JPH0962668A - Kana/kanji converter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate the demerits of respective learning systems as much as possible and to improve a conversion hit rate as a whole by selecting the learning system corresponding to the attributes of a conversion object in KANA(Japanese syllabary)/ KANJI(Chinese character conversion in the situation of performing learning. SOLUTION: A learning system selection information registration means 3 adds or changes the contents of learning system selection tables 12 by instruction input from a user and makes the user select one of the learning system selection tables 12 prepared for plural kinds beforehand corresponding to utilization purposes. A learning system selection means 11 switches learning processing means 2a-2e based on the attributes of the conversion object in the KANA/KANJI conversion such as the read of an input character string, the grammatical attributes and semantic attributes of a converted result or expression in the situation where the converted result of the KANA/KANJI conversion is established and the learning is performed. The learning processing means 2a-2e performs a learning processing for determining a priority order for defining the candidate of a conversion character string by the respectively different learning systems when the plural candidates of the conversion character string are present for the input character string.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a kana-kanji conversion device used for Japanese input in a computer, a document processing device (or word processor), an electronic notebook, etc.

[0002]

2. Description of the Related Art A conventional kana-kanji conversion device unconditionally learns the latest use learning or frequency learning or performs learning at all when there are a plurality of conversion character string candidates for an input character string. I was using either method.
Alternatively, it was a method of allowing the user to select one of these methods.

Here, the "latest usage learning" method and the "frequency learning" method have advantages and disadvantages, but conventionally, only one of the learning methods has been adopted.

[0004]

The advantages / disadvantages of these two types of learning methods in Kana-Kanji conversion and the advantages / disadvantages when learning is not performed are as follows.

1) First, as a merit when learning is not performed, even after the conversion candidate "child's" is fixed for the reading of "this", the conversion result of inputting "this hatsume""This + invention" can be issued as, and the frequency of use of the intended vocabulary when creating the kana-kanji conversion dictionary is maintained, so the hit rate of multiple clause conversion is not hindered, and the same conversion candidate is displayed each time. Therefore, it can be used to reproduce the same result, and it is convenient, for example, in the case where an instructor explains in a usage class.

[0006] As a disadvantage, "Shinichi"
Even if you only use "Shinichi" in reading,
This is inconvenient because the conversion candidates can always be obtained only in the order of the conversion candidates in the kana-kanji conversion dictionary such as "Shinichi".

2) Next, as a merit in performing "latest use learning", "shinichi" is read for "shinichi"
After that, "Shinichi" becomes the first candidate for conversion. Especially in the case of surnames, place names, proper nouns, etc., the hit rate is often improved.

As a demerit, after the conversion candidate "child's" is fixed for the reading of "this", the conversion result becomes "child's + invention" when "this" is input. If you learn to “change” for the reading of “frog”, you will also “change the dog” for the reading of “dog + frog”. Even if the AI (artificial intelligence) method that takes into consideration the conversion is adopted, it is not always possible to hit because the word "dog" can be said to "keep", "buy", etc. Kana kanji Since the order of conversion candidates in the conversion dictionary changes, the same result cannot be reproduced by the same operation.

3) Next, as a merit when performing "frequency learning", even after "learning dogs + changing" is learned by reading "dogs + frogs", more frequent "dogs can be kept" Can be obtained, and if the frequency of "replace" increases, "replace the dog +" can also be acquired. As a disadvantage, "shinichi" is the first candidate for "shinichi" reading. It is necessary to confirm it several times before it is acquired, and because the order of conversion candidates in the Kana-Kanji conversion dictionary changes, the same result cannot be reproduced by the same operation.

As described above, learning that is generally considered to improve the hit rate is not necessarily the only merit, and therefore there is also a kana-kanji conversion processing product that allows the user to select whether or not to perform learning. Many.

Therefore, there is a demand for a new learning method in kana-kanji conversion that has various learning methods and advantages of not learning and eliminates the disadvantages of each method as much as possible.

[0012]

The kana-kanji conversion device of the present invention comprises means as shown in FIG. 1 in order to solve the above problems. FIG. 1 is a block diagram of the principle of the present invention.
1 is a learning method control means, 11 is a learning method selection means, 12
Is a learning method selection table, 2a to 2e are learning processing means,
Reference numeral 3 is a learning method selection information registration means, and 4 is a kana-kanji conversion dictionary.

The learning method control means 1 comprises a learning method selection means 11 and a learning method selection table 12. The learning method selecting means 11 performs kana-kanji conversion such as reading of an input character string, grammatical attributes and semantic attributes (hereinafter referred to as part-of-speech), or an expression of an input character string when learning is performed after the conversion result of kana-kanji conversion is determined. Is a means for switching the learning processing means 2a to 2e based on the attribute of the conversion target in.

The learning method selection table 12 is an attribute (situation) of the conversion target in the kana-kanji conversion in the learning phase.
Stores information for selecting which learning method to use according to.

When there are a plurality of conversion character string candidates for the input character string, the learning processing means 2a to 2e perform learning processing for determining the priority order of the conversion character string candidates by different learning methods. It is a means. A plurality of learning processing means 2a to 2e are provided and switched by the learning method selection means 11. Each of the learning processing units 2a to 2e is, for example, a learning processing unit 2a that performs permanent frequency learning, a learning processing unit 2b that performs permanent latest use learning, a learning processing unit 2c that performs temporary frequency learning, and a temporary learning process. Learning processing means 2d for performing the latest usage learning, not learning (that is,
The learning processing means 2e does not rewrite the initial state of the kana-kanji conversion dictionary 4.

The learning method selection information registration means 3 is means for adding or changing the contents of the learning method selection table 12 in response to an instruction input from the user. Further, the learning method selection information registration means 3 has means for allowing the user to select one of the learning method selection tables 12 prepared in advance according to the purpose of use.

In brief, the kana-kanji conversion device of the present invention uses a plurality of learning methods for each specific case (situation) in kana-kanji conversion when changing the order of conversion candidates. It is equipped with a learning function that selects one of them for learning.

In the specific case, the conversion target is
In the case of a word (character string) that has a specific "reading" or that contains a specific "reading", has a specific "part of speech" in the case of a word that has a specific "expression" or a specific "expression" In the case of a word, it is the case where the above or the above is combined.

Here, "reading" is a reading character string of the input character string, and "expression" is the character string itself expressing the conversion result. “Part of speech” is binary data that represents various attribute parameters such as words in the conversion result, stores the type of inflectional endings, and also information about AI (artificial intelligence) conversion, whether or not user settings are registered, etc. It includes various types of information.

For these specific cases, in the learning phase, one of the plurality of types of learning methods prepared in advance is adopted for learning. Learning methods include "frequency learning", "latest usage learning", and "no learning". “Frequency learning” means that the display order of the conversion candidates is the fixed frequency order, or the conversion result with the highest frequency is the first.
This is a candidate learning method. "Learning use learning" means that the display order of conversion candidates is the temporal order in which they were used most recently.
Alternatively, it is a learning method in which the latest conversion result is used as the first candidate. “No learning” is a case where learning is not performed and the state of the kana-kanji conversion dictionary is not rewritten, but for convenience, it will be described as a learning method.

For example, in a specific case, the conversion result of the input character string "frog" is logically converted by AI conversion such as "return home + dog" + "keep dog" or "frog + song". It contains many words that are likely to be attached. Therefore, the input string "kaeru"
Since "reading" is not suitable for "least-used learning", in this case, "no learning" or "frequency learning" should be selected for a specific "reading" such as "frog". To do.

As a specific case, for example, "boil" is a word related to food, and there are many similar words such as "umani,""chinese," and "nimeme.""Latest use learning" is not suitable for effective use in AI conversion. Therefore, “frequency learning” is selected for a specific “expression” of “ni”.

As a specific case, "surname",
When many proper nouns such as "name" are used, the homonyms to be converted are almost fixed. In such cases, "least-use learning" is usually advantageous, so
"Latest use learning" for "part of speech" such as "surname", "first name"
To be selected.

Further, when the conversion target corresponds to a plurality of specific cases, a priority is given to the case (situation) in advance, and one learning method is adopted based on this priority to more reliably It is possible to decide the choice of learning method.

In another case, for example, "frog"
It is also possible to define a learning method for a condition consisting of a combination of a plurality of cases, such as when "reading" and "part of speech" is "animal". Therefore, a more appropriate learning method can be adopted.

Furthermore, it is also effective to distinguish permanent learning from temporary learning for both the latest use and the latest frequency. "Permanent" learning is a learning method that makes the result of the previous learning valid even if the kana-kanji conversion device is restarted. “Temporary” learning is a learning method in which the learning result is valid during a series of operations, but the learning result is discarded at the end of the operation. The effect of learning is temporary, and is effective, for example, when a large number of students use one device at an educational site or the like, the operation of the previous student does not affect the operation of another student.

The setting of "no learning" is particularly effective in a case where the effect of AI conversion is desired to be utilized. For example, the conversion result of the input character string "Anna" is "Anna" of "pronoun".
And "Anna" of "Name" is included, even for "Reading" of "Anna", even "frequency learning" will hinder the hit rate, and the first candidate must be the pronoun "Anna". desirable.

As described above, since the appropriate learning method corresponding to the specific case (situation) of the input character string is automatically selected, the conversion hit rate is improved. In addition, the learning method selection information registration means 3 assigns a more appropriate learning method by setting and registering a specific case or learning method based on the purpose and environment of use of the user.
The conversion hit rate can be improved.

For example, "last name", "first name", "place name", etc. are usually most suitable for "latest use learning".
However, it is inconvenient for the user who inputs a large number of addresses / names to use these part-of-speech as the latest use learning, and it is desirable to set “frequency learning” or “no learning”. In addition, for example, in an educational setting, it is desirable that the same conversion result can be obtained for the same operation each time the lesson starts by not performing permanent learning at all and only performing temporary learning. By providing the user with the function of setting the learning method selection information, it becomes possible to select the optimal learning method according to the purpose of use of the user.

[0030]

Embodiments of the present invention will be described below. FIG. 2 is a diagram showing a device configuration example according to the embodiment of the present invention.

In FIG. 2, 20 is a central processing unit, and 20
0 is a kana-kanji conversion program, 201 is a file controller, 202 is a display controller, 203 is an input controller, 204
Is a user interface control unit, 205 is a conversion logic control unit, 206 is a learning control unit, 207 is a situation-based learning control unit, 208 is a learning method selection table, and 209 is “reading”.
Another learning control unit, 210 is a “part of speech” learning control unit, 211
Is a learning control unit for each “expression”, 212 is a learning processing unit (permanent frequency), 213 is a learning processing unit (permanent latest use), 21
4 is a learning processing unit (temporary frequency), 215 is a learning processing unit (temporary latest use), 216 is a learning processing unit (no learning), 217 is a learning method selection information registration unit, 30 is a display device, and 40 is a keyboard. A device and 50 represent a kana-kanji conversion dictionary.

The Kana-Kanji conversion program 200 is a general term for a software component group which operates on the central processing unit 20 and controls various constituent elements for Kana-Kanji conversion. The display device 30 is a display device such as a CRT display or a liquid crystal display, and is a device for displaying a user interface for kana-kanji conversion, for example, a character string in the middle of kana-kanji conversion or a character string resulting from the conversion. Keyboard device 4
0 is a device for the user to input a reading character string and a conversion instruction.

The file control unit 201 is means for controlling input / output to / from the kana-kanji conversion dictionary 50, and the display control unit 202 is means for controlling the display device 30 for realizing a kana-kanji conversion user interface. The input control unit 203 is means for controlling the keyboard device 40 for realizing a user interface for kana-kanji conversion.

The user interface control unit 204 is a control unit that controls the user interface for Kana-Kanji conversion, such as passing the input "reading" to the conversion logic control unit 205 for conversion and managing learning. The conversion logic control unit 205 is a unit that realizes a logic that converts the reading character string passed from the user interface control unit 204 into a Japanese character string mixed with kanji and kana.

The learning control section 206 is a control means for controlling the learning function. The situation-based learning control unit 207 is a unit that is called by the learning control unit 206 and refers to the learning method selection table 208 to select a learning method for each situation, thereby performing control for learning by the selected method. The learning method selection table 208 includes information for selecting which learning method to use in accordance with the attribute (situation) of the conversion target in kana-kanji conversion, and priority between situations or between learning methods in the learning phase. It is a database that stores degrees and degrees. The learning method selection table 208 may be uniquely provided in the kana-kanji conversion program 200, or may be provided as a part of the kana-kanji conversion dictionary 50.

The "reading" -based learning control unit 209 is a means for controlling the selection of the learning method according to the "reading" of the situation, and the "expression" -based learning control unit 210 selects the "expression" of the situation. It is a means to control the selection of learning method according to
The “part-of-speech” -based learning control unit 211 is means for controlling the selection of the learning method according to the “part-of-speech” in the situation.

A learning processing unit (permanent frequency) 212 is means for performing learning by a permanent frequency learning method, and a learning processing unit (permanent latest use) 213 is means for performing learning by a permanent latest use learning method. , Learning processing unit (temporary frequency) 2
14 is a means for performing learning by a temporary frequency learning method,
A learning processing unit (temporary latest use) 215 is means for performing learning by a temporary latest use learning method, and a learning processing unit (no learning) 216 is a kana-kanji conversion dictionary 5 without learning.
It is a means for maintaining the initial state of 0.

FIG. 3 is a diagram showing an example of the data structure of a kana-kanji conversion dictionary. In FIG. 3, the index 31
Is an index table that consists of correspondence information between readings and addresses and is used to extract various information necessary for Kana-Kanji conversion using "reading" as a key. When "reading" is input, the index 31 is searched to obtain the address of the expression storage block 32 that is lexicographically matched (minimum match).

The expression storage block 32 is a storage area for storing information on a word group of similar readings. After searching the index 31 and taking out the expression storage block 32, next, the completely matching “yomi” is searched from the expression storage block 32. As a result, when you hit
"Expression", "part of speech" of the word corresponding to the "reading",
The “frequency of use” and the like can be obtained. Note that different words (same synonyms) may be obtained with the same reading.

"Reading" in the expression storage block 32 is a reading character string, and compressed readings may be stored instead of all readings. “Expression” is a character string that expresses the word, and may be actually compressed and stored as in reading. “Part of speech” is attribute parameter information of the word, such as single kanji, exclamation mark, verb, sahen (sa line change verb), plant, animal, surname, first name, place name, AI conversion information, user use information, etc. It is a composite of various information.
"Frequency of use" is information on the frequency of use of words,
Information is also stored for words prepared in advance in order to improve the hit rate. The frequency may be expressed by simply increasing the value by 1 each time it is used. For example, "high,
A stage display such as "medium and low" may be used. It is also possible to manage the frequency of use by including it as a type of attribute in the “part of speech”. In the following description, it is assumed that the value is increased by 1 each time it is used.

FIG. 4 is a diagram showing an example of the data structure of the learning method selection table. In the learning method selection table 208, the conditions that become conditions when selecting a learning method are
Information such as “reading”, “expression”, “part of speech”, the type of learning method, and priority is stored. “Yomi”, “expression”, and “part of speech” specify the learning method of the conversion candidates that match under the AND condition. If only one of them is specified, it becomes a single condition. For example, when the two conditions of "reading: frog" and "part of speech: animal" are set, the conversion candidates "frog" and "frog" that match both of the input character string "frog" are ， The learning method “Permanent latest use” that meets these conditions will be selected.

The "learning method" in the learning method selection table 208 specifies the learning method selected when the conditions match. In reality, it is binary data such as flags or numbers. The “priority” is an index for determining which condition is to be prioritized when a plurality of conditions are met in the learning method selection table 208, and a higher numerical value is set for an item with a higher priority. Please specify.

Learning method selection table 208 shown in FIG.
Is an example for general use. For example, prepare separate tables for specialized use such as "for educational sites" and "for address / name entry person", and specify multiple tables by user's specification. It is also possible to switch. The contents of the learning method selection table 208 can be changed by the learning method selection information registration unit 217.

Here, the learning method selection table 208 is shown as being independent of the kana-kanji conversion dictionary 50. However, as a part of the kana-kanji conversion dictionary 50, all words are used for distinguishing learning. The information may be stored. However, in this case, when adapting the learning method according to the purpose of use, change the entire dictionary each time, or
Alternatively, there is a drawback that another dictionary must be prepared.

FIG. 5 is a diagram showing an example of a learning method selection information registration screen when the contents of the learning method selection table 208 are changed by the learning method selection information registration unit 217. FIG. 5A is a display example of the registration screen when the user selects the learning method. On this registration screen, the user can select whether to use the prepared usage situation (a preset group of detailed learning method settings) or to individually finely set each learning method. It is a screen for making it. Here, 1. Is selected, the screen shown in FIG. When is selected, the screen shown in FIG. 5C is displayed.

In the screen of FIG. 5B, the learning method selection table prepared in advance according to the purpose of use of the user is registered as the table currently in use, so that the usage status can be automatically handled. By selecting one from the general purpose, educational, address / name input, demonstration, etc. screens, you can easily register the appropriate learning method selection table that matches the purpose of use. It was made possible.

In FIG. 5C, the contents of the learning method selection table currently in use are displayed to the user in almost the same format. The user operates the cursor movement keys etc. to select the target to be changed. For example, if the "read" item of "read" is selected, a screen displaying the contents of the selected item is displayed as shown in FIG. In FIG. 5D, each item can be changed by selecting each item with the cursor movement key, and an arbitrary learning method and its priority can be set.

In FIG. 5 (C), for example, the delete key may be defined as deleting the item and the insert key may be defined as adding a new item. FIG. 6 is a processing flowchart according to the embodiment of the present invention.

First, in step S1, ordinary kana-kanji conversion processing is performed. At this time, the conversion candidates are displayed according to the priority determined by the learning performed so far. In step S2, it is determined whether or not the conversion candidate is fixed. If the conversion candidate is not confirmed, step S1
When the conversion candidate is confirmed by continuing the normal kana-kanji conversion process, the learning condition is satisfied, and thus the process proceeds to step S3.

In step S3, the learning method selection table is searched by reading, expression, and part of speech which are conditions for learning. In step S4, it is determined whether any of the corresponding reading, expression, and part of speech can be found in the learning method selection table. If found, the process of step S5 is carried out, and if not found, the process of step S8a is carried out.

In step S5, it is determined whether a plurality of conditions are satisfied. If a plurality of conditions are satisfied, the process of step S6 is performed, and if not satisfied, the process proceeds to step S7. In step S6, the priorities of a plurality of conditions are compared and the one with the higher priority is adopted.

In step S7, one of the learning methods extracted according to the adopted conditions is selected. If the learning method is the permanent latest use, the processing of step S8a is performed, if the learning method is the permanent frequency, the processing of step S8b is performed, and if the learning method is the temporary latest use, the step S8c.
One of the processes of step S8 prepared corresponding to the branching condition of the selected learning method is performed, such as performing the process of.

In steps S8a, 8b, 8c, ...,
A learning process is performed according to each learning method. Step S
In the process of 8a, permanent latest use learning is performed, and step S8 is performed.
The process of b performs permanent frequency learning, the process of step S8c performs temporary latest use learning, and so on.

In the processing of step S4, if no corresponding reading, expression, or part of speech is found in the learning method selection table, "permanent latest use learning" is selected in this example. However, other learning methods may be selected as the default. Also,
It is also possible to adopt a table structure that cannot be undiscovered, for example, by covering all parts of speech.

[0055]

[Embodiment] In this embodiment, in order to allow a user to easily set an optimum learning method selection table according to a purpose of use, an application is selected on a screen as shown in FIG. 5B. Thus, one of the learning method setting groups prepared in advance can be preset and used. This selection enables automatic switching of the optimum learning method according to the purpose as follows.

[For General Purpose] -The word "frog" is read as "permanent frequency learning".・ Even when reading "frog", if the part-of-speech is an animal, it will be "permanent latest use learning".・ The words containing the expressions "boiled", "fried", "fried", etc. are called "permanent frequency learning".・ If the part of speech is "adjective", "verb", etc., it is "permanent frequency learning".・ In the case that does not correspond to any of the above, it is considered as "permanent latest use learning".

[Effects for General Purpose] The hit rate is increased by changing the learning method by finely setting the conditions as described above. In the conventional method, using the kana-kanji conversion dictionary may sometimes lower the hit rate. However, according to this learning method, the more the user uses the dictionary, the higher the hit rate becomes.
A high hit rate can be maintained over the long term by avoiding the risk of masking the conversion effect.

[For Educational Sites] The “permanent latest use learning” explained in the general purpose case is changed to “temporary latest use learning”, and “permanent frequency learning” is changed to “temporary frequency learning”. The learning method selection table changed to is used.

[Effect for Educational Site] When a student uses it in the classroom, while the student is using it, the same hit rate improvement result as intended for “general purpose” can be obtained, and By discarding the results when students start using them, they are not affected by the results used by previous students.

[For Demo Software] All cases are set to “no learning”. [Effects for demo software] The same result is always reproduced with the same operation. This is effective for inputting Kana-to-Kanji conversion with the demo software and for cases where it is guaranteed that the same conversion result as at a remote location will be obtained via the line.

[For Address / Name Input] “Place Name”,
Set "Last name", "First name", etc. to "No learning" or "Permanent frequency learning".

[Effects for Address / Name Input] It is possible to prevent a case in which a large number of addresses / surnames / first names are input, and a less frequently used one becomes the first candidate for conversion by the latest usage learning. it can.

[0063]

As described above, according to the present invention,
By selecting and assigning an effective learning method for each learning phase, the conversion hit rate improves as a whole.

Although it is conceivable that the allocation of the appropriate learning method may be different depending on the purpose of use of the user, the present invention allows the user to change the allocation of the learning method, so that the effective efficiency can be adjusted according to the purpose of use. Good kana-kanji conversion can be realized.

[Brief description of drawings]

FIG. 1 is a principle block diagram of the present invention.

FIG. 2 is a diagram showing an example of a device configuration according to an embodiment of the present invention.

FIG. 3 is a diagram showing an example of a data structure of a kana-kanji conversion dictionary.

FIG. 4 is a diagram showing an example of a data structure of a learning method selection table.

FIG. 5 is a diagram showing an example of a learning method selection information registration screen.

FIG. 6 is a diagram showing a processing flowchart.

[Explanation of symbols]

 1 Learning Method Controlling Means 11 Learning Method Selecting Means 12 Learning Method Selecting Tables 2a to 2e Learning Processing Means 3 Learning Method Selection Information Registering Means 4 Kana-Kanji Conversion Dictionary

Claims (9)

[Claims] 1. A kana-kanji conversion device for converting an input phonetic character string into a character string of an expression centered on kanji by using a kana-kanji conversion dictionary. When there are candidates, there are a plurality of learning processing means for performing learning processing for determining the priority order of conversion character string candidates by different learning methods, and the learning processing means according to the conversion target attribute in kana-kanji conversion. A kana-kanji conversion device, comprising: a learning method control means for switching between. 2. The kana-kanji conversion device according to claim 1, wherein selection information for determining a learning method is stored in the kana-kanji conversion dictionary or separately from the kana-kanji conversion dictionary.
The learning method control means uses selection information for reading the input character string or reading the word contained in the input character string as the attribute of the conversion target A kana-kanji conversion device, characterized in that the learning processing means is selected. 3. The kana-kanji conversion device according to claim 1, wherein selection information for determining a learning method is stored in the kana-kanji conversion dictionary or separately from the kana-kanji conversion dictionary.
The learning method control means includes selection information storage means for storing in correspondence with a specific expression of the conversion result, and the learning method control means selects the expression of the conversion character string or the expression of a word included in the conversion character string as the attribute of the conversion target. A kana-kanji conversion device, characterized in that the learning processing means is selected using. 4. The kana-kanji conversion device according to claim 1, wherein selection information for determining a learning method is stored in the kana-kanji conversion dictionary or separately from the kana-kanji conversion dictionary.
The learning method control means includes selection information storage means for storing in correspondence with a specific grammatical attribute or semantic attribute of the conversion result, and the learning method control means selects the grammatical attribute or the semantic attribute of the conversion result as the conversion target attribute. A kana-kanji conversion device, characterized in that the learning processing means is selected using information. 5. The kana-kanji conversion device according to claim 1, wherein the learning method control means includes a conversion target attribute when there are a plurality of conversion target attributes of kana-kanji conversion for which the learning processing means should be switched. A kana-kanji conversion device, wherein the learning processing means is selected in accordance with a predetermined priority order for. 6. The kana-kanji conversion device according to claim 1, wherein selection information for determining a learning method is stored in the kana-kanji conversion dictionary or separately from the kana-kanji conversion dictionary.
It is characterized by comprising selection information storage means for storing in accordance with a combination of a plurality of conversion target attributes, and the learning method control means selects a learning processing means based on selection information for the combination of conversion target attributes. Kana-to-Kanji conversion device. 7. The kana-kanji conversion device according to claim 1, wherein the plurality of learning processing means include learning processing means of a non-learning method, learning processing means for performing temporary latest use learning, and temporary learning processing means. A kana-kanji conversion device including at least one of a learning processing means for performing frequency learning, a learning processing means for performing permanent latest use learning, and a learning processing means for performing permanent frequency learning. 8. The kana-kanji conversion device according to claim 1, further comprising learning method selection information registration means for registering correspondence information between the attribute of the conversion target selected by the user and the learning processing means to be used. Kana-to-Kanji conversion device. 9. The kana-kanji conversion device according to claim 1, wherein a plurality of types of selection information patterns for determining the learning method according to the attribute of the conversion target are prepared in advance according to the purpose of use. A kana-kanji conversion device having means for allowing a user to select one of a plurality of types of selection information patterns.