JPH0748211B2 - Kana-Kanji converter - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a kana-kanji conversion device that converts a sentence consisting of a kana character string into corresponding kanji to generate a kanji-kana mixed sentence.

[Conventional technology]

In order to efficiently perform Kana-Kanji conversion for Japanese with many homophones, the conventional Kana-Kanji conversion device learns this Kanji character by phrase each time it determines the Kanji to be converted, and outputs the conversion candidate Kanji at the time of the subsequent conversion. When doing, the Kanji used last is given priority and output. Furthermore, not only words but also adjacent clauses are learned as a set to improve conversion efficiency.

The conversion procedure in the conventional Kana-Kanji conversion device gives a higher score to a Kanji when the last Kanji is used, and preferentially converts to a Kanji with a higher score during conversion. Higher priority.

In other words, "This was the reason why he was sick."
When the kana kana character string is converted from the beginning of the sentence and the kanji kana mixed sentence "The Minister of Health and Welfare has determined the composition" is obtained, the priority of the kanji corresponding to "Kosei" is "composition>welfare>
… ”

In addition, in a kana-kanji conversion device that learns adjacent clauses as one set, if you learned "welfare" and "ministry" as one set, kana-to-kanji conversion for a kana-kanji character string If you do
At the time of converting "Kousei", the adjacent words "daijin" are searched as one set, and the corresponding words are searched. Since "Minister of Health and Welfare" has been learned, it is converted to "Minister of Health and Welfare". At that time, the priority of the single clause of "Kousei" is "welfare>composition>
… ”

[Problems to be Solved by the Invention]

By the way, in a kana-kanji conversion device that learns simple phrases, when converting a sentence in which homonyms are used alternately, such as "The Minister of Health and Welfare has determined the composition", the last word used after each conversion. Since the priority of is 1st,
When deciding "welfare" first and converting the next "kousei", "welfare" is output as the highest priority conversion candidate, and the "composition" required this time has a priority of 2nd or lower. . Therefore, the operator has to press the conversion key until “composition” is output, and there is a problem that the conversion efficiency of homonyms is low.

A kana-kanji conversion device that learns adjacent bunsetsu as one set was devised for the purpose of compensating for the above-mentioned drawbacks caused by learning of single bunsetsu, but since it is configured to store each word combination, There is a problem that the effect of learning cannot be utilized if the alignment is different. For example, when the priority of “send” is “gift> send”, when kana-kanji conversion such as “send her” is performed, “she / send” is learned as one set, and the priority of a single phrase is “send>"Give".
Next, when kana-to-kanji conversion such as "gift present" is learned, "present / gift" is learned, and the priority of a single phrase is "gift>send". After the above learning,
When converting "Kare wo suru" to kanji, "she / send"
Has been learned, but "he / send" with a different combination has not been learned, so "send him" is converted into "give him" preferentially.

The present invention has been made to solve such a problem, and stores a word and a concept of the word in association with each other, and determines whether or not the word corresponding to the word should be converted. An object of the present invention is to provide a kana-kanji conversion device having a high kanji conversion efficiency by giving the highest priority to the conversion order of words having the same concept as the concept of a dependency word.

[Means for Solving the Problems]

The present application is a kana-kanji conversion device that converts a sentence composed of a kana character string into a word containing kanji corresponding to kana to generate a kanji-kana mixed sentence. The words "he", "send / gift" and the concept of this word " And a storage means (8) for storing the concept of "person" or a word having a dependency relationship with this word, "thing (send), person (send), thing (gift)" in association with each other. The word "he" having a dependency relationship with "kaku" that should be converted is specified, and the concept "person" of the specified word "he" is searched from the storage means (8) and converted. The storage means (contains the concept of a word that has a dependency relationship with the conversion candidate word “send / gift” corresponding to Ikarana “send” “thing wo (send), person wo (send), thing wo (gift)” 8), and the conversion candidate unit in which the retrieved two concepts are the same Characterized in that it comprises a priority priority means converting ranking of "Send" (7).

Further, according to the present invention, when the conversion order of conversion candidate words is determined according to the conversion frequency in the past, one conversion candidate word “send” has a plurality of concepts “a thing (send) and a person (send). When equipped with, means for learning for each concept.

[Action]

The kana-kanji conversion device of the present application is a kana that should be converted.
The word "he" having a dependency relationship with is identified, the concept "person" of the identified word "he" is searched, and the conversion candidate word "send / gift" corresponding to the kana "send"
Of the conversion candidate word "send" in which the two concepts found are searched by searching for the concept "object wo (send), person wo (send), object wo (gift)" Prioritize ranking.

〔Example〕

Hereinafter, the present invention will be described with reference to the drawings illustrating an embodiment thereof.

FIG. 1 is a block diagram showing the configuration of the kana-kanji conversion device of the present invention. In the figure, reference numeral 1 is an input unit including a keyboard and the like for inputting instructions such as kana character strings and kana-kanji conversion.
Kana character strings and instructions input from the input unit 1 are input to the control unit 2.
The control unit 2 controls the data exchange of each unit of the device according to the given instruction, and outputs the input kana character string to the output unit 3 including a CRT display or the like.

The conversion candidate creating unit 4 divides the input kana character string into phrases by using the kana-kanji conversion dictionary 5 that stores words corresponding to kana characters and their priority information, and corresponds to kana characters of the phrases. The kanji candidates and their priority information are read out and stored in the conversion candidate table 6.

The dependency analysis unit 7 refers to the conversion candidate table 6 and
The dependency of the input sentence is analyzed, the word / concept dictionary 8 in which the word and the concept of the word are stored in association with each other is searched, and the conversion candidate word corresponding to the conversion candidate and the conversion candidate kana We identify the concept of words that are in the receiving relationship, and assume the score of the conversion candidate according to this concept. The conversion candidate selection determining unit 9 refers to the conversion candidate table 6 to combine the conversion candidates, and outputs the conversion candidates to the output unit 3 in descending order of priority, that is, the conversion candidate having the highest assumed score. The output unit 3 sequentially displays the conversion candidates with the highest priority according to the Chinese character conversion instruction given from the input unit 1.

When the conversion is determined, the dependency analysis unit 7 analyzes the dependency of the bunsetsu of the converted input sentence again, and stores the determined word, the concept, and the score assumed during the conversion in the learning buffer 10. At that time, if the target word has been learned in the past, the score is updated.

A kanji-kanji conversion procedure performed by the kana-kanji conversion device having the above configuration will be described using a specific example.

Note that the word / concept dictionary 8 is in a state as shown in the conceptual diagram of FIG.
The concept that each has, that is, the part corresponding to "~" in "~ wo", which is a relationship between "send" and "depend,""send(send),""send(send),""send (send) Are stored in association with each other. Also, the word "she",
“He” and “present” are stored in association with the corresponding concepts “person”, “person”, and “thing”.
In this embodiment, one word and one concept are stored in association with each other. However, for example, a word such as “put” is not only “to” but also “to” and “to”. Since there are multiple concepts such as "to," if you combine these to store "(person), (thing) to (person), (gift)", you can narrow down the conversion candidates to a smaller number. be able to.

Further, the learning buffer 10 is in the state shown in the conceptual diagram of FIG. 3, and learns “send”, “send” and “gift” and their respective scores, which are classified according to the concept, as conversion candidates corresponding to “put”. Suppose

When the Kana-to-Kanji conversion of a kana character string that has been input as "Kanojo" is instructed, the conversion candidate creation unit 4 uses the Kana-Kanji conversion dictionary 5 to convert the input sentence to "Kanojo / It is divided into clauses such as "Okuru" and words and priority information corresponding to each clause are stored in the conversion candidate table 6.

FIG. 4 is a conceptual diagram of the conversion candidate table 6, and the conversion candidate having a higher score in each clause has a higher conversion priority.

The dependency analysis unit 7 refers to the candidate table 6 to analyze the dependency between “Kanojo” and “Okuru”, searches the word / concept dictionary 8 and performs conversion corresponding to “Okuru”. Search for candidate concepts and "she" concepts.

As a result of the search, "send" is the only conversion candidate having the concept "person" of "she" as the concept.

Therefore, it is assumed that the other conversion candidates “send” and “gift” have lower scores than “send”, and the learning buffer 10 is referred to based on this assumption.

That is, without this assumption, the priority is “gift>send> send”. However, based on the above assumption, "send" is given priority, and the priority is "send>give>send>".

Therefore, "send her" is the first conversion candidate.

After that, the conversion candidate selection determining unit 9 determines the conversion candidate table 6
, The conversion candidates are combined, and the conversion candidates having the highest priority are output to the output unit 3. The output unit 3 sequentially displays the conversion candidates with the highest priority according to the instruction given from the input unit 1.

When the operator determines the conversion to "send her", the dependency analysis unit 7 analyzes the dependency between the words of the converted input sentence and stores the concept and priority information of each determined word in the learning buffer. Store in 10. At that time, if the target word has already been learned, the contents of the learning buffer 10 are updated.

FIG. 5 shows the learning buffer in which the contents shown in FIG. 4 are updated.
In the 10 conceptual diagrams, the points assumed at the time of conversion are added to the points of "send".

Next, if you input something like "Send a present", the input sentence will be divided into phrases according to the same procedure as above,
The conversion candidates corresponding to each clause are stored in the conversion candidate table 6. Further, the dependency is analyzed to search for a word corresponding to the conversion target kana and a concept of a word having a dependency relationship with the kana. As a result of the search, it is assumed that the score of "send" that does not have the concept "object" of "present" as its concept is lower than that of other conversion candidates. When the learning buffer 10 is referred to under this assumption, the priority is “gift>send> send”, and as a result, “gift a present” is output as the first conversion candidate.

〔The invention's effect〕

The kana-kanji conversion device of the present invention identifies a word having a dependency relationship with a kana to be converted, and identifies the concept of the identified word and a conversion candidate word corresponding to the kana and a dependency relationship word. By comparing with the concept and prioritizing the conversion order of the conversion candidate words of the same concept, it is possible to prioritize the conversion order of an appropriate word according to the sentence and improve the kanji conversion efficiency.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a kana-kanji conversion device of the present invention, FIG. 2 is a conceptual diagram of a word / concept dictionary, FIGS. 3 and 5 are conceptual diagrams of a learning buffer, and FIG. 4 is a conversion candidate table. It is a conceptual diagram of. 1 ... Input part, 2 ... Control part, 3 ... Output part 4 ... Conversion candidate creation part, 5 ... Kana-Kanji conversion dictionary 6 ... Conversion candidate table, 7 ... Dependency analysis part 8 ... Word・ Concept dictionary, 9 …… Conversion candidate selection decision unit 10 …… Learning buffer

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Toru Asaeda, Toru Asaeda 2-18 Keihan Hon-dori, Moriguchi-shi, Osaka Sanyo Electric Co., Ltd. (56)

Claims (2)

[Claims] 1. A kana-kanji conversion device for converting a sentence consisting of kana character strings into a word containing kanji corresponding to kana to generate a kanji-kana mixed sentence. A storage means (8) for storing the concept of a related word in association with each other, and specifying a word having a dependency relationship with the kana to be converted,
The specified concept of the word is searched from the storage means (8), and the concept of the word having a dependency relationship with the conversion candidate word corresponding to the kana to be converted is searched from the storage means (8). The kana-kanji conversion device further comprises: priority means (7) for prioritizing the conversion order of the conversion candidate words having the same two retrieved concepts. 2. When determining the conversion order of conversion candidate words according to the conversion frequency in the past, when one conversion candidate word has a plurality of concepts, a means for learning for each concept is provided. The kana-kanji conversion device according to claim 1.