JPH04256162A - Word processor with learning function - Google Patents

Abstract

PURPOSE:To improve the learning function accuracy of a word processor which performs the KANA (Japanese syllabary)/KANJI (Chinese character) conversion with the batch conversion of continuous paragraphs. CONSTITUTION:A word processor which contains 8 learning function is provided with 8 dictionary means 1 which stores many words corresponding to the Japanese reading of the words together with the sorting information, a sorting-based dictionary 2 which stores with correspondence secured between the sorting information on a converted word end that on one place of a paragraph preceding the former word by one position, a learning means 3 which stores a selected word end the sorting information on the word of the paragraph preceding the selected word by one place into the means 2 when a homonym following the first place is selected in a homonym selection state, and a learning conversion means 4 which compares the word react out of the means 1 end its sorting information with a word contained in the means 2 end its sorting information and outputs the coincident word of the means 2 as the first conversion candidate if the coincidence is obtained from the comparison when the conversion is instructed with input of the Japanese reading of the word.

Description

[Detailed description of the invention]

0001

[Field of Industrial Application] This invention relates to a document processing device.
More specifically, the present invention relates to a document processing device that has a so-called learning function that rearranges once selected words as first candidates when converting them to have the same pronunciation.

0002

2. Description of the Related Art Conventional document processing devices equipped with a learning function are configured to learn only words selected as next candidates. Specifically, for example, if the kana character string "Zou no Hana" is incorrectly converted to the word "Zou no Hana" and the next candidate word "Zou no Hana" is selected, "Hana" will be changed to "Zou no Hana". ”, and the next time you convert it with the same reading, it will be “nose”.
is output as the first candidate.

0003

[Problem to be Solved by the Invention] However, with the above-mentioned conventional learning function, if you next input and convert the kana character string ``Amaryllis no Hana'', the above-mentioned learning function works and the character string ``Amaryllis no Hana'' is input and converted. ”, which caused the inconvenience of being erroneously translated and requiring time and effort to correct. This invention was made in consideration of the above circumstances, and when the next candidate is selected, the selected next candidate word and the classification information of the word of the previous clause are learned as a pair of information. This provides a document processing device with a learning function that improves conversion accuracy in the learning function and reduces the number of key operations required for selecting the next candidate.

0004

[Means for Solving the Problems] FIG. 1 is a block diagram showing the basic configuration of the present invention. In the figure, the present invention is a document processing device that performs kana-kanji conversion by batch conversion of continuous phrases, and includes a dictionary means 1 that stores a large number of words corresponding to the pronunciations of the words with classification information attached thereto, and a dictionary means 1 that stores converted words and A classified dictionary means 2 stores the classification information of the word in the clause immediately before the word in correspondence with the classification information of the word, and the selected word ,
learning means 3 for storing the classification information of the word of the clause immediately before the word in the classification dictionary means 2; Comparing the word and the classification information of the word with the word and the classification information of the word in the classification dictionary means 2,
This is a document processing device with a learning function, comprising a learning conversion means 4 which outputs the matching word in the classification dictionary means 2 as the first conversion candidate when the comparison results match.

[0005] As a second configuration of the present invention, the classification dictionary means 2 stores three combinations of the converted word, the classification information of the word of the clause immediately before the word, and the particle. Further, the learning means 3 includes a means for
When the first or subsequent homophones are selected in the homophone selection, the classification information of the word in the front clause, the particle, and the word in the rear clause in the two selected word clauses are respectively stored in the classification dictionary means 2. The learning conversion means 4 further comprises means for storing the information in the learning conversion means 4.
When inputting word pronunciation information and instructing conversion, the word read out from the dictionary means 1, the classification information of the word, and the particle are compared with the stored contents in the classification dictionary means 2. When the results match, the classification dictionary means 2
It is also possible to further include means for outputting a matching word among the conversion candidates as the first conversion candidate. In the present invention, the learning function according to the above-described basic configuration and the second configuration may be used separately, or both may be used selectively.

[0006]

[Operation] According to this invention, if the next conversion candidate is selected,
The next conversion candidate word and the classification information of the word in the previous clause are stored and accumulated in the classification dictionary means 2 as a pair of information, and then the reading of the same word is input and converted. When the key is pressed, the word read from the dictionary means 1 and the classification information of the word are compared with the word and the classification information of the word in the classification dictionary means 2, and when they match, the word and the classification information of the word read from the dictionary means 1 are The matching word is output as the first conversion candidate. Therefore, even if there are many conversion candidates with the same reading, the accuracy of the learning function is increased because the conversion candidates are output after checking whether the connections between the clauses in the continuous clause are correct, and as a result, the number of erroneous conversions is reduced. Improves the efficiency of document editing.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below based on embodiments shown in the drawings. Note that this invention is not limited by this. FIG. 2 is a block diagram showing the configuration of a document processing device that performs kana-kanji conversion. In the figure, reference numeral 10 denotes a keyboard, on which various editing keys such as character keys for inputting kana character strings as pronunciations of words, conversion/next candidate keys, etc. are arranged. Reference numeral 11 denotes a memory composed of ROM, RAM, magnetic disk, etc., which stores programs and data necessary for the kana-kanji conversion process. 1
2 is a CPU that functions as a learning means and a learning conversion means, and processes character strings input from the keyboard 10, such as whether they are character strings to be converted to kana-kanji or control character strings for conversion, cancellation, or document editing. It discriminates character strings based on the content, and controls temporary storage of data, activation of kana-kanji conversion processing, display of processing results, etc. as necessary.

Reference numeral 13 denotes a kana-kanji conversion processing unit that cooperates with the CPU 12, and converts a kana character string to be converted into kana-kanji characters inputted from the keyboard 10 into a kanji-kana mixture sentence. Reference numeral 14 denotes a dictionary memory as a dictionary means, which stores word information such as pronunciations and spellings of words, classification codes (information), etc. necessary for kana-kanji conversion. Reference numeral 15 denotes a display device as an output means, which is composed of a display unit such as a CRT or liquid crystal. 16 is a classification dictionary means for embodying the associative learning process that is a feature of this embodiment.
This is a word pair table buffer for learning pairs of word classification codes and words.

FIG. 3 shows part of the contents of the dictionary memory 14 described above, in which word notation and classification code are associated. Figure 4 shows the above classification/word pair table 1.
6, in which classification codes and word codes are stored in association with each other. FIG. 5 shows the concept of classification codes. In FIG. 5, the classifications are as follows: plants-trees, plants-flowers, animals-mammals, animals-fishes, and classification codes classified in this way are attached to each word in the dictionary memory 14. There is.

In the above configuration, the learning function according to this embodiment will be explained below based on the flowcharts shown in FIGS. 6 and 7, and with reference to the display example shown in FIG. When a character string is input from the keyboard 10 (step 3
0→31), the CPU 12 determines that this character string is a character string to be converted, stores it in the memory 11 (step 32), and displays display information indicating that it is a character string to be converted on the screen of the display device 15. Output visually. Specifically, kana string “
The same process is repeated for the string of characters input after the first character "I" of "Inu wa Hana Ga Good".
The information shown in FIG. 8(1) is displayed on the screen (steps 34→35).

In this state, when the conversion key for starting kana-kanji conversion is pressed on the keyboard 10 (step 37),
Determine the number of operations (step 38), and if it is 1 time, CP
U12 starts the kana-kanji conversion processing unit 13 (step 3
9) The kana-kanji conversion processing unit 13 refers to the dictionary memory 14, generates several candidates for character strings containing kanji and kana, and selects the combination with the highest priority as the first combination based on the information stored in the dictionary memory 14. Candidate. Here, the classification/
Refer to the memory contents of the word pair table buffer 16 (
In step 40), if there is no matching result among the combinations of conversion results (step 41), the first candidate for kana-kanji conversion is directly output. At the same time, the combination of conversion results is stored in the classification/word pair table buffer 16, and the conversion results as shown in FIG. 8(2) are displayed. If applicable, classify/word pair table buffer 1
The pair within 6 is selected as the first candidate (step 43).

Next, when the clause movement key is pressed in the display state shown in FIG. 8 (2), the homophone selection clause can be specified for each clause of the displayed converted character string (step 4
4 → 45). In this conversion example, it is necessary to correct "hana" to "nose", so the clause movement key will be pressed twice. When the next candidate key on the keyboard 10 is pressed in this state, the CPU 12 outputs the second candidate for the corresponding phrase from the combination of conversion results stored in the memory 11 (steps 38→46), thereby As shown in 3), a result is obtained in which the word "hanaga" is corrected to "hanaga".

[0013] When the second or subsequent clause candidates are selected using the next candidate key, the CPU 12 refers to the classification/word pair table buffer 16 and stores the word of the clause selected as the next candidate and the previous clause in this table buffer 16. If the classification code of the word of the clause, that is, the classification/word pair is not memorized, it is memorized (steps 47→50).
. In this example, the classification of “dog” is “08”.
However, the word "nose" is also stored in the classification/word pair table buffer 16.

Furthermore, in this state, when the kana character string to be converted, ``gazing at the morning glory flowers'' shown in FIG. Since the classification/word pair consisting of is not stored in the classification/word pair table buffer 16, "nose" obtained by normal word learning is output as the first candidate as shown in FIG. 8(5).

In this state, if you perform the same operation as in FIG. 8(3), that is, move the phrase twice and press the next candidate key to select "hana".
According to the classification shown in FIG. 5, "morning glory""02" and the word "hana" are respectively stored in the classification/word pair table buffer 16 (steps 51→54).

[0016] Here, when the character string "Elephant Hana Hanagai" to be converted shown in Fig. 8 (7) is input and the conversion key is pressed, the conversion result shown in Fig. 8 (8) can be obtained immediately ( Steps 53→55). In other words, with the conventional word-only learning function, "hana" was learned in the previous conversion, so "hana" would be output as the first candidate, but in this example, the classification/word pair table buffer 16, the animal-mammal pair "08" and "nose" is registered for learning (this is the content learned in "dog's nose"), so it appears to the operator as if it had been learned associatively. A correct conversion result can be obtained as the first candidate. Therefore, for the character string shown in FIG. 8(9), the character string shown in FIG.
As shown in 0), correct conversion results can be obtained immediately.

As is clear from the above description, in the present invention, the results of associative learning are obtained by giving higher priority to the learning contents of classification and word pairs than the conventional learning of words only. Further, FIG. 9 shows a part of the classification/particle/word pair table buffer as a second embodiment of the classification/word pair table buffer. As shown in the figure, the buffer contains classification codes, word codes,
It is structured so that particles are stored in correspondence with each other. Using a buffer with such a configuration, Figure 8 (3)
In the process, “08” is the classification code for “dog”.
The three combinations of ``no'' as a particle and ``nana'' as a word are respectively stored in the classification/particle/word pair table buffer.

When the conversion key is pressed, the classification code, particle, and word in the classification/particle/word pair table buffer are referenced. Therefore, even when inputting and converting the kana character string "Zou ga Hana wo Tateru" as shown in Figure 10 (1), the classification code, particles, and words are referenced, so the character string shown in Figure 10 (2) As shown in the figure, you can immediately obtain the correct conversion result for "An elephant eats a flower."

In the classification/word pair table buffer of the first embodiment, for example, in the conversion examples "I went to the classroom" and "I told the classroom", correct results cannot be obtained unless the particle is stored. On the other hand, in many cases, such as "I finished my meal," it can be determined based on only the relationship between "meal" and "finished" without limiting the particles. Therefore, it is preferable to use the classification/word pair table buffer of the first embodiment described above and the classification/particle/word pair table buffer of the second embodiment.

[0020]

According to the present invention, a classification code is attached to a word in a dictionary, and when the next candidate is selected in kana-kanji conversion, the classification code of the preceding bunsetsu word and the following word in the two clauses before the cursor is determined. Since the configuration is such that they are stored in the buffer as a pair, the classification code of the word in the front clause, the particle, and the word in the following clause in the two clauses before the cursor are stored in the buffer.
Since the configuration is such that two combinations are stored in the buffer, the number of key operations required to select the next candidate is reduced, thereby making it possible to easily obtain a desired word.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the basic configuration of the present invention.

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

FIG. 3 is an explanatory diagram showing the contents of a dictionary memory in the embodiment.

FIG. 4 is an explanatory diagram showing the contents of the classification/word pair table buffer of the embodiment.

FIG. 5 is an explanatory diagram showing the concept of classification codes in the embodiment.

FIG. 6 is a flowchart showing learning processing in the embodiment.

FIG. 7 is a flowchart showing learning processing in the embodiment.

FIG. 8 is an explanatory diagram showing an example of display by learning processing of the embodiment.

FIG. 9 is an explanatory diagram showing the contents of a classification/particle/word pair table buffer according to the second embodiment.

FIG. 10 is an explanatory diagram showing an example of display by learning processing using a classification/particle/word pair table buffer.

[Explanation of symbols]

1. Dictionary means 2 Classified dictionary means 3 Learning methods 4 Learning conversion means

Claims (2)

[Claims] Claim 1: A document processing device that performs kana-kanji conversion by batch conversion of consecutive phrases, comprising: a dictionary means storing a large number of words corresponding to the pronunciations of the words with classification information; and a converted word and one of the words. Classified dictionary means stores the classification information of the word of the previous clause in association with the classification information, and when the first or subsequent homophone is selected in the homophone selection, the selected word and the one of the word are stored. a learning means for storing classification information of a word of the previous clause in the classification dictionary means; and a word read from the dictionary means and classification information of the word when the pronunciation of the word is input and a conversion instruction is given. learning conversion means which compares the words in the classification dictionary means and the classification information of the words, and outputs the matched word in the classification dictionary means as the first conversion candidate when the comparison results match. A document processing device with a learning function. 2. The classification dictionary means further comprises means for storing three combinations of a converted word, classification information of a word of a clause immediately before the word, and a particle in correspondence, and the learning means However, when the first and subsequent homophones are selected in homophone selection, the classification information of the word in the front clause, the particle, and the word in the rear clause in the two selected word clauses are stored in the classification dictionary. The learning conversion means further comprises means for storing the word, the classification information of the word, and the particle read from the dictionary means when inputting reading information of a word and giving a conversion instruction, according to the classification. Claim 1 further comprising means for comparing the words stored in the dictionary means and, when the comparison results match, outputting the matched word in the classification dictionary means as the first conversion candidate.
The document processing device with a learning function described above.