JP2966473B2 - Document creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a document creation apparatus for creating a Japanese document by using kana-kanji conversion, and more particularly, to converting an input Japanese reading character string to a kanji-kana character. The present invention relates to a device for converting a sentence into a mixed sentence.

[Prior Art] A method of inputting Japanese readings and converting them into kanji / kana mixed sentences is described in Japanese Patent Application Laid-Open No. 60-189565. There is a method in which a candidate row with a high degree of frequency is displayed on a display unit, and a Japanese sentence is input while selecting the displayed candidate. According to this publication, a correct sentence can be displayed as a conversion result by a learning method in order to distinguish between homologous parts of "reporter" and "train" such as "news reporter is train".

[Problems to be solved by the invention]

However, such a conventional method has the following problems.

First, a method in which learning is performed only when the user performs a correction operation while confirming the contents of the kana-kanji conversion on the display device. If the learning is not performed, the homonymous part is obtained until the conversion result intended by the user is obtained. Operation is required.

For example, if the conversion result intended by the user is "listen to the lecture" and the conversion result is "listen to the protest", first correct the "protest" to "lecture" and then "listen" to "listen".
It is necessary to perform the correction operation twice, for example, to correct the error.

In addition, since learning is performed with compound words or affixes without decomposing into independent words, the determination of homonymous words in the distinctive word portion is not eliminated.

As described above, according to the conventional method, the homonymous word cannot be determined without the learning information, so that there is a problem that the operation of correcting the conversion result becomes complicated.

An object of the present invention is to solve the problem of the conventional method and reduce the number of candidate corrections by simply correcting the head term of the homonym part and automatically correcting the homonym of the succeeding term. Another object of the present invention is to provide a document creation device equipped with kana-kanji conversion with good operability.

[Means for Solving the Problems] The object of the present invention is to provide a kanji conversion dictionary that stores kanji or a mixture of kanji and kana characters for reading, and a semantic classification that classifies the semantic categories of the notation after conversion of the body language. A stored semantic classification dictionary, a case frame dictionary storing semantic classification in which semantic categories of the nominative dominated by the verbal are stored, input means for inputting Japanese readings, and Means for partitioning the phrase so that the phrase is included, and both the semantic classification of the noun of the noun phrase and the semantic classification of the declinable word of the declinable phrase match based on the semantic classification dictionary and the case frame dictionary. And a character string conversion candidate display means for displaying, on a display device, a character string including a noun and a word converted in accordance with the input reading as a conversion candidate, Conversion When an instruction is given from the user to re-convert the displayed nomenclature in the state where the nominating is performed in the complementary character string, a second A candidate for conversion of a nominative is displayed on the display device, and the semantics already displayed as the candidate for conversion are converted into the same meaning as the conversion candidate for the second nominal based on the semantic classification dictionary and the case frame dictionary. This is achieved by a document creation apparatus characterized in that the document is converted again so as to be classified and displayed on the display device as a kanji or a kanji mixed kana conversion candidate.

[Operation] When the homonym of the preceding term is corrected by the above technical means, the homonym of the following term is automatically changed, so that the number of times of correcting the homonym part can be reduced.

〔Example〕

Hereinafter, a document creating apparatus according to an embodiment of the present invention will be described with reference to FIGS.

FIG. 2 is a block diagram showing the configuration of the document creation device in one embodiment.

1 is a control device such as a microprocessor for controlling the whole operation, 2 is an input device for inputting Japanese reading and Kana-Kanji conversion instruction and the like, 3 is a device for mixing input Japanese reading character strings with Kanji and Kana characters. A kana-kanji conversion device for converting into a sentence, 4 analyzes a connection between morphemes, for example, a kana-kanji conversion dictionary used in morphological analysis processing, and 5 and 6 analyze a modification relation of terms from a result of the morphological analysis processing. In the dictionary used in the syntactic and semantic analysis processing, a semantic classification dictionary of 5 stores semantic classifications that classify semantic categories of the nominative, and a case frame dictionary of 6 stores notations of the verbs, part of speech, case information,
And a semantic classification in which the semantic category of the nominative dominated by the idiom is stored. Reference numeral 7 denotes storage information for storing the created documents, etc., 8 denotes a random access memory (hereinafter referred to as RAM) for temporarily storing kana-kanji conversion results and data, and 9 denotes a conversion candidate network generated by morphological analysis. An area 10 to be stored is a display device for displaying an input reading character string and a kana-kanji conversion result.

FIG. 7 is a keyboard showing an example of the input device 2; 21 is a Japanese syllabary key for inputting Japanese readings; 22 is a conversion key for giving an instruction to convert the input reading into a sentence mixed with kanji or kana; Left cursor key for giving a display to move the cursor displayed on the display device 10 to the left, 24 is a right cursor key for giving an instruction to move the cursor displayed on the display device 10 to the right, 25 is a 10 For example, selecting a candidate from a candidate selection matrix displayed on the display device,
It is a numeric keypad for inputting numbers.

 Next, the operation will be described with reference to FIGS.

First, as a reading for performing Kana-Kanji conversion from the keyboard 2 of the input device shown in FIG. The input reading is temporarily stored in a reading storage area 81 on the RAM 8 shown in FIG. When the reading input is completed, the conversion key 22 is operated to activate the kana-kanji conversion device 3. The operation of the kana-kanji conversion device will be described below with reference to FIG.

In the kana-kanji conversion process, first, a morphological analysis process 31 is performed. In this morphological analysis process, a conversion candidate network 91 shown in FIG. 3, for example, is created while searching the kana-kanji conversion dictionary 4 based on the reading stored in the reading storage area 81, and the conversion candidate network is temporarily stored. It is stored in area 9. The structure / meaning analysis process 32 is performed on all combinations stored in the conversion candidate network temporary storage area 9. The candidate sequence determined to be optimal by morphological analysis is RAM8
To be kept.

As a method of this morphological analysis processing, a method as described in Japanese Patent Application Laid-Open No. 60-189565 is known, and therefore detailed description is omitted here.

The operation of the syntactic and semantic analysis process 32 will be described in detail with reference to FIG.

First, the part-of-speech of the candidate sequence to be analyzed in the syntax analysis processing 321 is stored in the part-of-speech storage table of FIG.
HINSHI-TBL), and the morpheme string is divided into phrases according to the set part of speech information.

Next, in the single sentence extraction process 322, the pattern of the phrase is analyzed from the information divided in the unit of a phrase, and the phrase is divided into a sentence (hereinafter, referred to as a single sentence) including at least one verbal phrase and a body phrase. In the subsequent case frame analysis availability determination processing 323, the phrase information of the delimited simple sentence is examined, and a combination in which the body language and the word are not present in the simple sentence, for example, `` protest chrysanthemum '' is excluded from the target of the case frame analysis processing 324, "Listen to protest" is first sent to case frame analysis process 324. It is determined whether the case frame analysis process matches some of the patterns that are not possible. If none of the patterns match, it is determined that case frame analysis is possible, and case frame analysis processing is performed.
The process proceeds to 324, and if any of the patterns match, it is determined that case frame analysis is impossible, and the process proceeds to process 326.

If it is determined in the determination process 323 that the case frame analysis is impossible, it is checked in the process 326 whether or not there is a candidate column to be extracted next in the conversion candidate network temporary storage area 9. If there is a candidate column of the combination to be extracted, the candidate column is extracted by the next candidate extraction 327, and the process returns to the syntax analysis process 321. When there is no combination of candidate strings to be taken out, when the morphological analysis process 31 shown in FIG.
After the candidate string held in M8 is processed in the processing 33 shown in FIG. 1, the syntax / semantic analysis processing is terminated.

If the case frame analysis possibility determination process 323 can analyze the case frame analysis process 324, the case frame analysis availability flag (hereinafter, referred to as FLAG1) present in the RAM 8 is set to “0”, and the case frame analysis process 322 separates it. The case frame dictionary 6 and the semantic classification dictionary 5 are searched based on the word "listen" of the simple sentence "listen to protest". Analyze the dependency relationship. An example of a dictionary used for analysis is shown in FIG. 9 and FIG.
Shown in the figure. FIG. 9 shows the case frame dictionary 6, and FIG. 10 shows the semantic classification dictionary 5.

As an example of the dependency relationship analysis in the case frame analysis processing 324, it is checked whether or not the semantic classification of the nominative is included in the adjective to be converted. In the case of “listen to protest”, the word “listen” is registered in the case frame dictionary shown in FIG. 9, and “listen” in the case frame dictionary has the meaning classification of “voice, music, negotiation, command”. It is registered.

Also, the protest “protest” is registered in the semantic classification dictionary shown in FIG. 10, and the semantic classification of “protest” in the semantic classification dictionary is registered as “negotiation”. From the two dictionary information, "listen" includes the semantic classification "negotiation" of "protest", so "listen to protest" is determined to be correct as Japanese and RA
The simple sentence “Listen to protest” is stored in the simple sentence storage area (hereinafter referred to as SI-LIST) in M8, and “1” is set in FLAG1.

Next, a determination process 325 of whether or not the analysis is successful in the case frame analysis process 324 is performed.

The judgment process 325 proceeds to the judgment process 326 if FLAG1 = 0, and moves to the rose judgment process 326 where FLAG1 = 1.

In the determination process 328, it is determined whether the processing has been completed for all the single sentences separated in the single sentence extraction process 322, and if the processing has not been completed, the process returns to the single sentence extraction process 322 again. , The entire single sentence stored in the SI-LIST is passed to the process 33 shown in FIG.
The syntactic and semantic analysis processing ends.

After the above-described syntactic and semantic analysis processing 32 is completed, the passed candidate string 501 is set as a conversion result in the character string display buffer 502 in FIG.

In the subsequent process 34, another candidate group registered in the conversion candidate network is set in the display selection matrix 503 and displayed. At this time, the conversion display 501 is set to be in a re-convertible state.

Next, according to the key accepting process 35, if the conversion result is intended by the user, a key for performing the next process is input, and if a candidate for the conversion result is to be corrected, the ten keys 25 in FIG. 7 are input.

In the input key determination process 36, it is determined whether or not the numeric key 25 is used to instruct the conversion candidate to be corrected.

In the case other than the numeric keypad 25, the kana-kanji conversion processing is terminated and the conversion result is determined.

In the key reception processing 35, for example, when the numeric keypad 25 corresponding to the instruction key for correcting “protest” to “lecture” is input, the notation “lecture” is fixed, and the syntactic and semantic analysis of 32 is performed again in the subsequent conversion candidate network. Perform processing.

In the syntactic and semantic analysis process 32, first, the dependency relation is analyzed for the lexical word "listen" and the fixed display "lecture".

As a result, the meaning of "lecture" is classified into "listening"
Is not included, the following idioms “listen” and “lecture” are analyzed. As a result, since "listening" includes the semantic classification "discussion" of "lecture", "listening to lecture" is displayed on the display device 10 as an optimal conversion result as shown in FIG.

With this means, if the homonymous term of the term initially displayed on the display device is corrected to another term for the converted character string converted to Kana-Kanji, the subsequent term is automatically converted to the correct term as Japanese Becomes possible.

〔The invention's effect〕

As described above, according to the present invention, when the first homonym is corrected, the homonym of the succeeding term is automatically converted, so that the number of times of correcting the conversion candidates is reduced, and the document creator with good operability is reduced. Can be configured.

[Brief description of the drawings]

FIG. 1 is a flowchart showing kana-kanji conversion processing in one embodiment of the present invention, FIG. 2 is a system configuration diagram showing one embodiment of the present invention, FIG. 3 is a conversion candidate network showing kana-kanji conversion candidates, FIG. Is a flowchart showing a syntactic and semantic analysis process, FIGS. 5 and 6 are display examples of kana-kanji conversion results, FIG. 7 is an input device such as a keyboard, FIG. 8 is a part-of-speech storage table, and FIG. One example, tenth
The figure shows an example of the semantic classification dictionary. 1 ... Control device, 2 ... Input device, 3 ... Kana-Kanji conversion device, 7 ... Storage device, 10 ... Display device, 25 ... Candidate selection key.

Continuation of the front page (56) References JP-A-63-116269 (JP, A) JP-A-63-65566 (JP, A) JP-A-63-120363 (JP, A) JP-A-1-316863 (JP) JP-A-59-146381 (JP, A) JP-A-2-105968 (JP, A) JP-A-1-181154 (JP, A) JP-A-63-98068 (JP, A) 63-156273 (JP, A) JP-A-1-229369 (JP, A) JP-A-63-47868 (JP, A)

Claims (1)

(57) [Claims] 1. A kanji conversion dictionary that stores kanji or a kanji-kana character string corresponding to a reading, a semantic classification dictionary that stores a semantic category that classifies the semantic category of a notation after conversion of a body language, A case frame dictionary storing semantic classifications that classify the semantic categories of the dominant dominant, input means for inputting Japanese readings, and converting the input readings into phrases so as to include a physical phrase and a word phrase. Means for separating, and means for converting the reading so that both the semantic classification of the nominative of the nominal phrase and the semantic classification of the declinable word of the verbal clause match based on the semantic classification dictionary and the case frame dictionary, A character string conversion candidate display means for displaying a character string including a noun and a word converted in accordance with the input reading on a display device as a conversion candidate, and a nominative character among the displayed conversion candidate character strings. Is in identification When the user receives a command to re-convert the displayed nominative in a state where it can be identified, a second candidate for conversion of the nominative noun is displayed on the display device. Further, the declinable word already displayed as the conversion candidate is converted again based on the semantic classification dictionary and the case frame dictionary so as to have the same semantic classification as the conversion candidate of the second physical word, and the kanji is converted. Alternatively, a document creation apparatus is displayed on the display device as a kanji kana mixed conversion candidate.