JP3099425B2 - Kana-kanji conversion device and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character processing apparatus having a function of converting an input character string into a desired character string, for example, a kana-kanji conversion apparatus. When converting to a place name
The present invention relates to a place name high accuracy conversion method and apparatus capable of improving the conversion accuracy.

[0002]

2. Description of the Related Art A conventional kana-kanji conversion method is a kana-kanji conversion method that uses a dictionary in which reading, notation, and part-of-speech information are combined, and performs analysis between words or sentence units according to Japanese grammar. is there. In this Kana-Kanji conversion method, if the input and conversion target is an address, using a general Kana-Kanji conversion, for example, the place name such as “Surgasis urugo (Suruga-cho, Suruga-shi)” will be replaced by “ Thus, the place name is often converted strangely, that is, erroneously converted.

As a character input method limited to inputting an address, a method that does not use kana-kanji conversion is known. For example, as described in Japanese Patent Laid-Open No. 2-72472, a place name notation and hierarchical information are used. Using a place name dictionary that holds
There is an input method in which the displayed place names are input hierarchically from the upper place name of the place name to the lower place name.

[0004]

When a kana-kanji conversion is performed using a general dictionary and a kanji of a place name is to be input, if the place name having the same phonetic notation (having the same reading and different notation) is used, the notation is changed. It cannot be determined uniquely. For example, "Sendai"
There are two notations of "Sendai" and "Kawauchi" for the reading "Sendai". That is, conventionally, the place name was treated as a single noun, so that it was not possible to distinguish and convert the homophone notation. For this reason, there is a problem that the place name is converted to a place name that does not actually exist, such as "Sendai City, Kagoshima Prefecture".

Another problem with inputting place names and addresses using the kana-kanji conversion method is that correct kanji cannot be obtained if the user misreads place names. There are many place names that read in a special way and place names that can be read in many ways, and the problem is that the place name is not matched with the place name dictionary because the user inputs the reading in error. If the place name reading is narrowed down to one of the original readings, the user cannot know the correct reading of the place name and cannot obtain the desired notation. For example, “Kanazawa (Kanazawa)” is replaced with “Kanezawa”. In some cases, such as when inputting, conversion may not be performed correctly due to small reading fluctuations.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide a kana-kanji conversion method and apparatus for performing conversion that distinguishes homonymic notation of place names.

Another object of the present invention is to provide a kana-kanji conversion method and apparatus for performing conversion by absorbing reading fluctuations.

[0008]

In order to solve the above-mentioned problems, the present invention provides a place name dictionary that holds a hierarchical structure in addition to reading and writing place names, and examines the hierarchical connection of place names during conversion. , Made it possible to convert homonymic place names correctly.

According to another aspect of the present invention, a swing correction table for converting an input reading into another reading is provided, and at the time of reading matching with a dictionary, a match search is performed with reference to the shake correction table. By doing so, I was able to absorb the fluctuation of reading.

[0010] Further, means for storing the conversion result for each layer every time Kana-Kanji conversion is performed, and by referring to the stored contents at the time of the next conversion, priority is given to the place name notation subordinate to the recently used upper place name. Made it possible to convert.

[0011] Further, a configuration is adopted in which a place name dictionary and a dictionary other than a place name can be referred to at the same time, and when the conversion result of an input reading string is converted to the lowest place name, the remaining readings after that are converted. Made it possible to search dictionaries other than place names and create candidates.

[0012]

The operation of the present invention will be described with reference to the block diagram shown in FIG.

The reading of the place name input from the input unit 101 is passed to the control unit 103. The control unit 103 cuts out one or a plurality of characters from the beginning, and passes the characters to the notation search unit 104. The notation search unit 104 stores the notation for the reading passed from the control unit 103 in the dictionary unit 1
05, and returns the compatible notation and its evaluation value to the control unit 103. The evaluation value will be described later. The control unit 103 extends the parse tree based on the matching notation and the evaluation value passed from the notation search unit 104, compares the evaluation for each path, and
Display unit 102 as a conversion candidate in order from the route with the highest evaluation value
Pass to

The above-described evaluation values will be described. In the search of the dictionary unit 105 by the notation search unit 104, the notation corresponding to the reading is searched, and the evaluation value for each notation is determined from the connection relationship with the previous phrase. Control unit 103
Is compared with the hierarchical data for each notation obtained from the dictionary unit 105 as follows. (1) When the connection relationship (hierarchical relationship) between the upper hierarchical place name and the candidate notation matches, the first priority is given to the notation and the control unit 10
Pass to 3. (2) If the readings match but the joining relationship with the upper hierarchical place name does not match, the second priority is given to the control unit and passed to the control unit.

When performing a notation search in the dictionary unit 105 from the notation search unit 104, a reference to the shaking correction table 106 held to absorb the fluctuation of the reading is referred to, and the passed reading is determined as an object of the shaking correction table 106. In the case where the reading is included, the dictionary unit 105 is searched as a dictionary search target even if the reading is corrected for fluctuation.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 is a diagram showing a hardware configuration of an embodiment of the kana-kanji conversion device of the present invention. In FIG.
1 is a keyboard for reading and input, 202 is a display for displaying a conversion result, 203 is a CPU for controlling kana-kanji conversion, and 204 is a memory for storing data created by the CPU 203 and place names for each layer of the conversion result. , 205 and 206 are dictionary units for obtaining kanji notation based on readings inputted from a keyboard, and 204 is a ROM dictionary
A (common dictionary) 206 is a floppy dictionary (dictionary for each purpose). The floppy dictionary 206 is composed of a place name dictionary 207 and a personal name dictionary 208.
I have ten. However, the place name dictionary and the personal name dictionary are not limited to the FD dictionary, but may be provided as a ROM.

In the basic block diagram of FIG. 1, the present invention is roughly divided into three procedures. A user reads an address from the input unit 101. The control unit 103 reads the input character string obtained from the input unit 101 and passes it to the notation search unit 104 (first means). The notation search unit 104 searches the dictionary unit 105 while referring to the shake correction table 106 (second means). The control unit 103 creates an analysis tree while viewing the result (third means). As a result, the display unit 102 displays the evaluation values in descending order of the evaluation value in the analysis tree.

FIGS. 3 and 4 show the first means described above.
This will be described based on FIG. FIG. 3 is a flowchart for explaining the extraction of a character string for a match search from the input reading, and FIG. 4 is an analysis example for the explanation.

In order to cut out a reading string to be matched with the dictionary from the input character string 401, one pointer P (402 in FIG. 4) and one variable N (403 in FIG. 4) are set. The pointer P moves on the character string in the input character string and points to the number of leading characters to be cut out.
The variable N holds the number of characters cut out from the pointer. First, the pointer P points to the head of the input character string. The number of characters N holds 1 (step 301). N characters are passed from the pointer P to the notation search unit (step 302).

At this time, the notation search unit that has received the search reading string compares the received reading with the shake correction table shown in FIG. The configuration of the shake correction table is such that a first record for retrieval and two fields of one or a plurality of second read strings that are likely to have read fluctuations with respect to the read form one record, A table is a collection of its records. For example, “Sawa” for the search string “Sawa”, “Sawa” for the search string “Sawa”,
"Butterfly" corresponds to "gusset", "chow" corresponds to "gusset", and "hara" corresponds to "rose" and "straw". The received reading sequence is compared with the shake correction table. In other words, the first reading sequence of the shaking correction table is sequentially matched with the reading sequence for dictionary search, and if they match, a matching search of the place name dictionary is performed, including the place names rearranged in the second reading sequence. . For example, when the input reading sequence is “Yoshitachi Butterfly”, the following “Butterfly” matches the first reading sequence 803 of the shake correction table. Therefore, the place name dictionary is also searched for "Yoshidamachi" in which the second reading string 802 "gusset" corresponding to "butterfly" is rearranged. In other words, the dictionary search for both "yoshidacho" and "yoshidamachi" can be performed by inputting "yoshidacho".

It is checked whether there is a matched character string (step 303), and registration is performed based on the evaluation value (step 304). The registration based on the evaluation value will be described later. Next, 1 is added to the variable N (step 305), and if the number of characters does not exceed (step 306), the read string to be searched is increased by one character, and the dictionary search is performed again. If the number of characters is over,
The pointer P is set at the next cutout position on the analysis path (step 307). Here, the next cutout position is the sum up to the analyzed read character string of the place name in a certain analysis route, and is the analysis end position of the path at that time. If the set position exceeds the input character string, the first candidate is displayed and the process ends.

Next, the second means will be described with reference to FIG. FIG. 5 is a flowchart showing the procedure of the notation search. The character string read from the control unit is searched for a match with the place name dictionary (step 502). If there is no match, an evaluation value of "no candidate" is given, and the notation search is terminated (step 503). If there is, the matching place name is extracted as a candidate place name. Next, it is checked whether or not the pointer P is the head of the input character string (step 50).
4) If it is the top, since the hierarchical structure cannot be used, an evaluation value of "priority 2" is given, and all candidate place names are delivered to the analysis unit in the order of registration. (Step 505). If it is not the top, the immediately preceding place name is referred to as a higher place name, and it is checked whether or not the candidate place name can be joined. Whether or not joining is possible is checked using a parameter indicating hierarchical information, which will be described later. As for the parameter of the immediately preceding place name, the parameter that was previously received and stored at the same time as receiving the reading in step 501 is extracted (step 506). Further, the hierarchical information of the candidate place name is obtained (step 507), and the hierarchical information of the immediately preceding place name is compared with the hierarchical information of the candidate place name (step 508). A method for comparing the hierarchy information between the immediately preceding place name and the candidate place name will be described later. As a result of the comparison, if an inclusive relation is established between the immediately preceding place name and the candidate place name, an evaluation value of “priority 1” is given. An evaluation value of "priority 3" is given to a candidate whose inclusive relation is not established but whose reading matches. The hierarchical information is compared for all the conversion candidates having the same reading, and a priority evaluation value is attached to the conversion candidate, and the result is returned to the control unit.

Next, a method of comparing the above-mentioned hierarchy information between the immediately preceding place name and the candidate place name will be described with reference to FIGS. 6 and 7. FIG. FIG. 6 shows a place name dictionary data structure in the place name dictionary 207, and FIG. 7 is a processing flowchart of hierarchical information.

The hierarchy information is a data structure using coordinate parameters and range parameters. In this embodiment, a case will be described in which two coordinate parameters and one range parameter are provided. In FIG. 6, the coordinate parameters are 601 to 6
08 are the two parameters in parentheses
The range parameter is one parameter to the right of the brackets 601 to 608. The setting of the range parameter gives a larger value to the upper hierarchy, and is set by the hierarchical level of the place name. Since the range is unnecessary at the lowest hierarchical level and it is sufficient to indicate that the range is the lowest, the range parameter gives 0.
The setting of the coordinate parameters is set so as to fall within the range of “coordinate parameter ± range parameter” of the place name of the next higher place.
These three parameters are used as hierarchical information of the place name.
At this time, by adjusting the coordinate parameter and the range parameter, it is possible to set so that the hierarchy of the administrative division can be identified using the range parameter. For example, the range parameter “10” is set as a prefecture and the range parameter “3” is set as a city / ward group.

For example, after "Miyagi prefecture", "Sendai"
Is input, and if conversion candidates for “Sendai” in 603 and “Kawauchi” in 604 exist in the dictionary, the hierarchical information “(10
0, 50), 10 ”to the range (90 to 110, 40 to 6).
The place name candidate existing in 0) matches as a lower place name. Therefore, it is understood that the compatible notation in this case is “Sendai” belonging to this range.

The above hierarchical information comparison processing will be described with reference to FIG.

The position parameters (x, y) and range parameter h of the upper place name, and the position parameters (X, y) of the candidate place name
Y) is obtained (steps 701 and 702). Therefore, the condition “(x−h ≦ X ≦ x + h) and (y−h ≦ Y ≦ y +
h) ”(step 703), an evaluation value of priority 1 is given assuming that an inclusive relation is established between the immediately preceding place name and the candidate place name (step 705), and an evaluation value of priority 3 is given when the condition is not satisfied. (Step 704), the comparison of the hierarchy information ends.

Next, the third means described above will be described with reference to FIG.
This will be described based on FIG.

The control unit creates an analytic tree based on the result of the notation search. The notation searched is 404 to 408 together with the evaluation value.
Is stored as a tree structure. Hereinafter, a method of creating an analysis tree will be specifically described with reference to FIG.

In the case of the character string 401 shown in FIG. 4, "h" first passes to the notation search section, but in this case, it does not exist in the dictionary, and in the case of the character string 401, "chiba" passes. After all, character string 401
In the case of, since the priority 2 is obtained by “CHIKEN”, the path is registered in the parse tree. When n becomes large until the number of input character strings exceeds, the pointer is moved to the next position of the analyzed character (in the case of the character string 401, the fifth character since the fourth character has been analyzed), and similarly, N is set to 1 Notation search is performed while increasing one by one from. After a match for “Chibaken”, the pointer points to “Na” after “Chibaken”. Thereafter, a notation search is performed from "na" to obtain "narashi" and "narashinoshi", but "narashi" does not conform hierarchically and has a priority of 3, and "narashinoshi" hierarchically conforms and has a priority of 1. .
If the pointer reaches the end of the string, priority 1
And the priority 2 are scored and displayed in the order of the scores.

In the extraction of words, the words "ken" and "shi"
Is an effective extraction means, so that it is possible to perform extraction by using “ken” or “shi” as a delimiter of a place name to efficiently perform a notation search.

As in the case of the character string 401 in FIG. 4, even after the conversion to the lowest place name (larger section and town level) is completed by the conversion using the place name dictionary, if the input reading still remains, the lowest place name is used. After the place name, it is often an apartment name or a company name, and it is not appropriate to convert using a place name dictionary thereafter. If the least significant place name ("Okubo" in the case of the character string 401) is registered in the tree by conversion using the place name dictionary, and if the input reading remains (in the case of the character string 401, "Ikeda"), it is automatically The processing for switching the dictionary used for the conversion is performed. After the dictionary is switched, a standard conversion is performed by a built-in ROM dictionary (205) to convert the sentence to a Kanji-kana sentence other than a place name. At this time, a priority of 1 is given to the evaluation value for the candidates that have undergone the built-in standard conversion.

At the time of displaying the conversion result, the analytic tree and the evaluation value given by the relation between the place name and the place name are used. The evaluation value is converted into 3 points for "priority 1", 2 points for "priority 2", and 1 point for "priority 3", and accumulates the evaluation values for each path from the root to the leaf of the analysis tree. The path with the highest cumulative evaluation value is set as the first candidate. In the case of the example of FIG. 4, two analysis paths are generated, but “Nara-shi, Chiba” has priority 2
With (2 points) and priority 3 (1 point), the cumulative evaluation value is 3 points.
The cumulative evaluation value of "Okubo Ikeda Apartment in Narashino-shi, Chiba" is 9 points. After calculating the cumulative evaluation values for all the paths, the path having the highest cumulative evaluation value is set as the first candidate. In the case of the example of FIG. 4, "Okubo Ikeda Apartment in Narashino-shi, Chiba" is the first candidate.

When the user desires a place name of a kanji character other than the first candidate, the next candidate is sequentially displayed by inputting a next candidate display instruction from the input means, and when the desired candidate is displayed, a finalization instruction is issued. To obtain the conversion result.

Each time the kana-kanji conversion result is determined by the user, the place name of the conversion result is stored in the memory 204 for each layer. With this, when there is an input that lacks the input of the upper place name next, that is, when the upper-layer place name is omitted, the place name belonging to the place name is preferentially referred by referring to the upper-layer place name stored in the memory. Can output. For example, when “Ishinomaki-shi, Miyagi” is determined by the user after Kana-Kanji conversion, place name information (coordinate parameters and range parameters) for each hierarchy is stored in the memory as shown in FIG. Next, when the reading of "Sendaishi", in which the place name of the prefecture layer is omitted, is inputted, the coordinate information of "Miyagi prefecture" of 901 stored in the memory is referred to, and the above-described layer information of the layer information is obtained. Using a comparison algorithm, “Sendai City” (Miyagi Prefecture) is given priority over “Kawauchi City” (Kagoshima Prefecture) as a candidate output.

This is particularly effective in the case where addresses in a relatively small area are successively input, for example, in the case of inputting an address that does not require a high-order place name such as a list of neighborhood associations.

[0038]

As described above, according to the present invention, when a user inputs an address by Kana-Kanji conversion, it is possible to convert even a place name having the same phonetic notation into a correct notation.

Further, when a place name whose exact reading is not known is input by a kana, a slight fluctuation of the reading is absorbed by the kana-kanji conversion system, and the input can be performed without the user's awareness.

Further, even if the input of the upper place name is omitted, the operability of the input is improved by implicitly referring to the upper place name. Furthermore, if there is an input reading line after the lowest place name, conversion is performed using a dictionary other than the place name, so enter the apartment name, company name, school name, etc. after the address and convert at once And user operability is improved.

[Brief description of the drawings]

FIG. 1 is a basic block diagram,

FIG. 2 is a hardware configuration diagram of an embodiment;

FIG. 3 is a flowchart of a character string cutout process;

FIG. 4 is an example of place name extraction and parse tree,

FIG. 5 is a flowchart of a notation search process;

FIG. 6 is an example of a place name dictionary data structure;

FIG. 7 is a flowchart of a hierarchy information comparison process;

FIG. 8 is a diagram showing a shake correction table;

FIG. 9 is a diagram showing a conversion result storage area.

[Explanation of symbols]

101: character input device, 102: display device, 103: control unit, 104: notation search unit, 105: dictionary unit, 106 ...
Shake compensation table.

──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Akira Nakajima 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Inside Microelectronics Device Development Laboratory, Hitachi, Ltd. (72) Inventor Yasumasa Matsuda Totsuka-ku, Yokohama-shi, Kanagawa 292 Yoshidacho, Ltd.Microelectronics Device Development Laboratory, Hitachi, Ltd. (72) Inventor Shigeki Taniguchi 1-1-1, Tagacho, Hitachi, Ibaraki Pref. Hitachi, Ltd. Taga Plant, Hitachi, Ltd. (72) Inventor Hirofumi Kashiwa 1-1-1 Taga-cho, Hitachi City, Ibaraki Prefecture Inside the Taga Plant, Hitachi, Ltd. (72) Inventor Toyoichi Suzuki 1-1-1, Taga-cho, Hitachi City, Ibaraki Prefecture, Ltd. Inside the Taga Plant, Hitachi, Ltd. (72) Invention Person Masaki Kawase 1-1-1, Taga-cho, Hitachi City, Ibaraki Prefecture Inside the Taga Plant of Hitachi, Ltd. (72) Inventor Souichi Ishikawa 1-1-1, Tagacho, Hitachi City, Ibaraki Prefecture Inside the Taga Plant, Hitachi, Ltd. (72) Inventor Toshimichi Suzuki 1-1-1, Tagacho, Hitachi City, Ibaraki Prefecture Hitachi, Ltd. (56) References JP-A-62-140164 (JP, A) JP-A-1-184573 (JP, A) JP-A-59-49244 (JP, U) (58) Fields investigated (Int) .Cl. ⁷ , DB name) G06F 17/21-17/27

Claims (8)

(57) [Claims] 1. A method for inputting a reading character string to be converted into kana-kanji characters.
Input means to input and the first word for performing Kana-Kanji conversion
Kana-Kanji conversion to convert Kana-Kanji characters into Kana-Kanji characters
Means, the reading character string and the kana-kanji character
Output means for outputting a candidate string as a result of the conversion.
Show the hierarchical structure of place names in the Sukana Kanji conversion device
The first information and the second information indicating the joining relationship between the place name hierarchies.
A second method for performing Kana-Kanji conversion of place names with information
The second dictionary is provided with one or more coordinate parameters including one or more elements representing a position, and one or more representing a range based on the position represented by the coordinate parameters. A kana-kanji conversion device characterized by expressing a hierarchical structure of place names by a range parameter of 2. An input means for inputting a reading character string to be subjected to kana-kanji conversion, a first dictionary for kana-kanji conversion, first information indicating a hierarchical structure of place names, and a A second dictionary for performing kana-kanji conversion of place names having second information indicating a joining relationship, kana-kanji conversion means for converting a reading character string into kana-kanji characters using the dictionary, and kana-kanji characters by the kana-kanji conversion means Means for determining, as a first term complementary sequence to be output, a candidate sequence in which the relationship between the candidates matches the second information of the second dictionary among the converted candidate sequences, the read character string and the read character string A kana-kanji conversion device, comprising: a kana-kanji conversion device that outputs a candidate string that is a result of kana-kanji conversion of a kana-kanji character. Kanji conversion device. 3. A method according to claim 1, wherein the kana-kanji conversion candidate string based on the first dictionary and the second dictionary is a candidate string that cannot be determined by the determination means. 3. The kana-kanji conversion device according to claim 2 , further comprising a second determination unit that determines a candidate string that matches the second information of the second dictionary as a first candidate string to be output. 4. An input means for inputting a reading character string to be subjected to kana-kanji conversion, a first dictionary for performing kana-kanji conversion, and joining between first information indicating a hierarchical structure of place names and a place name hierarchy. A second dictionary having second information indicating the relationship;
A first kana-kanji conversion device comprising: a kana-kanji conversion means for converting the reading character string to a kana-kanji character; and an output means for outputting the reading character string and a candidate string which is a result of the kana-kanji conversion of the reading character string. And the second
A kana-kanji conversion method for the kana-kanji conversion of the read character string by simultaneously using the dictionaries. 5. An input means for inputting a reading character string to be subjected to Kana-Kanji conversion, a first dictionary for performing Kana-Kanji conversion, and joining between first information indicating a hierarchical structure of place names and a place name hierarchy. A second dictionary for performing kana-kanji conversion of place names having second information indicating a relationship, kana-kanji conversion means for converting the read character string to kana-kanji using the first dictionary and the second dictionary, In a kana-kanji conversion device, comprising: a determination unit that determines an output order of a candidate string that is a result of converting a reading character string into a kana-kanji character; and an output unit that outputs the reading character string and the candidate string. Kana-Kanji conversion characterized by determining a candidate column whose relationship between the candidates matches the second information of the second dictionary among the candidate sequences converted by the Kana-Kanji conversion, and defining the candidate column as a first candidate sequence to be output. Method. 6. An input means for inputting a reading character string to be subjected to kana-kanji conversion, a first dictionary for performing kana-kanji conversion, and joining between first information indicating a hierarchical structure of place names and a place name hierarchy. A second dictionary of place names having second information indicating a relationship, a unit for converting a read character string into kana-kanji characters using the first dictionary and the second dictionary, and a candidate sequence converted into kana-kanji characters by kana-kanji conversion means. Of which, the relationship between the candidates is the second
Means for judging a candidate that matches the second information of the dictionary as the first candidate string to be output, and outputting the read character string and a candidate string that is a result of the kana-kanji conversion of the read character string. A kana-kanji conversion method comprising: storing a place name candidate selected by an operator in the storage means for each hierarchy in a kana-kanji conversion device comprising means and storage means for storing place names for each hierarchy. 7. An input means for inputting a reading character string to be subjected to kana-kanji conversion, a first dictionary for performing kana-kanji conversion, and joining between first information indicating a hierarchical structure of place names and a place name hierarchy. A second dictionary of place names having second information indicating the relationship, means for converting the read character string to kana-kanji using the first dictionary and the second dictionary, and kana-kanji conversion by the kana-kanji conversion means First determining means for determining a candidate whose relationship among the candidate strings matches the second information in the second dictionary and setting the candidate as a first candidate string to be output, and a place name selected by the operator Storage means for storing place names for each of the hierarchies, and a first candidate string for outputting a candidate string having a relation between the place name candidates stored in the storage means and the second information in the second dictionary Second determining means for determining the Output means for outputting a candidate string as a result of converting the read character string into kana-kanji characters. In the kana-kanji conversion apparatus, if the candidate string cannot be determined by the first determination means, the second determination means should be used. Kana-Kanji conversion method characterized by 8. An input means for inputting a reading character string to be subjected to Kana-Kanji conversion, a first dictionary for performing Kana-Kanji conversion, and a junction between first information indicating a hierarchical structure of place names and a place name hierarchy. A second dictionary for performing kana-kanji conversion of a place name having second information indicating a relationship, means for converting a read character string to a kana-kanji character using the first dictionary and the second dictionary, and the read character string And a kana-kanji conversion device that outputs a candidate string that is a result of the kana-kanji conversion of the read character string. The kana-kanji conversion apparatus first performs conversion using the second dictionary. A kana-kanji conversion method characterized in that when the character string is converted to the lowest order, the next succeeding character string is converted using the first dictionary.