JP2847985B2 - Kana-Kanji conversion system for phrase segmentation learning information retrieval - Google Patents

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 apparatus for converting a reading character string into a kana-kanji mixed character string.
The present invention relates to a kana-kanji conversion information retrieval method in a kana-kanji conversion device that learns the features of kana-kanji conversion performed by a user and improves the conversion efficiency of the conversion of kana-kanji conversion thereafter.

[0002]

2. Description of the Related Art A kana-kanji conversion device is provided with a phrase separation moving means for moving a phrase separation position of a conversion result to a position desired by a user. The user can obtain the conversion result by the desired segment break by changing the position of the segment break by the segment break moving means and re-converting if the segment break position of the conversion result is different from the desired one. . Similarly, the kana-kanji conversion device is provided with a phrase-separation learning means for storing the phrase-separation position previously changed by the user as position information and reflecting this position information in the subsequent kana-kanji conversion. The conversion efficiency is being improved.

[0003] As an example of such a phrase separation learning means, Japanese Patent Laid-Open Publication No. 1-214967 discloses a system in which a reading character string and a phrase separation position for two phrases that are subject to phrase division correction are stored, and the subsequent Kana-Kanji characters. There is disclosed a character processing apparatus which searches for a read character string and phrase-separation learning information at the time of conversion and reflects the information in a conversion result.

[0004]

However, the phrase segmentation learning information retrieval system used in the conventional character processing apparatus has the following problems.

FIG. 4 is a flowchart showing a processing procedure of a conventional phrase segmentation learning information retrieval system at the time of kana-kanji conversion. First, words used in the flowchart of FIG. 4 will be described. "YOMI" is a reading character string to be converted into Kana-Kanji characters, "YLEN" is the number of characters of YOMI, "HPOSI" is the position of the first character of the reading character string in YOMI for searching the phrase-separation learning information file, "TPOSI" is a parameter indicating the position in the YOMI of the last character of the reading character string for searching the phrase separation learning information file, and "DPOSI" is a parameter representing the phrase separation position specified by the phrase separation learning information.
FIG. 5 is an explanatory diagram showing the contents of a phrase separation learning file used in the phrase separation learning information search method.
In the following, a description will be given of a processing procedure for performing a phrase segmentation learning information search using the phrase segmentation learning information in the state of FIG.

[0006] Now, it is assumed that a reading character string of "I have a waive" is input. Here, YOMI = “I have a buckwheat” and YLEN = 11. From this, a character string for searching for a phrase-separation learning file is generated using YOMI. First, assuming that HPOSI = 1 (step 201), TPOSI = HPOSI (step 202), and the read character string of the search is “O” (step 203). Next, the phrase-separation learning file is searched to determine whether a matching entry exists (step 204). Here, since there is no matching entry, TPOSI = 2 is set (step 20).
6) It is determined whether or not TPOSI has become YLEN (step 207). Here TPOSI is YLEN
Is returned, the process returns to step 203, and the phrase-separation learning information file is searched for using a reading character string of “put”. Since there is no matching entry in step 204, the process proceeds to step 206. Thus, step 2
By continuing the processing of step 206 from step 03, the read character string for searching for a phrase-separated learning file changes to “Oku ni”, “Oku ni”, “Oku ni” and “Oku ni”. Go. When the read character string is “Oku wa Niwa”, since there is a matching entry in step 204, the process proceeds to step 205. In step 205, the phrase separation position information to be used for the subsequent kana-kanji conversion is extracted from the matched phrase separation learning information, and DPOSI = 3. In this example, the segment break position is 3, so in the case of the reading character string of "Oku ni Niwa,"
The phrase break is "Okuni / Haniwa". Similarly, a search is made to see if there is any matching phrase segmentation learning information, and the process ends when HPOSI and TPOSI become YLEN (steps 207 to 20).
9).

As described above, in the conventional phrase segmentation learning information search method, the file search is performed at least for the number of characters (7 characters) and up to 66 times (11 characters) until the learning information "Oku-Niwa" is obtained. Must also be done.
In addition, if all the read character strings for searching for a phrase-separated learning file generated from the input read character string do not exist in the learning information file, all of the 66 searches are wasted. Further, when the contents of the phrase separation learning information file are as shown in FIG. 6, two patterns (“OKA "Naniwa" and "Kuniwa") match, but there is no particular criterion for selecting which one to select. Therefore, the conversion may be performed due to an incorrect segment break, which causes a reduction in the conversion efficiency.

An object of the present invention is to provide a phrase segmentation learning information retrieval system capable of efficiently searching registered phrase segmentation learning information and improving the processing speed of kana-kanji conversion using the phrase segmentation learning information. Aim.

[0009]

According to the phrase separation learning information retrieval method according to the present invention, first, when performing kana-kanji conversion using the phrase separation learning information, the phrase separation learning information stored by the phrase separation learning means is used. Performs continuous phrase conversion internally without using it. Then, from the conversion result, 2
A search for the learning information stored by the phrase separation learning means is performed using the read character string extracted from each phrase, and an internal search result is obtained for the conversion result in accordance with the phrase separation position information of the phrase separation learning information matched by the search. The phrase break position is corrected.

[0010]

The input character string is internally converted into a continuous phrase, and two phrases are extracted. Then, the learning information of the phrase segmentation learning means is searched based on the two phrases, and when there is matching learning information, the phrase segmentation position information corresponding to the learning information is extracted. Next, 2
The number of characters from the beginning of the first phrase of the phrase is compared with the phrase delimiter position information. Here, when the two values do not match, the clause break position is internally corrected according to the clause break position information. Then, based on the phrase segment position information, the preceding paragraph is converted into a single phrase, and the latter is converted into a continuous phrase. Also,
If the two values match, a similar search is performed for the next two clauses. By performing these processes for all the phrases, a kana-kanji conversion result reflecting the phrase segmentation learning information can be obtained. According to this method, the number of read character strings and the number of searches can be reduced as compared with the related art. In addition, since the search is performed on a phrase basis, even if learning information having a partially matched reading character string is registered, it is not searched for and is not converted by an incorrect phrase delimiter.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a kana-kanji conversion apparatus according to the present invention will be described below with reference to a phrase segmentation learning information retrieval system.

FIG. 1 is a block diagram showing a configuration of a kana-kanji conversion apparatus to which a phrase segmentation learning information search method according to the present invention is applied. The kana-kanji conversion device includes an input unit 1 for inputting a kana character string (or a Roman character string, the same applies hereinafter), a kana-kanji conversion unit 2 for converting an input kana character string into a kana-kanji mixed character string, A display unit 3 for displaying a kana-kanji mixed character string output from the kana-kanji conversion unit 2, a kana-kanji mixed character string output by the kana-kanji conversion unit 2, a phrase separation moving unit 4 for moving a kana-kanji mixed character string, A word dictionary unit 5 in which a plurality of character strings and kana-kanji mixed character strings and grammatical information are stored in association with each other.
And a phrase segmentation learning unit that stores, in the phrase segmentation learning file, the phrase segmentation positions after reading the two phrases and moving the phrase segmentation position related to the movement of the segmentation segmentation position by the segmentation segmentation movement unit 4 in the segmentation segment learning file. 6 is comprised.

The kana-kanji conversion unit 2 uses the phrase separation learning information stored by the phrase separation learning unit 6 when performing kana-kanji conversion using the phrase separation learning information in addition to the basic kana-kanji conversion processing. In addition to performing the continuous phrase conversion internally without performing the above, the learning information stored by the phrase segmentation learning unit 6 is searched using the reading character string obtained by extracting two phrases each from the conversion result, and the search result is matched. According to the phrase segment position information of the phrase segment learning information, a process of internally correcting the segment segment position on the conversion result is performed.

Here, the phrase separation learning function of the above-described kana-kanji conversion device will be briefly described. Now, suppose that the user performs a kana-kanji conversion by inputting a character string of "I have a crocodile" to obtain a character string of "There is a haniwa in the back". In the kana-kanji conversion unit 2, a word corresponding to the input read character string is searched in the word dictionary unit 5, and converted into a kana-kanji mixed character string. In this case, if the data having the reading character strings of two phrases included in the input reading character string is not registered in the phrase separation learning unit 6, the data of the phrase separation learning unit 6 will not be used. . As a result, it is assumed that the character string is converted into a character string “There is a garden in the back”. Since this character string is different from the character string desired by the user, the phrase must be corrected. Therefore, the user operates the phrase separation moving unit 4 to move the phrase separation “/” between the two phrases “Oku ni / Niwa” one character to the left and instruct reconversion. The kana-kanji conversion unit 2 converts each of the two phrases and displays the conversion result. Thus, a character string "There is a haniwa in the back" is obtained. The user obtains the desired character string, and performs an operation to determine the displayed character string.
When this operation is performed, as shown in FIG. 5, the kana-kanji conversion unit 2 registers a two-segment reading character string "Oku ni Niwaga" and the segment break position information "3".

Hereinafter, when the user inputs the same character string as the previous time, the kana-kanji conversion unit 2 searches the word corresponding to the input reading character string in the word dictionary unit 5 and converts the word into a kana-kanji mixed character string. In this case, since the data shown in FIG. 5 is registered in the phrase segmentation learning unit 6, the kana-kanji conversion unit 2 replaces the part of "Oku ni Niwa" in the input character string with "Oku ni", "Oku ni". Kana-Kanji conversion is performed at each paragraph break of "Haniwaga". As a result, the user can obtain a character string "There is a haniwa in the back" without correcting or re-converting the segment break.

Next, a description will be given of the processing procedure of the phrase segmentation learning information search method executed by the kana-kanji conversion apparatus of FIG. 1 with reference to the flowcharts of FIGS.

First, terms used in the flowcharts of FIGS. 2 and 3 will be described. “BUN” is a value of a set of clauses constituting the result obtained by the continuous clause conversion, “BNUM” is the number of clauses constituting the BUN, “BPO”
"SI" is a position in the BUN for obtaining a reading character string for searching a phrase separation learning information file in units of phrases, and "DPOSI" is a parameter representing a phrase separation position specified by the phrase separation learning information. . Also,
In the following description, the phrase segmentation learning information in the state of FIG. 5 is used.

In FIG. 2, it is assumed that a character string “I have an army” is input. First, continuous articulation conversion is performed without using the segmentation break learning information. As a result,
Assuming that a group of three clauses “OK”, “Niwa” and “Aru” has been obtained, these consecutive clauses are stored in the BUN (step 101). Next, BPOS which is a parameter for obtaining a reading character string from the BUN in units of clauses
I is set to 1 (step 102). And BUN (B
Using the reading character string “Oku ni Niwa” combining the POSI) and BUN (BPOSI + 1), the phrase separation learning file of the phrase division learning unit 6 is searched (step 10).
3) It is determined whether a matching entry exists (step 104). As shown in FIG. 5, the phrase-separation learning file has the learning information “Oku-ni-niwa”. Therefore, a value “3” indicating the phrase-separation position of the phrase-separation learning information is extracted, and DPOSI = 3 ( Step 105). Subsequently, it is determined whether or not the number of characters of BUN (BPOSI) does not match the value of DPOSI (step 106). In step 106, BUN (B
Since the number of characters (POSI) (4) does not match the value (3) of DPOSI, the process proceeds to step 107. In this step, the read character string “Oku ni Niwa” is written in DPOSI
At the front and back at the position. Then, single phrase conversion is performed on "Okuni", and continuous phrase conversion is performed on all the character strings "Haniwa ga Iru" after DPOSI in BUN (step 107). In step 107, the segment break position is internally corrected. Next, BP
It is determined whether the OSI has become BNUM (step 108). At this stage, since BPOSI has not reached BNUM, the loop processing from step 103 to step 107 is repeated. By the way, BUN (BPOS
The phrase-separated learning file is searched using the reading character string "Oku-Niwa-gama" that combines I) and BUN (BPOSI + 1). As described above, the phrase-separation learning file contains the learning information “Oku-ni-niwa”, so the value “3” indicating the phrase-separation position is extracted, and the DPOS
Let I = 3. Subsequently, it is determined whether the number of characters of BUN (BPOSI) matches the value of DPOSI. In step 106, since the number of characters (3) of BUN (BPOSI) matches the value (3) of DPOSI, the BPOSI
Is incremented to change the combination of clauses for search (step 109), and the routine goes to step 108. Since BPOSI has not reached BNUM even at this stage,
The loop processing from step 103 to step 107 is repeated.

Next, BUN (BPOSI) and BUN
The phrase-separated learning file is searched using the reading character string "Haniwa wa I" with (BPOSI + 1).
This time, there is no matching entry, so step 1
From 04, the process moves to step 109, the BPOSI is incremented again to change the combination of phrases for search, and the process moves to step 108. Where BPOSI is BN
Since the UM has been reached, the loop processing ends.

Next, in FIG. 3, BUN (BPOS
The phrase-separated learning file is searched with the one-syllable reading character string of I) (step 110), and it is determined whether a matching entry exists (step 111). In this example, the phrase-separation learning file is searched using the reading of the last one phrase, "Aru". Since there is no matching learning information in FIG. 5, the process ends. If there is matching learning information in step 111, a value indicating the segment break position is extracted and put into DPOSI (step 112). Then, the read character string is divided into front and rear parts at the position of DPOSI, single phrase conversion is performed from DPOSI to front, and continuous phrase conversion is performed to the character string after DPOSI (step 113).

According to the above-described processing, a kana-kanji conversion result reflecting the phrase-separation learning information of "there is a haniwa in the back" can be obtained by three searches.

[0022]

As described above, in the phrase segmentation learning information search method according to the present invention, when a phrase segmentation learning information file is searched, continuous phrase conversion is performed without using segmentation segmentation learning information. Since the phrase-separation learning information is retrieved using the combination of the retrieved phrases, the number of character strings to be retrieved and the number of retrievals can be reduced. In addition, since the search is performed based on the phrase unit reading, even if learning information having a partially matched reading character string is registered, the retrieval is not performed, and conversion is not performed due to an incorrect phrase delimiter. For this reason, the phrase segmentation learning information can be efficiently searched, and the processing speed of kana-kanji conversion can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a kana-kanji conversion device to which a phrase separation learning information search method according to the present invention is applied.

FIG. 2 is a flowchart showing a processing procedure of a phrase segmentation learning information search method according to the present invention.

FIG. 3 is a flowchart showing a processing procedure of a phrase segmentation learning information search method according to the present invention.

FIG. 4 is a flowchart showing a processing procedure of a phrase separation learning information search method at the time of conventional kana-kanji conversion.

FIG. 5 is an explanatory diagram showing the contents of a phrase separation learning file.

FIG. 6 is an explanatory diagram showing the contents of a phrase segmentation learning file.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input part, 2 ... Kana-Kanji conversion part, 3 ... Display part, 4 ... Phrase division moving part, 5 ... Word dictionary part, 6 ... Phrase division learning part

Claims (1)

(57) [Claims] An input means for inputting a kana character string or a Roman character string, a kana-kanji conversion means for converting an input kana character string or a Roman character string into a kana-kanji mixed character string, and an output from the kana-kanji conversion means A display unit for displaying a kana-kanji mixed character string, a kana-kanji mixed character string output by the kana-kanji conversion unit, a kana-kanji mixed character string separating unit, and a kana-kanji mixed character unit for moving the kana-kanji mixed character string. Concerned 2
When performing kana-kanji conversion using phrase-separation learning information in a kana-kanji conversion device equipped with a phrase-separation learning unit that stores the phrase-separation position after reading the two phrases and moving the phrase-separation position as phrase-separation learning information, The continuous phrase conversion is performed internally without using the phrase separation learning information stored by the phrase separation learning means, and stored by the phrase separation learning means by using a read character string extracted two by two from the conversion result. A kana-kanji conversion device for performing a search for the searched learning information, and internally correcting a phrase break position for the conversion result according to the phrase break position information of the phrase break learning information matched by the search. Delimiter learning information search method.