the conversion result (scores etc. are given so that candidates are selected in order from top to bottom in Figure 8).

Next, we will explain the process by which the operator rearranges the phrases to change the phrase ``avoid an experiment'' to ``the content of the experiment.'' In FIG. 3, first, in step Sll, the uncorrected conversion result is saved. That is, the kanji character ``Jikken no Nai Machiu'' and the kana character ``Jitsuken no Naiyo'' are stored in the pre-correction conversion result buffer 7. The operator specifies the phrase cut (step 512) by specifying that the fifth character "na" is the beginning of the phrase, and stores "5" in the phrase cut designation data buffer 5 accordingly. Thereafter, the kana-kanji conversion unit 1 is called in response to an instruction for kana-kanji conversion (step 513).

The candidate generation process is the same as described above even in the case of a conversion called inadvertently in response to an instruction for kana-kanji conversion (step 513).

In the conversion result determination (step 514), the clause cut designation data "5" is referred to and a town una candidate in which the fifth character "na" is the beginning of the clause, ie, "experiment content" is selected (FIG. 8).

In saving the modified conversion result OUT (step 514), the kanji ``content of the experiment'' and the kana ``shitsuken no -naiyo''(``-'' represents a bunsetsu boundary) are stored in the modified conversion result buffer 8. Instructions for learning later (step 515)
In this case, the learning section 5 is called.

Learning of bunsetsu boundaries is performed as shown in FIG. In FIG. 4, an independent word comparison immediately before a break is performed (step 521). That is, the first independent word in the unmodified conversion result buffer 7 and the first independent word in the post-modified conversion result buffer 8 are compared. In this case, since both are "experiments", −
It becomes a scattered output, and control is passed to type 1 learning (step 522). In type 1 learning, the matching independent word ``shitsuken'' is deleted from the kana ``jitsuken no-naiyo'' stored in the corrected conversion result buffer 8, and ``no-nai machiu'' is recut to type 1. Add it to learning dictionary 3 (Figure 12).

After learning this way, when the operator tries to convert the kana ``Shitsuken no Naiyo'' again, the kana-kanji converter 1 is called and the process shown in FIG. 2 is performed. The candidate generation process is the same as above. Conversion result determination (step S
In 5), the candidate "experiment content" that matches the data "no-naimau" in the first type recut learning dictionary 3 is selected as the conversion result.

Also, consider the process of converting the input kana character string ``chosa no naiyo'' which involves a similar erroneous conversion to the above-mentioned ``shitsuke ken no naiyo'' after the above bunsetsu re-cutting and learning have been performed. Similar to the conversion of the input kana character string "Shiraken no Naiyo", candidates as shown in FIG. 9 are stored in the candidate buffer CB. Conversion result determination (
In step S5), the candidate "contents of investigation" that matches the data "no-naiyo" in the first type recut learning dictionary 3 is selected as the conversion result. In this case, the erroneous translation of "no investigation" is avoided.

Next, the process of converting, re-splitting, and re-converting the input kana character string "Gokansei" will be described.

The process of generating candidates is similar to the conversion of the input kana character string ``Shitsuken no Naiyo'', and candidates as shown in FIG. 10 are stored in the candidate buffer CB. In the conversion before bunsetsu re-learning, the conversion result is determined (step S5, second
(Figure) selects "Completion" as the conversion result.

Then, in the process where the operator re-cuts the phrases and corrects ``gosakusen'' to ``compatibility,'' the pre-correction conversion result is saved (S11, Figure 3) in the pre-correction conversion result buffer 7 as kanji rIII Kanji. ” and the kana “Go-kansei” are stored. Specifying bunsetsu cut by operator (312)
stores "4" in the bunsetsu cut designation data buffer 5 (gogan-sei). The kana-kanji conversion unit 1 is called by the instruction for kana-kanji conversion (S13). Conversion result determination (
SS (FIG. 2) refers to the clause cut designation data "4" and selects the candidate "compatibility" in which the fourth character "se" is the beginning of the clause from among the candidates shown in FIG. To save the modified conversion result (S14, FIG. 3), the kanji character "compatibility" and the kana "dokan-sei" are stored in the modified conversion result buffer 8. The learning instruction (815) calls the learning section 5.

In the independent word comparison immediately before the break (step S21, the fourth
Figure) The first independent word in the conversion result buffer 7 before correction and the first independent word in the conversion result buffer 8 after correction are compared. In this case, since it is "compatible" with "go", a mismatch output occurs and control is passed to the second type learning S23. In the second type learning 23, the kana "Gogan-sei" stored in the modified conversion result buffer 8 is added to the second type recutting learning dictionary 4 (
Figure 13).

After this, when the operator attempts to convert the kana "Dokansei" again, the kana-kanji conversion unit 1 is called. The candidate generation process is the same as above.

In the conversion result determination (S5), the candidate "compatibility" that matches the data "gogan-sei" in the second type recut learning dictionary 4 is selected as the conversion result.

Next, the process of converting the input kana character string ``Kikiko next'', which is similar to ``Gokanjikuto'', after the above-mentioned bunsetsu re-cutting and learning is performed, will be described. Candidates as shown in FIG. 11 are stored in the candidate buffer CB (FIG. 2). Conversion result determination (S5) determines the same candidate "equipment configuration" as if nothing has been learned, since there is no matching candidate in the first type recutting learning dictionary 3 and the second type recutting learning dictionary 6. Select as the conversion result.

In addition, in this example, when learning the sentence i boundary of compatibility (“gokan-sei”), “(any independent word)-sei” (the first attached word is an empty string) is entered in the first type reshuffling dictionary 3. If it is registered, "Device-K" will be erroneously converted and output (the third candidate in Fig. 11), but in this example, by determining the first independent word, such clause boundaries can be detected. If you do individual learning instead of all at once, you won't have any of the above problems.

In order to confirm the effectiveness of this example, a text consisting of 954 clauses (Science Council of Japan, 14th term activity plan, November 1988)
I tried converting the month) into kana and kanyu. In this case, if learning according to this embodiment was not performed, there would be 21 boundary errors as shown in FIG. Seven of them were able to be prevented by learning through practical examples. Due to comprehensive learning of boundary error EO2, E07 and E15 are avoided, and EO5 causes EO9
is avoided, E16 is avoided by Ell, and E19, E20, and E21 are avoided by E18.

F6 As described in the detailed explanation of the invention, in this invention, when learning bunsetsu cutting, an independent word included in the bunsetsu-cutting error clause that includes the correct bunsetsu boundary is placed directly before the correct bunsetsu boundary. When the word is the same as an independent word included in a clause, learning is performed using an arbitrary independent word display instead of the independent word of the clause.

In this way, comprehensive learning of bunsetsu cuts becomes possible.

On the other hand, if the independent word included in the clause at the time of bunsetsu cut error that includes the correct clause boundary is different from the independent word included in the clause at the time of correct positioning immediately before the correct clause boundary, the independent word itself of the clause is used. , they stop comprehensive learning and prevent the negative effects of comprehensive learning from occurring.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention;
FIG. 2 is a flowchart explaining the operation of the kana-kanji converter 1 in the embodiment shown in FIG. 1, FIG. 3 is a flowchart explaining the operation of the phrase cutting unit 6 in the embodiment, and FIG. 4 is the embodiment Flowcharts illustrating the operation of the learning section 9 of FIGS. 5 to 13 rI! J is a diagram for explaining the operation of the same embodiment, and FIG. 14 is a diagram for explaining the effects of the same embodiment. 1...Kana-kanji conversion section, 2...Dictionary, 3-...1st
Seed cutting learning dictionary, 4-... Second type cutting learning dictionary, 5... Bunsetsu cutting specification data buffer, 6...
- Phrase re-cutting unit, 7... uncorrected conversion result buffer, 8... corrected erroneous conversion result buffer, 9... learning unit. Applicant International Business Machines
Corporation sub-agent Patent attorney Sawa 1) Toshio 1'
, BB#2 exchange section flowchart Fig. 2 UT 1 , -5i4 clause buffer Fig. 7

Claims (7)

[Claims] (1) In order to convert an input kana character string into a sentence containing kana and kanji, one of the words in the first group registered in the first dictionary part is followed by a word in the second group registered in the second dictionary part. Sequentially generate phrase units formed by combining one of the two groups of words or an empty string, and construct one or more chains of phrase units that match the input kana character string, and create the chain of 1 above or the above. In a continuous clause kana-kanji conversion method that outputs a kana-kanji-mixed sentence selected from a plurality of chains according to a predetermined rule, the method includes a step of receiving correction information on boundaries of bunsetsu units, and a step of receiving correction information on boundaries of bunsetsu units. re-selecting one of the above chains and re-outputting it as a kana-kanji-mixed sentence; The step of comparing the word of the first group immediately before the corrected boundary included in the kana-kanji mixed sentence contained in the re-outputted kana-kanji mixed sentence when the comparison result shows a mismatch. , the kana character string corresponding to the bunsetsu unit before the boundary after the above correction,
and included in the above re-outputted sentence containing kana and kanji,
A kana character string corresponding to at least a word of the first group among the bunsetsu units after the modified boundary is registered as a first series of kana character strings, and a kana character string in the first series of kana character strings A step of registering information indicative of the post-correction boundary position, and when the above comparison result indicates a match, one of the bunsetsu units before the post-correction boundary included in the re-output kana-kanji mixed sentence. Kana character string corresponding to the part excluding the first group of words, and at least the first group of clause units after the corrected boundary included in the re-outputted kana-kanji mixed sentence. registering a kana character string corresponding to the word as a second series of kana character strings, and registering information indicating the modified boundary position in the second series of kana character strings; When the first series of kana character strings is input, or when the second series of kana character strings is input following any of the first group of words, the registered modified boundary 1. A continuous clause kana-kanji conversion method comprising the step of forcibly dividing a clause unit by position. (2) Included in the above re-outputted sentence containing kana and kanji,
2. The continuous clause kana-kanji according to claim 1, wherein one of the kana character strings corresponding to the part of the clause unit before the modified boundary excluding the words of the first group includes an empty kana character string. Conversion method. (3) In order to convert the input kana character string into a sentence containing kana and kanji, create clauses in sequence by combining 1 independent word followed by 1 attached word or empty string, and match the above input kana character string. In the continuous clause kana-kanji conversion method, which configures one or more chains of clauses, and outputs the above chain 1 or the chain selected according to a predetermined rule as a kana-kanji mixed sentence, a step of receiving correction information; a step of reselecting one of the above-mentioned chains to match the correction information of the bunsetsu boundary and outputting it again as a sentence containing kana and kanji; The step of comparing the independent words contained in The kana character string that is included in the kana-kanji mixed sentence and corresponds to the clause before the above-mentioned corrected boundary, and the clause that is included in the above-mentioned re-output kana-kanji mixed sentence after the above-mentioned corrected boundary. registering at least a kana character string corresponding to an independent word as a first series of kana character strings, and registering information indicating a modified boundary position in the first series of kana character strings; When the result shows a match, the kana character string corresponding to the part of the bunsetsu unit before the boundary after the above correction, excluding the independent word, included in the above re-outputted kana-kanji mixed sentence, and the above re-outputted kana character string A kana character string corresponding to at least an independent word among the clauses after the corrected boundary included in the outputted kana-kanji mixed sentence is registered as a second series of kana character strings, and registering information indicating the modified boundary position in the series of kana character strings; and when the first series of kana character strings is input or following any independent word, 1. A continuous phrase kana-kanji conversion method comprising the step of forcibly cutting phrases at the registered boundary position after correction when a series of kana character strings is input. (4) In order to convert the input kana character string into a sentence containing kana and kanji, one of the words in the first group registered in the first dictionary part is followed by a word in the second group registered in the second dictionary part. Sequentially generate phrase units formed by combining one of the two groups of words or an empty string, and construct one or more chains of phrase units that match the input kana character string, and create the chain of 1 above or the above. In a continuous clause kana-kanji conversion device that outputs a kana-kanji mixed sentence selected from a plurality of chains according to a predetermined rule, there is provided a means for receiving correction information on the boundaries of bunsetsu units, and a method adapted to the correction information on the boundaries of bunsetsu units. A means of re-selecting one of the above chains and re-outputting it as a sentence containing kana-kanji; a means of comparing the word of the first group immediately before the corrected boundary included in the kana-kanji mixed sentence, and when the comparison result shows a mismatch, the word included in the re-outputted kana-kanji mixed sentence. , the kana character string corresponding to the bunsetsu unit before the modified boundary, and at least the first group of clause units after the modified boundary included in the re-outputted kana-kanji mixed sentence. registering a kana character string corresponding to the word as a first series of kana character strings,
and a first storage means for registering information indicating the corrected boundary position in the first series of kana character strings; The kana character string that corresponds to the part of the bunsetsu unit before the boundary after the above correction excluding the words of the first group, and the above correction included in the re-outputted sentence containing kana and kanji. The kana character string corresponding to at least the first group of words in the clause units after the boundary of
a second storage means for registering information indicating the corrected boundary position in the second series of kana character strings; means for determining that the first series of kana character strings or the second series of kana character strings has been input by referring to the storage means; In response, said registration occurs when said first series of kana character strings is input, or when said second series of kana character strings is input following any said first group of words. and means for forcibly dividing a bunsetsu unit at the corrected boundary position. (5) When the above comparison result indicates a match, the part of the bunsetsu unit before the above-mentioned corrected boundary excluding the words of the first group included in the above-mentioned re-outputted kana-kanji mixed sentence 5. The continuous clause kana-kanji conversion device according to claim 4, wherein one of the corresponding kana character strings also includes an empty kana character string. (6) In order to convert the input kana character string into a sentence containing kana and kanji, sequentially generate clauses by combining 1 independent word followed by 1 attached word or empty string, and match the above input kana character string. In a continuous clause kana-kanji conversion device that configures one or more chains of clauses, and outputs the above chain 1 or the chain selected according to a predetermined rule as a kana-kanji mixed sentence, means for receiving correction information; means for reselecting one of the above chains to match the correction information for the boundary of the bunsetsu and re-outputting it as a sentence containing kana and kanji; and the clause before correction that includes the border after correction means to compare the independent word included in the above re-outputted sentence with the independent word immediately before the above-mentioned corrected boundary included in the above-mentioned re-outputted kana-kanji mixed sentence, and when the above-mentioned comparison result shows a mismatch, the above-mentioned re-output The kana character string that is included in the kana-kanji mixed sentence and corresponds to the clause before the above-mentioned corrected boundary, and the clause that is included in the above-mentioned re-output kana-kanji mixed sentence after the above-mentioned corrected boundary. A first storage means for registering at least a kana character string corresponding to an independent word as a first series of kana character strings, and for registering information indicating a modified boundary position in the first series of kana character strings. and, when the above comparison result indicates a match, the kana character string corresponding to the part of the bunsetsu unit before the above-mentioned corrected boundary excluding independent words, included in the above-mentioned re-outputted kana-kanji mixed sentence. , and registering a kana character string corresponding to at least an independent word among the clauses after the corrected boundary included in the re-outputted kana-kanji mixed sentence as a second series of kana character strings, and a second storage means for registering information indicating the corrected boundary position in the second series of kana character strings; means for determining that a series of kana character strings has been input or that the second series of kana character strings has been input; When a character string is input, or when the second series of kana character strings is input following any independent word, bunsetsu cutting is forcibly performed at the registered boundary position after correction. 1. A continuous clause kana-kanji conversion device characterized by having means. (7) It has a central processing unit, a storage device, a display device, and a character input device to perform data processing and to convert an input kana character string into a sentence containing kana and kanji. One of the words of the first group is successively combined with one of the words of the second group registered in the second dictionary section or an empty string to generate phrase units, and the above-mentioned input kana characters are generated. In a data processing device that configures one or more chains of clause units that match a column and outputs the above chain 1 or the chain selected according to a predetermined rule from the above plural chains as a kana-kanji mixed sentence, a means for receiving correction information on the boundaries of the sentence; a means for reselecting one of the above chains so as to match the correction information on the boundaries of each bunsetsu unit and outputting it again as a sentence containing kana and kanji; means for comparing the first group of words included in the bunsetsu unit before correction with the first group of words immediately before the corrected boundary included in the re-outputted kana-kanji mixed sentence; When the result shows a mismatch, the kana string corresponding to the bunsetsu unit before the boundary after the above correction, which is included in the re-outputted kana-kanji mixed sentence, and the kana character string included in the above-mentioned re-outputted kana-kanji mixed sentence registering, as a first series of kana character strings, kana character strings that correspond to at least the first group of words among the clause units after the modified boundary;
and a first storage means for registering information indicating the corrected boundary position in the first series of kana character strings; The kana character string that corresponds to the part of the bunsetsu unit before the boundary after the above correction excluding the words of the first group, and the above correction included in the re-outputted sentence containing kana and kanji. The kana character string corresponding to at least the first group of words in the clause units after the boundary of
a second storage means for registering information indicating the corrected boundary position in the second series of kana character strings; means for determining that the first series of kana character strings or the second series of kana character strings has been input by referring to the storage means; In response, said registration occurs when said first series of kana character strings is input, or when said second series of kana character strings is input following any said first group of words. and means for forcibly dividing a bunsetsu unit at the corrected boundary position.