JPS63308665A - Kana/kanji converting device - Google Patents

Abstract

PURPOSE:To execute the clause delimitting learning of a KANA (Japanese syllabary)/ KANJI (Chinese character) conversion more securely by connecting information for two clauses and registering them to a clause delimitting learning dictionary. CONSTITUTION:When the clause of an analyzing position is a clause delimitting change clause, a clause to which a clause length change is executed and the next clause are succeed and in a condition a clause delimitting mark is placed, they are registered to a learning dictionary 12. However, when the same index exists in the dictionary 12, after this is deleted, the said clause delimiting data are registered. Next, when the independent word part of the established clause is not smaller than the independent word part in the other clause candidacy, to the homonym of the said independent word, a word dictionary 11 is rewritten so that the said independent word can be a forefront. When the independent word part of a selected and shown clause is smaller than the independent word part of the other clause candidacy, with a unit not smaller than the length of the longest independent word, in a condition an attached word is combined to the independent word, it is newly registered to the word dictionary 11. Thus, the clause delimitting learning of a KANA/KANJI conversion can be executed more securely.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field J] The present invention relates to a kana-kanji conversion device that converts a kana character string input for inputting a Japanese sentence into a sentence containing kana and kanji.

[Prior art] In order to learn bunsetsu divisions, conventional kana-kanji conversion devices mix both bunsetsu whose bunsetsu lengths have been changed and bunsetsu whose bunsetsu lengths have not changed, treat the bunsetsu themselves as one word, and convert them into words. Separately from the dictionary, the phrase information to be learned was registered in the learning phrase dictionary.

[Problems to be Solved by the Invention] According to the prior art, the phrase itself is registered as one word, so for example, in response to the input of "Today is a flower," it is possible to input "Today is / Flowers / the most.".'', and here, if you select ``bloomed'' from the phrase candidates (``most often'', ``bloomed'', ``torn'', ``ejected'', etc.) for the phrase ``most,'' then ``bloomed'' will be memorized. In the following analysis, the input ``Yesterday the flowers bloomed'' can be converted to ``The flowers bloomed yesterday too.'' However, the input ``Here I wipe my kimono.'' here/
take off/take off one's footwear. '' is converted to ``Here I take off my 7 kimonos.'' If we extend the length of the three-character phrase ``here'' to the length of the four-character phrase ``here'', we get "Here," is learned, and if this learned phrase break is given priority in the bunsetsu division process, the next time the same input is made, "Here, wipe the kimono." Although the input can be converted to ``Here I take off my seven kimonos.'', since we are learning only one phrase, the input ``Here I sharpen things'' should normally be converted to ``Here I sharpen my cutlery.'' Even though it is wanted to be translated as ``.'', here we are sharpening /things/. ” and “here” are converted with priority.

In other words, the learning result of ``Here, 7 kimono'' is converted to ``Here / cutlery'', but it is converted to ``Here / thing'', and the phrase should be separated based on the relationship between two phrases. This will give unnecessary information to the postscript process, leading to incorrect analysis.

Furthermore, in response to the input "Ask whether you want to do something.", if you change the phrase length from "Calligraphy or visiting/visiting." to "Calligraphy/home/visiting.", "Calligraphy" becomes The problem was that it was only learned, and even if the same input was made again next time, it would only be converted to ``Calligraphy or visiting.''

The present invention solves the above-mentioned drawbacks, and provides a kana-kanji conversion device that performs more accurate phrase separation learning for kana-kanji conversion by concatenating information for two phrases and registering the same in a phrase separation learning dictionary. With the goal.

[Means for Solving the Problems] In order to solve these problems, the present invention provides an apparatus for converting a solid kana character string into a sentence with kana and kanji mixed into segments. A word dictionary that has character strings and parts of speech that are their grammatical information, a phrase dividing means that uses this to divide phrases into phrases, and clearly divide phrases into sentences mixed with kana and kanji that have been divided into phrases. a phrase candidate display means that displays phrase candidates one after another for each divided phrase, and a phrase candidate display means that displays phrase candidates one after another for each phrase that has been divided, and a phrase candidate display means that displays phrase candidates one after another for each phrase that has been divided, and for phrases whose phrase length is not changed, On the other hand, there is a phrase candidate learning means that learns phrase candidates into a word dictionary, and a phrase length change that expands or contracts the target phrase length based on the phrase segmentation that has been done in advance, and reanalyzes the sentence after the target phrase. a bunsetsu break learning means that concatenates the bunsetsu whose bunsetsu length has been changed and the next bunsetsu, has a bunsetsu break mark, and stores the bunsetsu break, and a bunsetsu break learning dictionary that has this bunsetsu break information; In the text writing process, if both the bunsetsu break learning dictionary and the word dictionary are analyzed, and as a result, the contents of the bunsetsu break learning dictionary become bunsetsu targets, the bunsetsu break marks are used to separate them into two bunsetsu. [Action] When converting a solid kana character string into a sentence with kana and kanji mixed in, the kana character string inputted by the input means is converted into a sentence divided into segments. When the word dictionary and the bunsetsu break learning dictionary are searched in the same way, both are grammatically analyzed as one independent word, and the bunsetsu break is performed, and the sentence registered in the bunsetsu break learning dictionary becomes a candidate for one bunsetsu. divides it into two clauses based on the delimiter mark therein, displays the sentence containing kana and kanji using a separate display means with clear clause divisions, and displays clause candidates for homophones for each clause. If the bunsetsu separation is incorrect, the bunsetsu length is lengthened or shortened using the bunsetsu length changing means, and the text after the target clause is reanalyzed to produce the correct kana-kanji conversion result. get.

In order to learn the conversion results here, first, for a clause whose clause length has not been changed, the independent word part of the clause selected and displayed by the clause candidate learning means is smaller than the independent word part of other clause candidates. If not, rewrite the homophone group of the independent word so that the independent word comes first, and if the independent word part of the selected and displayed clause is smaller than the independent word part of other clause candidates. is newly registered in the word dictionary in a state in which an independent word is combined with an adjunct word in units not smaller than the length of the longest independent word 66.

Furthermore, for the phrase whose phrase length has been changed, the phrase whose phrase length has been changed is concatenated with the next phrase by the phrase break learning means, and the phrase is registered in the phrase break learning dictionary with a phrase break mark attached. However, if the same heading exists in the phrase break learning dictionary, the phrase break data is registered after deleting it.

[Example] Here, the present invention will be explained based on an example showing details of the present invention.

FIG. 1 is a diagram showing a kana-kanji conversion device of the present invention.

Reference numeral 1 in the figure is a central processing unit (CPU) that controls the entire system based on data input from the input section 2. This CPU controls all devices.

The kana character string inputted in the input section 2 is sent to the kana-kanji conversion device 8 where kana-kanji conversion is performed, and while the phrase candidates are displayed on the phrase candidate display section 3, the phrase candidate selection section 5 converts them into phrases in units of phrases. Display and select phrase candidates, and phrase candidate confirmation section 6
A desired clause candidate is determined by the above steps, and the kana-kanji character string is stored in the text storage section 7 as a text.

In the kana-kanji conversion device Nl2, the kana characters are picked up by the kana character acquisition unit 9, and the phrase separation unit 10 searches the word dictionary 11 and the phrase separation learning dictionary 12, and performs the phrase separation process while performing grammar processing. conduct.

If the sentence registered in the phrase separation learning dictionary becomes a candidate for one phrase as a result of this separation, it is divided into two phrases based on the break mark therein, and the homophone and allophone phrases of the two phrases are separated. The word dictionary 11 is searched and created, and the phrase candidate storage unit 1 is created.
6.

Here, the division display section 14 reads out the first candidate for each clause from the clause candidate storage section 16 and displays it on the clause candidate display section 3. Thereafter, the phrase candidate selection section 5 reads candidates one after another from the phrase candidate storage section 16 and the phrase candidate output section 15 in phrase units, and while displaying them on the phrase candidate display section 3, selects desired kana-kanji character strings in phrase units. .

If there is an error in the bunsetsu division, the bunsetsu length changing unit 4 extends or shortens the clause and instructs the clause re-analysis unit 13 to re-analyze the clause after that clause. Clause reanalysis part 13
In this example, the analysis of a more limited part is performed using the bunsetsu segmentation section 10.
The analysis result is stored in the phrase candidate storage unit 16 by adding information indicating that the phrase length has been changed to the phrase length changed phrase data.

Then, the division display section rewrites the kana-kanji character string after the clause whose clause length has been changed.

Once the desired kana-kanji character string is obtained through the above operations, it is determined by the phrase candidate determining section 6, and the result is stored in the text storage section 7 and sent to the determined phrase information receiving section 17.

Based on the information sent to the confirmed clause information receiving unit 17, the clause candidate learning unit 18 determines whether the independent word part of the selected and displayed clause is among other clause candidates for clauses whose clause length has not been changed. If it is not smaller than the independent word part of the selected and displayed clause, the word dictionary 11 is rewritten so that the independent word comes first in the homophone group of the independent word, and the independent word part of the selected and displayed clause is If it is smaller than the independent word part in the clause candidate, the independent word is newly registered in the dictionary 1111 in a state in which the attached word is combined with the independent word in units not smaller than the length of the longest independent word. Furthermore, for the phrase whose phrase length has been changed, the phrase whose phrase length has been changed is concatenated with the next phrase and is registered in the phrase break learning dictionary 12 with a phrase break mark attached. However, if the same heading exists in the phrase break learning dictionary, the phrase break data is registered after deleting it.

FIG. 2 is a flowchart illustrating the operation of the kana-kanji conversion device of the present invention regarding the creation of clause candidates.

As mentioned above, the kana-kanji conversion device 201 of the present invention
The kana character string is received, the kana analysis position is initialized, and the word dictionary 11 and the clause break learning dictionary 12 are searched from the predetermined kana analysis position in <202> to extract word candidates. Based on this, clause separation is performed by <203>.

If there is a learning dictionary clause in the clause separated by <204>, then by <205>, these clauses are separated into two clauses according to the delimiter marks, and one
The word dictionary is searched at the same level as the Bunzuru wheel, and the same number of different word candidates for each clause is created. When all the clause candidates are obtained, the clause break is determined by <206>, and the division table is displayed by 207>. .

FIG. 3 is a flowchart illustrating the operation of the kana-kanji conversion device of the present invention regarding phrase candidate learning.

As described above, the phrase candidate learning of the kana-kanji conversion device of the present invention receives the confirmed phrase information in <301>, and first sets the analysis position to the first phrase.

Then, if the phrase at the analysis position is a phrase whose length has been changed by <302>, the phrase whose length has been changed by <303> is concatenated with the next phrase, and with a phrase delimiter mark attached, the phrase delimiter learning dictionary Register on 12th. However, if the same heading exists in the phrase break learning dictionary, the phrase break data is registered after deleting it. (Not shown) Furthermore, if the independent word part of the clause determined by <304> is not smaller than the independent word part in other clause candidates, <
305>, the word dictionary 11 is rewritten so that the independent word comes first in the homophone group of the independent word. In addition, if the independent word part of the selected and displayed bunsetsu is smaller than the independent word parts of other bunsetsu candidates, <306> combines the independent word number and attached word in units that are not smaller than the length of the longest independent word. In this state, the word is newly registered in the word dictionary 11.

When these processes are completed, it is checked in <307> whether the clause being processed is the final clause, and if it is not the final clause,
The process from <302> (not shown) is repeated to transfer the processed clause to the legal clause. Note that if the phrase is the final phrase, the process ends here.

[Effect of the invention] The present invention has the following effects.

(1) Clause break learning device! In step 12, the two clauses are stored with delimiter marks, and this is used for analysis by the clause divider 10, so that the learning results limited to the relationship between the two clauses can be reflected, allowing for detailed analysis. There is no danger of it being lost.

(2) Two clauses are stored in the clause break learning device 12 with a break mark, and the clauses are used for analysis in the clause divider 10, so that the clause length changing unit 4 converts the clauses into two clauses. If you do this, you can learn accurately.

(3) By making the bunsetsu division learning dictionary 12 independent from the word dictionary 11, the word dictionary is not overwhelmed.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a kana-kanji conversion device of the present invention. FIG. 2 is a flowchart showing the bunsetsu division writing operation of the kana-kanji conversion device of the present invention. FIG. 3 is a flowchart showing the phrase learning operation of the kana-kanji conversion device of the present invention. 10...Bunsetsu division learning section 12...Bunsetsu separation learning dictionary 18...Bunsetsu candidate learning section Above, '〆, - Agent Patent attorney Tsutomu Mogami and 1 other person ・Second figure

Claims (1)

[Claims] In a device that converts a solid kana character string into a kana-kanji mixed sentence written in segments, it uses a word dictionary that has kana headings, corresponding kanji character strings, and parts of speech that are grammatical information, and uses this to convert text into segments. A bunsetsu segment writing means for performing bunsetsu writing, a split writing display means for displaying the bunsetsu segmented and mixed kana-kanji sentences with clear bunsetsu divisions, and a bunsetsu candidate for each segmented bunsetsu unit. a phrase candidate display means for displaying phrase candidates one after another; a phrase candidate learning means for learning phrase candidates in a word dictionary for phrases selected and displayed for phrases whose phrase lengths are not changed; It has a clause length changing means that expands or shortens the target clause length for writing and reanalyzes the sentence after the target clause, and a clause break mark that connects the clause whose clause length has been changed with the next clause. , has a bunsetsu break learning means that stores bunsetsu breaks and a bunsetsu break learning dictionary that has this bunsetsu break information, and in the bunsetsu break writing process, analyzes both the bunsetsu break learning dictionary and the word dictionary as analysis targets, and calculates the results. This kana-kanji conversion device is characterized in that when the contents of a phrase separation learning dictionary become a phrase, it is divided into two phrases using a phrase delimiter mark, and the phrases are written separately.