JPH05189417A - Japanese word processor - Google Patents

Abstract

PURPOSE:To provide a Japanese word processor which selects only one candidate and obtains the optimum result of analysis even if plural homonym candidates exist in a sentence analyzing process. CONSTITUTION:A relative information dictionary 107 stores the correspondence between the prescribed words and their relative words. An evaluating means 106 retrieves the information out of the dictionary 107 for decision of the homonyms included in the KANA (Japanese syllabary)-KANJI (Chinese character) character strings. Then the means 106 counts the number of relative words for each homonym and outputs the homonym having the largest KANA-KANJI count value with preference. The KANA-KANJI character string which includes the decided homonym candidate is held in an output buffer 108 and then outputted through an output device 109.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Japanese sentence processor for converting a reading character string into a character string containing kana-kanji characters.

[0002]

2. Description of the Related Art Conventionally, in this type of device, when a kana or romanized reading is instructed to be converted to kanji, the kanji corresponding to that reading is searched in a word dictionary and converted into a kana-kanji mixed sentence. It When outputting Japanese, from among many analysis result candidates including homonyms,
It is necessary to select one candidate using some evaluation method. As one of the means, when selecting a homonym, there is a method of preferentially outputting a homonym related to another word in the same sentence from candidates for the homonym. As another means, there is also a method of adding semantic classification information to a word, registering a related semantic classification for each word, and using the information.

[0003]

However, the above-mentioned conventional Japanese sentence processing apparatus has a problem that further analysis cannot be performed when there are a plurality of homonymous synonyms having related words in the analysis sentence.

In particular, in the method of preferentially outputting a homonymous word related to another word in the same sentence from the candidates of the homonym when selecting a homonym, a certain word Using the information of the dictionary (related information dictionary) that stores the related words indicating the related words as a pair, if there are a plurality of homonym candidates having the related words in the sentence analysis process, , It is not possible to determine which homonym candidate should be prioritized, and therefore it is not possible to select the most suitable candidate.

It is an object of the present invention to select a single candidate and obtain an optimum analysis result even if there are a plurality of homonym candidates in the sentence analysis process. To provide.

[0006]

In order to achieve the above object, the first invention is a Japanese sentence processing device for converting a reading character string into a character string containing kana-kanji, and a predetermined word and a word related to the word. And a related information dictionary that stores the related words associated with each other, and when determining the homonyms contained in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched to determine each homophone. The evaluation means counts the number of related words associated with each synonym candidate, and preferentially outputs the homonym synonym candidate having the largest count value.

A second aspect of the present invention relates to a Japanese sentence processing device for converting a reading character string into a kana-kanji mixed character string, in which a predetermined word and a related word indicating a word related to the word are associated and stored. Information dictionary, when determining the homonyms included in the character string related to the Kana-Kanji conversion process, search the information in the related information dictionary, for each homonym candidate, for the related words related, The sum of the values indicating the ease of connection calculated according to the attribute including the positional relationship between the homonym synonym candidates and the related words in the character string is calculated, and the homonym synonym candidate having the largest calculated value is selected. It is provided with an evaluation means for giving priority to output.

In a third aspect of the present invention, in a Japanese sentence processing device for converting a reading character string into a character string containing kana-kanji characters, a predetermined word, a related word indicating a word related to the word, and the strength of the connection between the two are determined. When determining the related information dictionary that stores the associated degree of association and the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched to determine each homonym. The evaluation means calculates the sum of the degrees of association of related words for each word candidate, and preferentially outputs the homonymous word candidate having the largest calculated value.

According to a fourth aspect of the present invention, in a Japanese sentence processing apparatus for converting a Yomimi character string into a character string containing Kana-Kanji characters, a predetermined word, a related word indicating a word related to the word, and the strength of the connection between the two are provided. When determining the related information dictionary that stores the associated degree of association and the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched to determine each homonym. For each word candidate, a value indicating the ease of connection calculated according to the attribute including the positional relationship between the homonymous word candidate and the related word in the character string is set to the degree of association corresponding to the related word. Multiplying, and evaluating means for preferentially outputting a homonym candidate having a maximum calculated value obtained by calculating the sum of the results obtained by these multiplications.

[0010]

In the first aspect of the invention, the evaluation means searches the information in the related information dictionary when determining the homonyms included in the kana-kanji mixed character string, and determines the related words for each homonym candidate. The number is counted, and the homonym synonym candidate having the largest count value is output with the highest priority. Therefore, according to the present invention, the optimum analysis result can be obtained.

Further, in the second invention, the evaluation means determines the homonyms included in the kana-kanji mixed character string,
The information in the related information dictionary was searched and calculated for each homophone synonym candidate according to the attribute including the positional relationship between the homophone synonym candidate and the related word in the character string for the related word. The sum of the values (weights) indicating the ease of connection is calculated, and the homonym candidate having the maximum calculated value is preferentially output. Therefore, according to the present invention, the optimum analysis result can be obtained.

Further, in the third invention, the evaluation means determines the homonyms included in the kana-kanji mixed character string,
The information in the related information dictionary is searched, the sum of the degrees of association of related words related to each homonym candidate is calculated, and the homonym candidate having the largest calculated value is output with the highest priority. .. Therefore, according to the present invention, the optimum analysis result can be obtained.

Further, in the fourth invention, the evaluation means determines the homonyms included in the kana-kanji mixed character string,
An attribute including the positional relationship in the character string between the homonym candidate and the related word in the degree of relevance corresponding to each related related word for each homonym candidate by searching the information in the related information dictionary Multiply the value (weight) indicating the ease of connection calculated according to the above, and calculate the sum of the results obtained by these multiplications, and give priority to the homonym candidate with the largest calculated value. To do. Therefore, according to the present invention, the optimum analysis result can be obtained.

[0014]

Embodiments of the present invention will be described below with reference to the accompanying drawings.

An embodiment of the first invention will be described.

FIG. 1 is a functional block diagram showing an embodiment of a Japanese sentence processing apparatus according to the first invention.

In the figure, the Japanese sentence processing apparatus includes an input device 101 for inputting a reading character string, an input buffer 102 for temporarily storing the input reading character string, and a sentence grammatically correctly for each word. A morphological analyzer 103 that divides and outputs this as an analysis result, a word dictionary 104 that stores information for each word, a grammar dictionary 105 that stores grammatical information necessary for sentence analysis, Evaluation device 106 for evaluating the analysis result and outputting an appropriate analysis result
And related word information in which a certain word and a word (related word) related to the word are registered in word units, and related meaning classification information in which information of a meaning classification related to a certain word is registered. Related information dictionary 107, an output buffer 108 that temporarily stores the output result, an output device 109 that outputs the output result, and a control device 11 that controls each device described above.
It is configured with 0 and.

FIG. 2 shows a related information dictionary 107 according to the present invention.
It is a conceptual diagram which shows the structure of.

In this dictionary 107, as shown in FIG.
Corresponding to the reading 21 and the notation 22 of a homonym as a word, the reading 23 and the related word 24 of the related word as the related word information indicating the word related to the certain word, or the meaning classification related to the certain word. The relation type 26 indicating which of the relation classification 25 as information, the relation word information or the meaning classification information is registered as the relation information for the certain word.
Is registered. It should be noted that when the relation type 26 is “meaning”, the meaning classification information is registered, and when it is “word”, the related word information is registered as related information.

Next, the basic processing operation of the Japanese sentence processing apparatus having the above-mentioned configuration will be described with reference to the flowchart shown in FIG.

As shown in FIG. 3, the character string input from the input device 101 is temporarily stored in the input buffer 102 (step 301). The morphological analyzer 103
Based on an analysis request from the control device 110, a sentence to be analyzed (reading character string) is read from the input buffer 102 (step 302), and the read sentence to be analyzed is each information in the word dictionary 104 and the grammar dictionary 105. While performing the morphological analysis while referring to, the candidate of the analysis result (analysis sentence) is sent to the evaluation device 106 (step 30).
3). When there are a plurality of homonymous synonym candidates included in the analytic sentence (kana-kanji mixed character string) corresponding to the reading character string, the number of candidates for the analysis result is the number according to the number of homonymous synonym candidates. The evaluation device 106 selects an optimum candidate from the candidates of the analysis result by an evaluation method (details will be described later) using information such as part-of-speech, frequency, and related word information (step 304). The candidates of the analysis result are temporarily stored in the output buffer 108 (step 305) and then output through the output device 109 (step 306).

Next, the procedure of evaluating the analysis result candidates by the evaluation device 106 will be described with reference to the flowchart shown in FIG.

As shown in FIG. 4, in the evaluation device 106,
It is determined whether or not it is necessary to evaluate the received analysis result as a candidate of the analysis result based on the related information (related word information or semantic classification information) (step 40).
1). When it is necessary to evaluate with related information,
This is when there are a plurality of candidates for a certain word (generally a homonym) contained in the analysis sentence (character string mixed with kana-kanji) corresponding to the same reading character string. In other words, it is when there are a plurality of analysis sentence candidates.

When it is necessary to evaluate the analysis sentence (analysis result) by using the information of the related words in the related information dictionary 107,
It is checked whether or not there is a word candidate (for example, a homonym) having related information in the analysis sentence (step 4).
02). When the word candidate exists, it is determined whether or not the related word corresponding to the word candidate exists in the analysis sentence (step 403). If there is a related word, it is determined whether there are a plurality of word candidates in which the related word exists (step 404). When there are a plurality of word candidates, the number of all related words existing in the analysis sentence is counted for each word candidate, the count values are compared, and the word candidate having the largest comparison value is detected. That is, the information of the word candidate having the most related words is preferentially used as the related information (step 405). An evaluation is performed using this related information (step 406), and as a result, it is determined whether or not an optimum candidate has been found (step 407). If an optimum analysis result candidate is found,
The evaluation is completed, and if not found, an optimum candidate is found by another evaluation means (step 408).

Next, the processing operation of the first embodiment of the present invention will be described with reference to specific examples.

Here, for example, consider a case where a character string "Niwatori wa tamago kae kae kae" is converted into a kana-kanji mixed sentence.

The morphological analysis device 103 uses the input buffer 10
When the above character string is received from 2, the morphological analysis is performed using the word dictionary 104 and the grammar dictionary 105, and it is converted into a phrase “Niwatori wa / tamago / kae” (note that the symbol “/” indicates a phrase delimiter). After dividing, each word is searched. In this case, <return>, <return>, <
Multiple homophones such as "hatching>" and "kaesu" are searched. The analysis result is input to the evaluation device 106. The evaluation device 106 that has received this analysis result will perform analysis using the part of speech, frequency, etc. of the word, but there are cases where the first candidate sentence cannot be determined using only this information. In this example, "chicken / egg / return" and "chicken / egg /
The two sentences "Hatchachi" remain as candidates, but both are grammatically correct, and it is not possible to determine which is the first candidate based on the part of speech or grammatical information alone.

In such a case, the information in the related information dictionary 107 is used. Of the above two sentences, the related word of <return> present in one sentence is “egg” (semantic classification = article), while the related word of <hatching> present in the other sentence is “egg”. And "chicken" (semantic classification = person / animal). When both related words are present in this way, the candidates (homogeneous word candidates) cannot be narrowed down only by the presence or absence of the related information. Therefore, in the present invention, a method is adopted in which the numbers of related words in the analysis sentences of the respective candidates are compared, and the candidate having a large number of related words is set as the first candidate. In the case of this example, the
The related word of "return" is only "egg", while the related word of "hatching" has two words "chicken" and "egg" in the analysis sentence. It will be used preferentially, so the sentence "Ken hatches eggs" mixed with Kanji will be selected as the first candidate.

As described above, according to the present embodiment, when there are a plurality of homonyms having related words in the analysis sentence in the sentence analysis process and the candidates cannot be narrowed down any more, the homonyms of the homonyms are not used. An optimal analysis result (analysis sentence) can be selected by comparing the number of related words with respect to the candidate and setting the homonymous word candidate having many related words as the first candidate.

Next, an embodiment of the second invention will be described.

Also in this embodiment, the configuration of the Japanese sentence processor, the basic processing operation, and the analysis result evaluation processing operation are the same as in the first embodiment of the invention. This example and the first
The difference from the embodiment of the invention of FIG.
The process of No. 6 is performed as follows: "For each word candidate (each homonym), a weight calculated according to an attribute including a positional relationship between the word candidate and the related word in the analysis sentence (indicating the ease of connection Value) is calculated, and the word candidate having the largest calculated value is preferentially used as related information. " In addition to the positional relationship, the attributes include the degree of ease of connection between the word candidate (synonyms) and the related words, and between the word candidates (synonyms) and the related words. The degree of connection of particles to be connected, etc. are included.

Therefore, according to this embodiment, when there are a plurality of homonym candidate words having related words in the analysis sentence in the sentence analysis process and the candidates cannot be narrowed down any more, the same nephew word candidates are respectively extracted. Calculate the sum of the weights of related words,
An optimal analysis result (analysis sentence) can be selected by setting the homonym candidate having the largest calculated value as the first candidate.

Next, an embodiment of the third invention will be described.

In this embodiment, the structure of the Japanese sentence processor is basically the same as that of the embodiment of the first invention shown in FIG. 1, but the dictionary information of the related information dictionary 107 is Different. Related information dictionary 1 in this embodiment
A conceptual diagram of the configuration of No. 07 is shown in FIG.

The dictionary 107 in this embodiment has a data structure in which a related particle and a degree of association are added to the data structure as shown in the dictionary 107 in the first embodiment of FIG. That is, it is composed of a phonetic reading 51 and a notation 52, a related word reading 53 and related words 54 as related word information, a related classification 55 as semantic classification information, a related type 56, a related particle 57, and a degree of association 58. It has a data structure. In the related particle 57, a particle (particle connected after the related word) inserted between a certain word and a related word or a certain word and a related classification (semantic classification) is registered. In the degree of association 58, a degree of association indicating the strength of connection between a certain word and a related word or between a certain word and a related classification (semantic classification) is registered.

The basic processing operation of the Japanese sentence processing apparatus is the same as that of the first embodiment of the invention, and the evaluation apparatus 106
The analysis result evaluation process by is basically the same as the process in the embodiment of the first invention, but a part of the process is different.

FIG. 6 shows a processing procedure for evaluating the analysis result by the evaluation device 106 in this embodiment. This processing procedure (steps 601 to 608) is basically the same as the processing procedure shown in FIG. 4, but differs in the following points. That is, the processing of step 405 shown in FIG.
"For each candidate of each word, calculate the sum of the degrees of association for all related words present in the analysis sentence, compare these operation values, and use the word candidate with the maximum operation value as the related information. The process is changed to "priority use" (step 605). The processing of other steps is the same.

Next, the processing operation of the third embodiment of the invention will be described with reference to a specific example.

Here, for example, consider the case of converting a sentence such as "I am a part and a child" to a sentence mixed with kana and kanji.

The morphological analyzer 103 has the input buffer 1
When the character string from 02 is received, morphological analysis is performed using the word dictionary 104 and the grammar dictionary 105, and "I am /
After dividing into bunsetsu: "departo / kotori wo / kaou" (note that the symbol "/" indicates a bunsetsu delimiter), each word is searched. In this example, <buy>, <
A plurality of homonyms such as “keep”, “cross”, and “support” are searched for. The analysis result is input to the evaluation device 106. The evaluation device 106 that has received this analysis result will perform analysis using the part of speech, frequency, etc. of the word, but there are cases where the first candidate cannot be determined using only this information. In the case of this example, “I ’m at a department store /
"I keep a bird / keep it" and "I / at a department store / buy a bird /"
Although two sentences remain as candidates, both are grammatically correct, and it is not possible to determine which is the first candidate based on only the part of speech or grammatical information.

In such a case, the information in the related information dictionary 107 is used. Of the above two sentences, the related word of <keep> in one sentence is “small bird” (semantic classification = animal / organism), while the related word of <buy> in the other sentence is "Department store" (semantic classification = store)
And "small bird" (semantic classification = animal, creature). In this way, when there is a related word that is common to both, the degree of relevance of each degree of relevance is used for evaluation. The evaluation value is the sum of the relevance of all related words that are related to the word candidate (synonyms) and that exist in the analysis sentence, and the evaluation values of each word candidate (synonyms) are compared, The homonyms with the highest evaluation value are preferentially used as related information.

In the case of this example, since the only related word of <keeping> present in the sentence is “small bird”, the evaluation value is “3” which is the same as the degree of association “3” with “small bird”. On the other hand, since the related words of <buy> are two words, “small bird” and “department store”, the evaluation value is the sum of the related degrees, that is, the related degree “3” with “small bird” + the relationship with “department store” The degree “1” = 4. Therefore, since the evaluation value “4” of <buy> is larger than the evaluation value “3” of <buy>, the related information of <buy> is preferentially used, and “I / in the department store / small bird” A sentence mixed with kanji, such as "/ buy", will be selected as the first candidate.

As described above, according to this embodiment, as the related information, the information of the related degree indicating the degree of the related degree of the related word is held, and the related word is included in the analysis sentence at the time of sentence analysis. When a plurality of homonym synonyms candidates exist, the evaluation of each candidate is obtained using the degree of association, and the candidate with a high evaluation value is prioritized, so that the optimum analysis result (analysis sentence) can be selected.

Next, an embodiment of the fourth invention will be described.

In this embodiment, the configuration of the Japanese sentence processor, the basic processing operation, and the evaluation processing operation for evaluating the analysis result are the same as those of the third invention. The difference between this embodiment and the embodiment of the third invention is that the processing of step 605 in FIG. 6 is performed in the following manner: "For each word candidate, the degree of relevance corresponding to all related words, Of the candidate and related words are multiplied by the weights calculated according to the attributes including the positional relationship in the analysis sentence, and the sum of the results obtained by these multiplications is calculated. The word candidate with the maximum calculated value is related information. Is used as a priority ”. The attributes include the same elements as those described in the second embodiment of the invention.

In this embodiment, the evaluation value for the analysis result is not simply the sum of the degrees of association, but the degrees of association are weighted according to some factors such as the distance between the candidate word and the related words. The sum of the values (weighted relevance) obtained is calculated, the word candidates are determined based on the sum, and the optimum analysis result is selected.
Therefore, according to this embodiment, when a plurality of homonyms with related words are present in the analysis sentence during sentence analysis, the evaluation of each homonym candidate is obtained using the weighted relevance. An optimal analysis result (analysis sentence) can be selected by giving priority to a homonym candidate having a large evaluation value.

[0047]

As described above, according to the present invention, even when a plurality of homonymous synonyms candidates exist in the analysis sentence in the sentence analysis process, the count value of the related words and the degree of relevance related to the related words are One homonym candidate is selected based on the sum value, the sum of the weights of related words, or the weight of related words, and the value of the sum of related degrees. As a result of analysis (analysis sentence) can be obtained, a desired kana-kanji character string can be quickly obtained. That is, it is possible to improve the efficiency of document creation / editing processing.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing an embodiment of a Japanese sentence processing apparatus according to the first invention.

FIG. 2 is a conceptual diagram showing a configuration of a related information dictionary in the embodiment according to the first invention.

FIG. 3 is a flowchart showing a basic processing operation of the Japanese sentence processing apparatus in the embodiment according to the first invention.

FIG. 4 is a flowchart showing an evaluation processing procedure of an evaluation result by the evaluation device in the embodiment according to the first invention.

FIG. 5 is a conceptual diagram showing the structure of a related information dictionary in an embodiment according to the third invention.

FIG. 6 is a flowchart showing an evaluation processing procedure of an evaluation result by the evaluation device in the embodiment according to the third invention.

[Explanation of symbols]

101 ... Input device, 102 ... Input buffer, 103 ... Morphological analysis device, 104 ... Word dictionary, 105 ... Grammar dictionary,
106 ... Evaluation device, 107 ... Related information dictionary, 108 ... Output buffer, 109 ... Output device, 110 ... Control device.

Claims (4)

[Claims] 1. A Japanese sentence processing device for converting a phonetic character string into a character string mixed with Kana-Kanji, and a related information dictionary storing a predetermined word and a related word indicating a word related to the word in association with each other. When determining the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched,
A Japanese sentence processing device, characterized in that it comprises an evaluation means for counting the number of related words related to each homonym candidate and preferentially outputting the homonym candidate having the largest count value. . 2. A Japanese sentence processing device for converting a reading character string into a kana-kanji mixed character string, and a related information dictionary storing a predetermined word and a related word indicating a word related to the word in association with each other. When determining the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched,
For each homonym synonym candidate, for the related related words, calculate the sum of the values indicating the ease of connection calculated according to the attributes including the positional relationship in the character string between the homonym synonym candidates and the related words. A Japanese sentence processing device, comprising: an evaluation unit that calculates and preferentially outputs a homophone synonym candidate having the largest calculated value. 3. A Japanese sentence processing device for converting a reading character string into a kana-kanji mixed character string, a predetermined word, a related word indicating a word related to the word, and a degree of association indicating the strength of the connection between the two. And a related information dictionary that stores the corresponding information, and when determining the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched,
Japan, characterized in that it comprises an evaluation means for calculating the sum of the degrees of association for related words related to each homonym candidate, and preferentially outputting the homonym candidate having the largest calculated value. Word and sentence processor. 4. A Japanese sentence processing device for converting a reading character string into a kana-kanji mixed character string, a predetermined word, a related word indicating a word related to the predetermined word, and a relevance degree indicating the strength of the connection between the two. And a related information dictionary that stores the corresponding information, and when determining the homonyms included in the character string related to the kana-kanji conversion processing, the information in the related information dictionary is searched,
For each homonymous synonym candidate, the degree of relevance corresponding to each related related word, the ease of connection calculated according to the attributes including the positional relationship between the homonymous synonym candidate and the related word in the character string. The present invention is characterized by comprising an evaluation means for multiplying by the indicated value and giving priority to the homonym candidate having the largest calculated value obtained by calculating the sum of the results obtained by these multiplications. Text processing device.