JPH05342194A - Continued kana/kanji clause conversion system - Google Patents

Abstract

PURPOSE:To smoothly input Japanese by improving accuracy for the converted result of Japanese syllabary (KANA)/Chinese character (KANJI) conversion. CONSTITUTION:Word information corresponding to KANA character strings inputted from a data input part 1 through an index input part 2 is read from a dictionary file 8 and outputted to a clause extraction part 4 by a dictionary word read part 3. Based on the word information from the dictionary word read part 3, the clause extraction part 4 prepares 'clause' candidates and outputs the 'clause' candidates to a priority evaluation part 5. The priority evaluation part 5 evaluates the priority of the 'clause' candidates from the clause extraction part 4 by using the use frequency information of 'independent words' among those 'clause candidates' and the 'clause' candidate having the highest point is decided and outputted to a converted result output part 6. The converted result output part 6 displays the 'clause' candidate from the priority evaluation part 5 through a data output part 7 onto a screen.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a continuous sentence kana-kanji conversion system, and more particularly, to segment break processing in a continuous sentence kana-kanji conversion system.

[0002]

2. Description of the Related Art Conventionally, in a bunsetsu kana-kanji conversion system, the length of a "bunsetsu" composed of a retrieved "independent word" and an "adjunct word" attached to this independent word is targeted for priority evaluation.

Since the dictionary provided by the Japanese input system is constructed for the general public, a wide variety of words are stored. Therefore, the dictionary stores a large number of words that are not needed by individual users or that are rarely used. As a result, the retrieved "independent word" is often irrelevant.

However, if priority evaluation is performed only by the length of the "bunsetsu" composed of "independent word" and "adjunct word", the validity of the retrieved "independent word" cannot be directly evaluated. Therefore, an incorrect “bunsetsu” is created, which causes erroneous conversion.

Further, the longer the "bunsetsu" is set, the more advantageous it is in priority evaluation. Therefore, the "adjunct word" is unnecessarily connected, which is a cause of erroneous conversion.

[0006]

SUMMARY OF THE INVENTION The present invention has been made to eliminate the above-mentioned problems of the conventional ones, and can improve the accuracy of the conversion result of the kana-kanji conversion and facilitate the Japanese input. The purpose is to provide a kanji-kana-kanji conversion system.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The continuous sentence-kana-kanji conversion system according to the present invention is a continuous sentence-kana-kanji conversion system for converting a hiragana character string input in a continuous sentence clause into a kanji-kana kanji mixed character string. Storage means for storing each of the words, and creation means for creating evaluation information for determining a segment break for the hiragana character string based on the usage frequency information stored in the storage means; Means for separating the bunsetsu from the hiragana character string based on the evaluation information created by the creating means.

[0008]

An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, when a Hiragana character string is input from the data input unit 1, the Hiragana character string is output to the dictionary word reading unit 3 via the index input unit 2. Here, when the hiragana character string is input from the data input unit 1 by roman character input or kana input, the headline input part 2 converts the character string input by roman character input or kana input into hiragana notation.

The dictionary word reading unit 3 reads word information corresponding to the character string in hiragana notation (hereinafter referred to as a headline) received from the headline input unit 2 from the dictionary file 8 and outputs it to the phrase extracting unit 4. Here, since the dictionary file 8 stores word information regarding “independent words”, the dictionary word reading unit 3 can read all “independent words” that can be read from the dictionary file 8 by the headline received from the headline input unit 2. Search for "word".

The phrase extracting unit 4 creates "bunsetsu" candidates based on the word information received from the dictionary word reading unit 3. That is, the phrase extracting unit 4 determines whether or not the next character can be connected to the “independent word” searched by the dictionary word reading unit 3, that is, whether or not the next character can be an “adjunct”. Is checked and "bunsetsu" created based on the check result
The candidates are output to the priority evaluation unit 5.

The priority evaluation unit 5 evaluates the priority of the "bunsetsu" candidate received from the phrase extraction unit 4 by using the usage frequency information of the "independent word" in the "bunsetsu" candidate. The priority evaluation unit 5 determines a “bunsetsu” candidate to be adopted by selecting a combination that gives the highest points obtained from the usage frequency information of each word, and outputs the “bunsetsu” candidate to the conversion result output unit 6. Output. The conversion result output unit 6 displays the “bunsetsu” candidates received from the priority evaluation unit 5 on a screen (not shown) via the data output unit 7.

FIG. 2 is a diagram showing an example of Roman character input from the data input unit 1 of FIG. 1, and FIG. 3 is a data input unit 1 of FIG.
It is a figure which shows the example of kana input from. In these figures, the text "Go to Miyazaki today" is shown as "KYOUMIYAZAKI MADEIKU" in romaji, and "Kyoumiyazaki" as kana.

FIG. 4 is a diagram showing an example of Hiragana notation by the index input unit 2 of FIG. In the figure, the headline input unit 2
From the data input section 1 "KYOUMIYAZAKI M
"ADEIKU" and "Kyomiyazaki Made"
Convert to Hiragana notation such as "Go to Kyomi and Zaki".

FIG. 5 is a diagram showing an example of processing by the dictionary word reading unit 3 and the phrase extracting unit 4 of FIG. FIG. 5A shows an example of word information obtained by the dictionary word reading unit 3 based on the Hiragana notation of FIG. 5 (b) is shown in FIG.
An example of "bunsetsu" candidates created by the phrase extraction unit 4 based on the word information example of (a) is shown.

In FIG. 5 (a), the hiragana notation "Kyomi and Zakigo Iku" in FIG. 4 is replaced with "ki", "kyo",
"Today", "Today", "U", "Umi",
"Mi", "Miya", "Miyazaki", "Ya", "Yazaki", "Za", "Ma", "To", "De", "I",
It shows a case where the search is performed by dividing the heading information into "ku". According to this search, the word notation information and the usage frequency information corresponding to each of the above-mentioned headline information are read from the dictionary file 8 by the dictionary word reading unit 3 and output to the phrase extracting unit 4.

In FIG. 5 (b), "bunsetsu" candidates created by only the "independent word" indicated by the word notation information in FIG. 5 (a) or by connecting the immediately following character to the "independent word" are selected. Shows. These "bunsetsu" candidates are output to the priority evaluation unit 5 together with the corresponding "independent words" and usage frequency information.

FIGS. 6 to 11 show examples of the results of priority evaluation by the priority evaluation unit 5 of FIG. In these figures, the heading information indicating how the hiragana notation “Kyomi and Zakigo Iku” in FIG. 4 is divided, corresponding “bunsetsu” candidates and usage frequency information, and the usage frequency information are accumulated. The total points thus obtained, the number of clauses indicating how many hiragana notations are divided, and the priority consisting of the average value of the usage frequencies obtained by “total points / number of clauses” are shown.

In FIG. 6, the hiragana notation “go to Kyomi and Zaki” in FIG. 4 is replaced with “ki”, “yo”, “u”,
"Mi", "ya", "za", "ki", "ma", "de",
It shows the case where it is divided into "i" and "ku". In this case, the total points obtained by accumulating the usage frequency information corresponding to each “bunsetsu” candidate is “47”, and the number of phrases is “11”.
Therefore, the priority is 47/11 = 4.3.

In FIG. 7, the hiragana notation “go to Kyomi and Zaki” in FIG. 4 is replaced with “Kyo”, “Miyazaki”,
It shows the case where it is divided into “up” and “go”. In this case, since the total points are “33” and the number of phrases is “4”, the priority is 33/4 = 8.3.

FIG. 8 shows a case where the hiragana notation “Kyomi and Zaki to go” in FIG. 4 is divided into “Kyo”, “to Miyazaki” and “Iku”. In this case, since the total points are “28” and the number of phrases is “3”, the priority is 28/3 = 9.3.

FIG. 9 shows a case where the hiragana notation “Kyomi and Zaki to go” in FIG. 4 is divided into “Kyo”, “to Miyazaki” and “Iku”. In this case, “Kyo” which is used less frequently than “Today” selected in FIG. 8 is selected for the heading information “Kyo”, so the total points are “21” and the number of phrases is “3”. And the priority is 21/3 = 7.

FIG. 10 shows a case where the hiragana notation “Kyomi and Zaki to Iku” in FIG. 4 is divided into “Kyomi”, “Yazaki to”, and “Iku”. In this case, the frequency of use of "interest" is high, but the frequency of use of "Yazaki" is low, so the total points are "25" and the number of phrases is "3".
And the priority is 25/3 = 8.3.
The pattern shown in FIG. 10 is an output pattern often seen in the conventional kanji-kana-kanji conversion.

In FIG. 11, the hiragana notation “Kyomi and Zaki go to” in FIG. 4 is replaced with “Kyomiya”,
The case is divided into "za", "ki", "up", and "iku". In this case, the first phrase is taken as long as possible, but since "Zaki" does not exist in the dictionary file 8, the second and subsequent phrases cannot be well separated, and a word with a low frequency of use must be selected. ing. Therefore, since the total points are "39" and the number of phrases is "5", the priority is 39/5 = 7.8.

In the processing results shown in FIGS. 6 to 11, since the priority of the delimiter shown in FIG. 8 is the highest, the “bunsetsu” candidate shown in FIG.
Is displayed on the screen via the data output unit 7.

As described above, the priority evaluation unit 5 determines the segment breaks by using the usage frequency information stored in the dictionary file 8 corresponding to each word information, thereby giving priority to words with high usage frequency. Can be used. Therefore, the accuracy of the conversion result of the Kana-Kanji conversion can be improved, and Japanese input can be smoothly performed.

[0027]

As described above, according to the present invention, use frequency information indicating the number of times each word is used is stored,
Based on this stored usage frequency information, create evaluation information for determining the punctuation of the hiragana character string, and perform the punctuation of the hiragana character string based on the created evaluation information. With this, it is possible to improve the accuracy of the conversion result of the kana-kanji conversion, and it is possible to smoothly input Japanese.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is a diagram showing an example of Roman character input from a data input unit in FIG.

3 is a diagram showing an example of kana input from the data input unit of FIG.

FIG. 4 is a diagram showing an example of Hiragana notation written by the index input unit of FIG. 1.

5 is a diagram showing a processing example by a dictionary word reading unit and a phrase extracting unit of FIG. 1;

6 shows an example of a result of priority evaluation by a priority evaluation unit of FIG.

FIG. 7 shows an example of the result of priority evaluation by the priority evaluation unit of FIG.

8 shows an example of a result of priority evaluation by the priority evaluation section of FIG.

9 shows an example of a result of priority evaluation by a priority evaluation unit of FIG.

10 shows an example of a result of priority evaluation by the priority evaluation unit of FIG.

11 shows an example of a result of priority evaluation by the priority evaluation section of FIG.

[Explanation of symbols]

 2 heading input part 3 dictionary word reading part 4 phrase extraction part 5 priority evaluation part 8 dictionary file

Claims (2)

[Claims] 1. A continuous sentence kana-kanji conversion system for converting a hiragana character string input in a continuous sentence clause into a kanji kana kanji mixture character string, wherein usage frequency information indicating the number of times each word is used is stored in correspondence with each word. Storing means, creating means for creating evaluation information for determining a punctuation of the hiragana character string based on the use frequency information stored in the storing means; and evaluation information created by the creating means. And a means for separating the bunsetsu from the hiragana character string on the basis thereof. 2. The extracting means for extracting, in the creating means, a usage frequency for each break of the phrase based on the usage frequency information of the storage means, and a use for each break of the phrase extracted by the extracting means. An accumulating means for accumulating frequencies and a calculating means for calculating an average value for each segment of the phrase in the hiragana character string from the accumulation result of the accumulating means are provided, and the average calculated by the calculating means The continuous sentence kana-kanji conversion system according to claim 1, wherein a value is used as the evaluation information.