JPH08123806A - Kana-kanji conversion method and device therefor - Google Patents

Abstract

PURPOSE: To provide a KANA (Japanese syllabary)-KANJI (Chinese character) conversion device which has high accuracy of conversion. CONSTITUTION: The words corresponding to the inputted KANA are retrieved out of an independent word dictionary 11 which stores the reading, writing and part-of-speech information on the independent words and an annex word dictionary 12 which stores the reading, writing and grammar information on the annex words. Then connection properties are decided between the retrieved independent and annex words and also between the retrieved annex words based on the part-of-speech information and the grammar information so that the paragraph candidates are generated. The presence or absence of the modifying relation is decided between the generated paragraph candidates based on the part of speech of the first independent word serving as a modifying paragraph, the part of speech or the second independent word serving as a modified paragraph and the modifying relation rule that prescribes the conditions to satisfy the modifying relation between the first and second independent words. Then the priority is decided among the paragraph candidates based on the presence or absence of the decided modifying relation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a kana-kanji conversion method for converting a Japanese sentence given in kana expression into a kana-kanji mixed sentence.

[0002]

2. Description of the Related Art In recent years, Japanese word processors have become widespread as conversion means for inputting reading information of Japanese sentences as kana information and converting it into sentence information containing kana-kanji characters.

In such a Japanese word processor,
When text reading information is entered by entering hiragana or romaji using the keyboard, the operation timing of specific keys that indicate clauses and sentence breaks, or punctuation marks are entered or entered during kana information entry. At the timing when the number of characters exceeds the predetermined number, the process of converting the input kana information to the corresponding kana-kanji mixed notation is performed, and the conversion process result is displayed on a display such as a CRT. I am trying. By repeating this series of conversion processing and display, the user can create a kana-kanji mixed notation for a desired sentence.

In the process of converting the input of kana information into the kana-kanji mixed notation in such a Japanese word processor, it is necessary to be able to accurately convert the kana-kanji notation intended by the user. If the conversion cannot be performed accurately, the user must make corrections for the incorrect conversions, which requires a great deal of effort. Therefore, in the development of the kana-kanji conversion device, when converting the reading into kanji, from the viewpoint of the conversion candidate corresponding to the reading, the word that the user wants to input can be converted as the first candidate. The technology is being developed from.

In the conventional conversion processing, unlike Japanese such as English, there is no habit of separating words into words. Therefore, input kana information is first divided into each word and a clause is recognized. Next, a process of selecting the first candidate from the phrase candidates generated in the phrase recognition process or a process of determining the priority of each phrase candidate is performed. Conventionally, in the latter process, co-occurrence relation information and frequency information as the likelihood of the appearance of the word are used. Also, the modified relational rules applicable to all words belonging to the same part of speech are defined and checked. For example, "Auxiliaries are all modifiable by numbers with no adjuncts""Suffixes are not modifiable by numbers with no adjuncts"
And so on.

However, in some independent words, there are differences in terms of modification and modification. In the past, we didn't check that point, so
There is a problem that conversion errors that select conversion candidates that are grammatically impossible cannot be avoided only by using co-occurrence relation information and frequency information as the likelihood of appearance of the word.

[0007] For example, there are self-supporting words that have a qualifying property by a numeral that does not have an adjunct word and those that do not. For example, in the case of a noun, an example of the former is "union"
"Election district", the latter example is "maximum""tentative", the verb example is "add" and "is", the latter example is "fly", "stand", and adjective The examples of the former are "small" and "large", and the examples of the latter are "painful" and "spicy". For example, "union"
Is modified by a number with no adjunct such as "10 unions participated", but "maximum" is not modified by such a number.

Conventionally, since the information on the modificativity of each independent word by the number is not used, "a number without an adjunct modifies an independent word with no modifier by a number without an adjunct. Expression was allowed. Therefore, this may cause a conversion error. An example thereof is shown as conversion candidate 1 in FIGS. 46 (a) to 46 (c).

[0009] In addition, when common bunsetsu are composed of nouns with no attached words, there are some common nouns that have modifiers for proper nouns and others that do not. For example, examples of the former are “land” and “religion”, and examples of the latter are “delegation” and “cause”. For example, “Michi” has a qualifying property with respect to proper nouns when constructing a clause by itself without accompanying words, like “Michi Fukuoka City”, but “delegation” does not It does not have such a modifying property.

Conventionally, since the information on the modifier of a common noun to a proper noun has not been used, "a common noun without a modifier does not have a modifier to a proper noun without an accessory. Expression that modifies proper nouns. Therefore, this caused a conversion error.
An example thereof is shown as conversion candidate 1 in FIG.

In addition, the verb has a property of being modified by "adjective or final form of verb or adjective + case particle" ni "" or "adjective or final form of auxiliary verb + case particle" ni "" , There are those that do not have. For example,
Examples of the former are "determined" and "extent", and examples of the latter are "put out" and "tsuku". For example, "determine"
Is "smaller" or "not eaten", such as "verb or adjective adnominal form + case particle" ni "" or "auxiliary adnominal form + case particle" ni ""
However, "put out" does not have such a modification property.

Similarly to verbs, adjectives do not have the qualification of "verb or adjective adnominal form + case particle" ni "" or "auxiliary adnominal form + case particle" ni "" Things and things exist. For example, examples of the former are "no difference" and "nothing else", and examples of the latter are "hot" and "beautiful". For example, "no difference" means
Like "I must be small" and "I must have not eaten", "A verb or adjective adjunct form + case particle" ni ""
Or, it has a modifier property of “an adjunct form of auxiliary verb + case particle“ ni ””, but “hot” does not have such a modifier property.

In the past, since such information on the modified property of a verb and an adjective has not been used, "a verb or an adjective adnominal form or an ending form + a case particle" ni "" can also be used to describe "an auxiliary verb adjunct form or Final form + case particle "ni""
Verbs or adjectives that are not modified by allow the expression to be modified in this way. Therefore, this caused a conversion error. An example thereof is shown in FIG.
It is shown in the conversion candidate 1 of (b).

[0014] Further, there are word differences in nouns with respect to the modality of an annotated body.
Although there is a particle "no", which is based on a direct noun, but there are differences in terms of the connectedness modification), information about the difference was not used. Differences in adnominal modification include adnominals, adnominal forms, and the particle "no".
There are “2 to the power of 3 = 8” combinations of items to be modified, and since there is information on whether or not a modification relationship is indispensable for each combination, there are “(2 to the power of 3) × 2 = 16” combinations. There is a possibility.

Conventionally, since the information based on the above combination is not used for the noun modificativity of a noun, “an expression that a noun modifier modifies a noun that is not modifiable by the adnominal modifier” is allowed. Was there. Therefore, this caused a conversion error. An example is shown in FIG.
It is shown in the conversion candidate 1 of (a)-(e). For example, in the example of the conversion candidate 1 in (a), the noun “or” having no modifiability by the adnominal is an expression modified from the adnominal “the”. In the example of the conversion candidate 1 in (b), the noun "large and small" which is not modified by the adnominal form is an expression modified from the adjunct form "select". In the example of the conversion candidate 1 in (c), the noun "or later" that is not modified by the particle "no" is an expression modified from "no". In the example of the conversion candidate 1 in (d), the noun “zone” that is indispensable to be modified in any of the above three items is an expression that is not modified by anything. In the example of the conversion candidate 1 in (e), the noun "hara" that is indispensable to be modified by the particle "no"
However, the expression is not modified by anything.

[0016] Further, there are differences in terms of adnominal modification in nouns (adnominal modification may involve the particle "no" or may directly modify the noun without adjuncts. Despite the differences in terms of sex), no information about the differences was used. The difference in adnominal modification is that there are two combinations of two items when the particle "no" is involved and when a noun is directly modified without an adjunct word, and there is a modification relation for each combination. Since there is information whether it is essential or not, "(2 squared)
× 2 = 8 ”.

Conventionally, since the information based on the above combination is not used for the noun adnominal modifier, "an expression in which a noun makes an incorrect adnominal modifier" is allowed. Therefore, this caused a conversion error. For example,
It is shown in the conversion candidate 1 of FIGS. For example,
In the example of the conversion candidate 1 in (a), the noun "Keiju" having only the adnominal modifier with the particle "no" is used in other expressions. In the example of the conversion candidate 1 in (b), the noun "middle grade", which has only the adnominal modifier with the particle "no" or the property of directly modifying the noun without the adjunct, is used in other expressions. . In the example of the conversion candidate 1 in (c), the noun "honkai", which has neither the adnominal modifier with the particle "no" or the property of directly modifying the noun without the adjunct, is used in other expressions. ing.

Further, although the adverb has word differences regarding whether or not it becomes a constituent element of a noun phrase, information regarding the difference has not been used. for that reason,
"Adverbs that do not form a noun phrase are expressions that form a noun phrase"
Was forgiven. Therefore, this caused a conversion error. An example thereof is shown in the conversion candidate 1 of FIGS. 51 (a) and 51 (b). For example, in the example of the conversion candidate 1 in (a), the adverb "Moshi", which has no noun phrase constituent, is used immediately after the adnominal "Sono" and is used in the expression forming the noun phrase. In the example of the conversion candidate 1 in (b), the adverb “since” having no noun phrase constituent is used immediately after the adnominal “that” and is used in the expression that constitutes the noun phrase.

[0019]

As described above, in the conventional kana-kanji conversion method, in the process of selecting the first candidate from each phrase candidate or the process of determining the priority of each phrase candidate, grammatically In some cases, an unforgivable candidate may be selected as the first candidate or given a high priority.
There is a problem that high conversion accuracy cannot be obtained. As a result, the user of the Kana-to-Kanji conversion device is required to instruct the selection of the next candidate, and further the mental burden is imposed.

The first invention is made in consideration of the above circumstances, and an object thereof is to provide a kana-kanji conversion method with high conversion accuracy.

More specifically, as described above, in the conventional kana-kanji conversion method and device, since each independent word does not have information from the viewpoint of the qualification by the number, "the number is the number There is a problem that it is not possible to avoid "a conversion error that modifies a word that does not have the property to be modified".

The second and third inventions have been made in consideration of the above circumstances, and the conversion accuracy that can avoid the conversion error caused by each word difference of the number modifiers of each independent word. It aims to provide a high-quality kana-kanji conversion method.

Further, as described above, in the conventional kana-kanji conversion method, since each common noun does not have the information from the viewpoint of qualification to proper nouns, "a common noun without proper nouns is , The conversion error that modifies the proper noun "cannot be avoided.

A fourth invention is made in consideration of the above circumstances, and a kana with high conversion accuracy which can avoid a conversion error caused by each word difference of proper noun modifier possessed by each ordinary noun. The purpose is to provide a kanji conversion method.

In addition, as described above, in the conventional kana-kanji conversion method, the verb and adjective are "adjective or verbal or adjective adjunct form + case particle" ni "" or "adjective adjunct or terminating form + There was a problem that the conversion error could not be avoided because it did not have the information from the viewpoint of the qualification by the case particle "ni". The fifth invention is made in consideration of the above circumstances,
Caused by each verb and adjective's "adjective or adjective adjunct form or ending form + case particle" ni "" or "auxiliary adjunct form or or ending form + case particle" ni "" It is an object of the present invention to provide a kana-kanji conversion method with high conversion accuracy that can avoid such conversion errors.

In addition, as described above, in the conventional kana-kanji conversion method, since each noun does not have information from the viewpoint of adjoint modifier, "noun makes an incorrect adnominal modifier". There was a problem that could not be avoided.

A sixth invention is made in consideration of the above circumstances, and is a kana-kanji with high conversion accuracy which can avoid a conversion error caused by each word difference of the connectedness modifier of each noun. The purpose is to provide a conversion method.

Further, as described above, in the conventional kana-kanji conversion method, since each noun does not have information from the viewpoint of adnominal modifier, "the adnominal modifier is modified by the adnominal modifier" There is a problem in that it is not possible to avoid "a conversion error that modifies a word that does not have".

A seventh invention is made in consideration of the above circumstances, and is a kana-kanji conversion with high conversion accuracy which can avoid a conversion error caused by each word difference of adnominal modifier of each noun. The purpose is to provide a method.

Further, as described above, in the conventional kana-kanji conversion method, since each adverb does not have information from the viewpoint of whether or not each adverb becomes a constituent element of a noun phrase, "adverb not forming a noun phrase" However, there is a problem in that it is not possible to avoid "a conversion error that constitutes a noun phrase".

The eighth invention is made in consideration of the above circumstances, and it is a kana-kanji with high conversion accuracy capable of avoiding a conversion error caused by each word difference of the noun phrase composition of each adverb. The purpose is to provide a conversion method.

[0032]

A kana-kanji conversion method according to a first aspect of the present invention relates to an independent word dictionary and adjunct words in which a word corresponding to an input kana is stored as reading and notation for independent words and part-of-speech information. Based on the part-of-speech information and the grammatical information, a search is made from an adjunct word dictionary that stores readings, notations, and grammatical information, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and the adjunct word based on the part-of-speech information and the grammatical information The bunsetsu candidate is determined and the presence or absence of a qualifying relation between the bunsetsu candidates generated is determined by determining the part of speech of the first independent word that becomes the qualifying bunsetsu, the part of speech of the second independent word that becomes the modified bunsetsu, and the first Determination is made based on a modification relation rule that defines a condition for establishing a modification relationship between the independent word and the second independent word, and the priority order of the phrase candidates is determined based on the presence or absence of the determined modification relationship. It is characterized by doing.

In the kana-kanji conversion method according to the second aspect of the present invention, the word corresponding to the input kana is read and written for the independent word and the independent word dictionary that stores the part-of-speech information and the reading, the notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. Whether or not there is a qualifying relationship between the generated bunsetsu candidates and the qualification information indicating whether or not there is a qualification by a number that independently forms a bunsetsu without an adjunct to the independent word and between the number and the independence word. Is determined on the basis of a modification relation rule that defines a condition for establishing the modification relation, and the priority order of the phrase candidates is determined based on the presence or absence of the determined modification relation.

A kana-kanji conversion device according to a third aspect of the present invention comprises an independent word dictionary storing readings, notations and part-of-speech information for independent words, an auxiliary word dictionary storing readings, notations and grammatical information for auxiliary words, and an independent word. For a word, a qualifying memory that stores information about the qualifying property of a number that independently forms a clause without an adjunct, and a qualifying relationship that defines the conditions under which the qualifying relationship is established between the number and the independent word. Modification relation rule storage means for storing rules, word search means for searching a word corresponding to an inputted kana from the independent word dictionary and the auxiliary word dictionary, and an independent word and an accessory word searched by the word search means Based on the information of the phrase candidate generation means for determining the connectivity between the and and the adjunct word and the adjunct word to generate a clause candidate, and the modified property storage means and the modified relation rule storage means. Sentence A modification relationship determination unit that determines whether there is a modification relationship between the phrase candidates generated by the generation unit, and a priority order determination that determines the priority order of the phrase candidates based on the presence or absence of the modification relationship determined by the determination unit. And means.

In the kana-kanji conversion method according to the fourth aspect of the invention, the word corresponding to the input kana is read and written for the independent word and the independent word dictionary for storing the part-of-speech information and the reading, notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. Whether or not there is a qualifying relation between the generated bunsetsu candidates and whether or not each common noun has a qualifying property with respect to a proper noun when forming a bunsetsu alone without accompanying words Judgment is made based on the qualification relationship information that defines the conditions for establishing a qualification relationship between nouns, and based on the presence or absence of the determined qualification relationship,
The priority order of the phrase candidates is determined.

In the kana-kanji conversion method according to the fifth aspect of the invention, the word corresponding to the inputted kana is read and written for the independent word and the independent word dictionary for storing the part-of-speech information and the reading, the notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. Whether or not there is a qualifying relation between the generated bunsetsu candidates, the verb and adjective's adjunct form or final form + case particle “ni” or “auxiliary adnominal form or final form + case particle“ Information regarding the presence or absence of the modifiable property by "ni""and the adnominal or final form of the verb or adjective +
Judgment was made based on the qualifying relation rule that stipulates the condition that a qualifying relationship is established between the case particle "ni" or "adjective or final form of auxiliary verb + case particle" ni "" and a verb or adjective. It is characterized in that the priority order of the phrase candidates is determined based on the presence or absence of a modification relationship.

In the kana-kanji conversion method according to the sixth aspect of the present invention, the word corresponding to the input kana is read and written for the independent word and the independent word dictionary storing the part-of-speech information and the reading, notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. Whether or not there is a qualifying relation between the generated phrase clauses, the qualification information indicating whether or not the noun is categorized by the adjoint, and the qualifying relation between the noun and the noun and the noun It is characterized in that determination is made based on a modification relation rule that defines a condition to be established, and the priority order of the phrase candidates is determined based on the presence or absence of the determined modification relation.

In the kana-kanji conversion method according to the seventh aspect of the invention, the word corresponding to the input kana is read and written for the independent word and the independent word dictionary for storing the part-of-speech information and the reading, notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. Specifies the presence / absence of a qualifying relation between generated phrase clauses, the qualification information indicating the adnominal qualification of each noun, and the condition for the qualifying relation to be established between the noun or case particle “no” and the noun. It is characterized in that it is determined based on the modified relational rules described above, and the priority of the phrase candidates is determined based on the presence or absence of the determined modified relations.

In the kana-kanji conversion method according to the eighth aspect of the present invention, the word corresponding to the input kana is read and written for the independent word and the independent word dictionary for storing the part-of-speech information and the reading, notation and grammatical information for the attached word. Is searched from the adjunct word dictionary that has been stored, and the connectivity between the retrieved independent word and adjunct word and between the adjunct word and adjunct word is determined based on the part-of-speech information and the grammatical information, and bunsetsu candidates are selected. A modifier relation rule that specifies the presence / absence of a modifier relation between generated phrase clauses, the noun phrase constituent element information regarding whether or not each adverb can be an element of a noun phrase, and the condition that an adverb is an element of a noun phrase. Based on the presence or absence of the determined modification relationship,
The priority order of the phrase candidates is determined.

[0040]

According to the first aspect of the present invention, the presence or absence of the modifying relation between the generated bunsetsu candidates is determined by the part of speech of the first independent word that becomes the qualifying phrase, the part of speech of the second independent word that becomes the modified bunsetsu, and the first Is determined based on a modification relation rule that defines a condition for establishing a modification relation between the independent word and the second independent word, and the priority order of the bunsetsu candidates is determined by this result.

As a result, it is possible to finely check the modification relation between the bunsetsu candidates, avoid conversion to an erroneous expression that has been missed in the past, and obtain high conversion accuracy.

In the second and third aspects of the invention, the modifier information indicating the presence or absence of the modifier by a number that independently forms a clause without an adjunct to the independent word and the modifier relationship between the number and the independent word. Based on the modification relation rule that defines the condition to be satisfied, the priority order of the bunsetsu candidates is determined. This allows
It is possible to avoid conversion of an independent word that is not modified by a number into an incorrect expression that is modified by the number.

According to the fourth aspect of the invention, the modifier information indicating whether or not each common noun is modifiable with respect to the proper noun when the clause is independently formed without accompanying words, and the modifier relation between the common noun and the proper noun. Based on the modification relation information that defines the conditions to be satisfied, the priority order of the phrase candidates is determined. This allows
It is possible to avoid conversion of an ordinary noun that has no modifier to a proper noun into an incorrect expression that modifies the proper noun.

In the fifth invention, the qualification of a verb or an adjective by "a verb or an adjective adnominal form or an end form + a case particle" ni "" or "an auxiliary verb adjunct or end form + a case particle" ni "" Modifiable information indicating the presence or absence of "adjective or verb or adjective adjunct form +
Priority of bunsetsu candidates based on the modifier relation rule that defines the condition that a modifier relation is established between the case particle "ni" or "adjective or endive form of auxiliary verb + case particle" ni "" and a verb or adjective To decide. As a result, verbs or adjectives that are not modified by either the verb or adjective adnominal form or the ending form + the case particle "ni" or the verb or adjective adjunct form or the ending particle + the case particle "ni" do not It is possible to avoid receiving the wrong expression.

According to the sixth aspect of the present invention, the modification information that indicates the presence or absence of the modality of a noun to the noun, and the modifier that defines the condition that a noun and a noun and a noun and a noun are established. The priority order of bunsetsu candidates is determined based on the relation rules. As a result, it is possible to avoid conversion of an independent word that is not modified by the adnominal to an erroneous expression modified by the adnominal.

In the seventh invention, based on the modifier relation information indicating the adnominal modifier of each noun and the modifier relation rule defining the condition that the modifier relation is established between the noun or the case particle "no" and the noun. , Decide the priority order of bunsetsu candidates. As a result, it is possible to avoid conversion of the adnominal modifier into an erroneous expression that modifies a word that is not modified by the adnominal modifier.

According to the eighth aspect of the present invention, the bunsetsu candidate is based on the noun phrase constituent element information regarding whether or not each adverb can be an element of the noun phrase and the modifying relation rule defining the condition that the adverb is an element of the noun phrase. Determine priority. This allows
It is possible to avoid conversion of an adverb that does not form a noun phrase into an erroneous expression that forms a noun phrase.

[0048]

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

(First Embodiment) A first embodiment of the present invention will be described.

FIG. 1 is a block diagram showing the schematic arrangement of a kana-kanji conversion device according to this embodiment. The kana-kanji conversion device according to the present embodiment includes an input unit 1, a word search unit 2, and an independent word dictionary 1.
1, an adjunct word dictionary 12, a clause candidate generation unit 3, a connection table 13, a sequence candidate selection unit 4, a syntax determination unit 5, a modification relation determination unit 6, a modification relation rule 14, and an output unit 7.

As shown in FIG. 1, the input unit 1 is connected to the word search unit 2, the word search unit 2 is connected to the phrase candidate generation unit 3, and the phrase candidate generation unit 3 is connected to the sequence candidate selection unit 4. The sequence candidate selection unit 4 is connected to the syntax determination unit 5 and the output unit 7. Further, the modification relationship determination unit 6
Is connected to the syntax determination unit 5. The word search unit 2 includes an independent word dictionary 11 and an auxiliary word dictionary 12, the phrase candidate generation unit 3 includes a connection table 13, and the qualification relation determination unit 6 includes a qualification relation rule 14.

Next, with reference to FIG. 2, an overall flow of processing in the kana-kanji conversion device of this embodiment will be schematically described. The main part of the kana-kanji conversion process in this embodiment is roughly divided into two processes (steps S21 and S22) shown in FIG.

In step S20, a kana character string is input from the input unit 1 such as a keyboard.

In step S21 (first conversion process),
With respect to the inputted kana character string, a process of recognizing the range of the bunsetsu is performed by using information about the connectivity between the independent word and the attached word, information about the connectivity between the attached word and the attached word, and the like. This first
The conversion process of is performed by the word search unit 2 and the phrase candidate generation unit 3 illustrated in FIG. The first processing is, for example, Japanese Patent Application Laid-Open No. 60-72014, Japanese Patent Application Laid-Open No. 3-2.
The processing described in detail in 65061 may be used.

In step S22 (second conversion process),
A process of determining the first candidate (or the first candidate to a desired candidate) is performed for each candidate that has been identified as a phrase by the first conversion process. The second conversion process is performed by the sequence candidate selection unit 4, the syntax determination unit 5, and the modification relation determination unit 6 shown in FIG.

In step S23, the first candidate and the like determined by the second conversion process are output to the output unit 7 such as a CRT by the sequence candidate selection unit 4 shown in FIG.

After such processing is performed, the user
A predetermined operation such as confirming the first candidate output to the output unit 7 as the conversion result can be performed.

Next, the first conversion process and the second conversion process which constitute the main part of the kana-kanji conversion process in this embodiment will be described.

First, the first conversion process will be described.

Reading information to be converted is input from the input unit 1 and sequentially sent to the word search unit 2.

The word search unit 2 refers to the independent word dictionary 11 and the auxiliary word dictionary 12 to extract word candidates. As shown in FIG. 3, the independent word dictionary 11 stores the reading (31), the notation (32), the part of speech (33), the number-qualified flag (34), and the word number (35) for each word of the independent word. Has been done. As shown in FIG. 4, the adjunct word dictionary 12 stores an adjunct word reading (41) and grammatical information (42).

The word candidates extracted by the word searching unit 2 are sent to the phrase candidate generating unit 3. The phrase candidate generating unit 3 refers to the connection table 13 to generate a phrase candidate from the given word candidate, and sends the result to the sequence candidate selecting unit 4.
As shown in FIG. 5, the connection table 13 stores independent word and adjunct word, and adjunct word and adjunct word connection information. In FIG. 5, adjuncts 001, adjuncts 002, ... Correspond to adjuncts such as “ga” and “o”. Figure 5
The value 1 indicates that the connection is possible, and the value 0 indicates that the connection is not possible. The term “endability” indicates whether or not one phrase can be terminated by the independent word or its adjunct word. 1 indicates that it can be terminated, and 0 indicates that it cannot be terminated. There is.

Here, with respect to the kana input examples "30 Saidai""3Mai""4Tai" shown in (a), (b) and (c) of FIG. 6, the independent word dictionary shown in FIG. 11, a structure example of a sequence candidate generated by performing the above-mentioned first conversion process (here, the same process as the above-mentioned publication) using the auxiliary word dictionary 12 shown in FIG. 4 and the connection table 13 shown in FIG. 7, FIG. 8 (a), and FIG. 8 (b), respectively.

The structures of these sequence candidates are sequence numbers (7
01), syntax judgment (702), clause number (703), word number (704), part-of-speech (705), numbe modifier flag (706), modifier grammar information (707), modifier clause (70)
8).

Each item will be described with reference to FIG.

The sequence number is a sequence candidate number. The sequence candidate with sequence number 0 is “30 maximum” (second number in FIG. 6A).
In the candidates, the sequence candidate with the sequence number 1 is “30s” (FIG. 6).
It corresponds to the first candidate of (a).

The syntax judgment is information indicating the syntactic validity of the series candidate, and 0 is set as an initial value.
If this value is -1, it means that there is a clause that cannot be modified. In the case of 0, it means other possibility, that is, the clauses forming the sequence candidate have the modification destination in the sequence candidate (or may be inside or outside the sequence candidate).

The phrase number is the number of the phrase candidate. The phrase "30" is the phrase number 0, the phrase "maximum" is the phrase number 1,
The phrase “year” corresponds to the phrase number 2, and the phrase “substitute” corresponds to the phrase number 3.

The part of speech is a part of speech of an independent word here,
For example, nouns and verbs.

The qualifier modifier flag is information about the qualifier information by a numerical word that independently forms a clause without accompanying words. A value of 0 means that there is no modification, and a value of 1 means that there is modification. This corresponds to the item 34 in FIG. In this embodiment, no value is set for a suffix which always has a value of 1 like a classifier and a suffix which always has a value of 0 as a suffix does not prescribe a modification relation rule described later.

The modified grammatical information is the morphological or syntactical grammatical information of the last word forming a phrase, for example,
Nouns, verb adnominal forms, adjective conjunction forms, case particles "wo"
For example, "no". In this example, the phrase number 0 is “30”.
The modified grammar information of is a number, and the "maximum" modified grammar information of clause number 1 is a noun.

The modifier clause represents the clause number of the clause candidate modified by the word. In addition, -1 is set when the clause to modify cannot exist. However, 0 is set when there is a possibility that the modification destination exists regardless of whether the modification candidate is the phrase number of the candidate phrase, whether it is inside or outside the sequence candidate.

The above-described syntax judgment and the value of the modifier clause are judged and entered by the processing by the syntax judgment unit 5 and the modification relation judgment unit 6 in the second conversion processing described below.

Next, the second conversion process will be described in detail.

The sequence candidate selection unit 4 sends the sequence candidates sent from the phrase candidate generation unit 3 to the syntax determination unit 5. Figure 9
Is a flowchart showing the flow of this processing. In step S901, the number of sequence candidates is set to N. In step S902, j representing a series candidate is set to 0. In step S903, the sequence j is sent to the syntax determination unit 5 first. This process is performed for all the series candidates (step S90).
4, S905).

The syntax judging unit 5 to which the sequence candidate is given is, as will be described later, a syntax judging process (process for judging whether there is a phrase that cannot have a modification destination, based on a predefined modification relation rule. )I do. The value (0 or -1) of the syntax determination is returned to the sequence candidate selection unit 4 for each sequence candidate.

The sequence candidate selection unit 4 selects the first candidate (or the first candidate to the desired candidate) from the sequence candidates based on the result of the syntax determination processing by the syntax determination unit 5, and outputs it to the output unit 7. send. FIG. 10 is a flowchart showing the flow of this processing. In step S1001, the number of sequence candidates is set to N
Set to. In step S1002, j representing a series candidate is set to 0. In step S1003, it is determined whether the syntax determination item of the sequence candidate is 1, and in step S1004 it is 0. If it is 1, step S1005
Is stored in the memory 1, and if 0, step S1006.
Save in memory 2 with. Otherwise (ie -1
In the case of) is stored in the memory 3 in step S1007.
This process is performed for all the streams (step S100).
8, S1009), and then in step S1010, the data is sent to the output unit 7 in the order of the memory 1, the memory 2, and the memory 3.

Here, the above process will be specifically described by taking the input shown in FIG. 6 as an example.

As described above, in the first conversion process,
For example, FIG. 7 shows the processing result by the phrase candidate generation unit 3 obtained for the input “30 years” in FIG. 6A. Given this result, the sequence candidate selection unit 4 sends the sequence candidates to the syntax determination unit 5 in the order of “30 maximum” with sequence number 0 and “30s” with sequence number 1.

The syntax judging section 5 firstly determines the modification relation judging section 6
Using, the qualifying clause of each bunsetsu is obtained by analyzing the qualifying relation of the bunsetsu for each series candidate. For example, FIGS. 6A to 6C.
In the case of the input example of FIG. 11, FIG. 12 (a), FIG. 12 (b)
As shown in, the value of the modified clause (clause number or 0 or -1) is obtained. For example, the sequence number 0 in FIG.
In "30 maximum", since "30" of the clause number 0 cannot modify the "maximum" of the clause number 1, the modifier clause of the clause of the clause number 0 becomes -1.

Next, the syntax judgment section 5 carries out the syntax judgment of each series candidate based on these results and obtains the value of the syntax judgment. For example, in the case of the input examples of FIGS.
1, FIG. 12A, and FIG. 12B, the syntax determination value “−1” is returned for sequence number 0, and “0” is returned for sequence number 1.

The sequence candidate selection unit 4 determines the first candidate or the like based on the value of the syntax determination of each phrase candidate. In this case, the sequence candidate “3 with sequence number 1 whose syntax determination value is 0
"0 generation" -Syntax determination value -1 is sent to the output unit 7 in the order of the sequence candidate "30 maximum" having the sequence number 0.

Thus, as shown in FIG. 6A, the sequence candidate “30s” with the sequence number 1 is the conversion candidate 1, and the sequence candidate “30 maximum” with the sequence number 0 is the conversion candidate. It is output as 2. In addition, similarly, FIG.
As shown in (c), “3 dance” and “4th place tie” are output as conversion candidates 1, and “3 dance” and “4 pain” are output as conversion candidates 2.

The processing of the syntax determining section 5 (and the modification relation determining section 6) will be described in more detail below.

The syntax judging section 5 judges the syntactic validity of each series candidate. FIG. 13 is a flowchart showing the flow of the processing. In step S1101, the total number of clauses of the sequence candidate is set to B, and in step S1102, b representing a clause is set to 0. In step S1103, it is checked whether or not a modifier clause exists in the clause b (whether the term of the modifier clause of the sequence candidate is -1). In the case of -1, in step S1106, the item of syntax determination of the relevant series (702).
To -1. This processing is performed for all the clauses except the rightmost clause in the sequence (step S1104,
(S1105), when completed, returns to the sequence candidate selection unit 4.

Here, the input of FIG. 6A will be specifically described as an example. In the case of this example, the value entered in the item (708) of the qualifying clause in FIG. 11 is in a state of being processed by the qualifying relation determining unit 6 described below. Therefore, in the case of the sequence number 0, the item of the modifier clause is -1, so -1 is entered in the item (702) of the syntax judgment. In the case of the sequence number 1, nothing is written in the syntax judgment, so that it remains 0.

Next, the processing of the modification relation judging section 6 will be described. Here, a process of determining the presence or absence of a qualifying relation between phrases is performed. FIG. 14 is a flowchart showing the flow of the processing. In step S1201, the total number of clauses in the series candidate is set to B, and in step S1202, b representing the clause in the series candidate is set to 0. In step S1203, the processing for determining the modification destination of the phrase b is performed. This determination process is performed for all phrase candidates other than the rightmost phrase (steps S1204 and S1205). For example, the sequence candidate “30 maximum” with sequence number 0 obtained from the input shown in FIG.
Since it is composed of the phrase number 0 “30” and the phrase number 1 “maximum”, the decoration destination of the phrase number 0 “30” is obtained.

The process of checking the modification relation for each clause in step S1203 will be described below. FIG. 15 is a flowchart showing the flow of the processing.

In step S1300, K is initialized to -1, and in step S1301, the total number of modification relation determination rules defining the relationship between the modified clause and the modified clause is set in R, and in step S1302, r representing the modification relation rule is set. Set to 0.

FIG. 16 is an example of a modification relation rule. The part of speech (141) of the last word of the phrase to be modified, the condition (142) to be satisfied by the modified part of speech, the part of speech (143) of the modified phrase, and the target part. The condition to be satisfied by the modified part of speech (144),
It consists of the applicable range (145) of the rule. For example, rule (1) states that "the case particle" ga "modifies an adjective", and rule (2) states that "a continuous verb modifies the next verb". Rule (3) states that "when a numerator forms a bunsetsu alone without an accompanying word, the noun can be modified by the noun having the qualifying property of the noun and the noun is the next bunsetsu. Only in case. "

In step S1303, whether the phrase b is the modified part of speech of the rule r, and in step S1304 whether the phrase b satisfies the part of speech of the rule r, the modified grammar information of the sequence candidate information as shown in FIG. Check 707). If the conditions are met, step S1
Proceed to 305. If the conditions are not met, the next qualifying relation rule is applied. This process is repeated until all rules are applied (steps S1313 and S1314).

In step S1305, b + 1 is set in j representing the phrase to be modified. In step S1306, whether the phrase j is the modified part of speech of the rule r, and step S13
At 07, it is checked whether the phrase j satisfies the part-of-speech condition of the rule r.

If the two conditions are met, step S
At 1308, the clause j is set in the modified clause (708) of the clause b.

If either condition is not satisfied,
The next clause in the sequence candidate is checked as a modification destination (step S1309). In this case, if there is no next clause, 1 is set to k indicating that there is a clause to be modified in the subsequent input. If there is a next clause, it is checked in step S1311 if the applicable range of the rule r is satisfied. If so, the processing from step S1306 is repeated. If not satisfied, the next rule is applied (steps S1313 and S1314).

If none of the rules match, the process advances to step S1315 to check whether k is 1. When k is 1, 0 is added to the modifier clause (708) of clause b.
Is set (step S1316). If k is -1, -1 is set in the qualified clause of clause b (step S).
1317).

Here, the above-mentioned processing will be specifically described by taking a case where a modification relationship is obtained for the input of FIG. 6A as an example.

First, the case of checking the modification relation of the phrase number 0 “30” will be described.

First, the rule (1) will be examined. It is checked whether the phrase “30” is the modified part-of-speech “case particle” in rule (1). Since "30" is a number, this rule does not apply and attempts to apply the next rule.

In Rule (3), "30" satisfies the modified part-of-speech "numerical word". In this rule, since there is no part-of-speech condition, it is checked whether the clause "maximum" on the right side of the next one satisfies the modified part-of-speech "noun". Although the "maximum" modified part of speech is a noun, it does not satisfy the part-of-speech condition "with number modified part", so it can be seen that there is no modification relationship with "30".

Then, application of each rule is tried one after another. In this case, since it does not match the rule after this, "3
"0" has no modification destination.

Therefore, the modifier clause of the clause is -1.
Set.

Next, similar processing is performed for sequence number (1). In this case, since there is no matching modified relational rule, 0 is set in the modified phrase (708) of the phrase “30” and the phrase “age”.

Then, as described above, the syntax determination is performed using these pieces of information, and the sequence candidate selection unit 4 selects the sequence number (1) "30s" first, and selects it as the conversion candidate 1. It will be displayed on the output unit 7.

In the above process, when the modified phrase is always an unqualified item such as a classifier,
Step 1308 without making a judgment based on the modified relation rule.
You may go only. If the modified phrase is a suffix that does not always have the non-qualifying property, only step 1317 may be performed without making a determination based on the modifying relation rule. Also, for those whose qualification is determined by the part of speech as described above, 1 or 0 is set without blanking the numbe modified flag of FIG. 3 and the necessary qualification relation rule is defined. good. In this respect, the same applies to each of the examples described later.

As described above, according to the present embodiment, it is possible to avoid conversion of an independent word that is not modified by a number into an erroneous expression modified by the number, and obtain a high conversion accuracy. it can.

(Second Embodiment) A second embodiment of the present invention will be described.

A block diagram showing the configuration of the kana-kanji conversion device of this embodiment is as shown in FIG. 1 as in the first embodiment described above, and the outline of the processing is also the same as in the first embodiment. . Here, differences from the first embodiment will be mainly described.

In the present embodiment, the independent word dictionary 11 has the structure shown in FIG.
As shown in Fig. 7, the reading (15
1), the notation (152), the part of speech (153), the noun modifier flag (154), the proper noun modifier flag (155), and the word number (156) are stored.

FIG. 19 shows the structure of the generated sequence candidates for the input example “Inijyousho” shown in FIG. The structure of this sequence candidate is sequence number (1701), syntax determination (1702), clause number (1703), word number (1704), part of speech (1705), and adjective modification flag (1
706), proper noun modification flag (1707), modification grammar information (1708), and modification clause (1709).

The modification relationship rule shown in FIG. 20 is applied to such a series candidate to determine the modification relationship.
The syntax judgment as shown in 1 is made. As a result, the "proxy paper" of sequence number (1) is first selected and displayed on the output unit 7.

As described above, according to the present embodiment, it is possible to avoid the conversion of an ordinary noun having no modifier to a proper noun into an incorrect expression that modifies the proper noun, and obtain a high conversion accuracy. be able to.

(Third Embodiment) A third embodiment of the present invention will be described.

A block diagram showing the configuration of the kana-kanji conversion device of this embodiment is as shown in FIG. 1 as in the first embodiment described above, and the outline of the processing is also the same as in the first embodiment. . Here, differences from the first embodiment will be mainly described.

In the present embodiment, the independent word dictionary 11 has the structure shown in FIG.
As shown in 2, the reading (20
1), notation (202), part-of-speech (203), verb + "ni"
The modified flag (204) and the word number (205) are stored.

For the input example “especially hot” shown in FIGS. 23 (a) and 23 (b), the structure of the generated sequence candidates is shown in FIG.
4, shown in FIG. The structures of these sequence candidates are sequence number (2201), syntax determination (2202), and clause number (2
203), word number (2204), part of speech (2205),
It is composed of a verb + "ni" modified flag (2206), modified grammar information (2207), and modified clause (2208).

The modification relationship rule shown in FIG. 26 is applied to such a series candidate to determine the modification relationship.
7, the syntax judgment as shown in FIG. 28 is made. As a result, the sequence number (1) “especially hot” is selected first and displayed on the output unit 7.

As described above, according to the present embodiment, "a verb or an adjective adnominal form or an end form + a case particle" ni ""
Verbs or adjectives that are not modified by either the adnominal form or the ending form of the auxiliary verb + the case particle "ni" can be prevented from being converted into an erroneous expression that receives such a modification, and high conversion accuracy can be obtained. You can

(Fourth Embodiment) Next, a fourth embodiment of the present invention will be described.

A block diagram showing the configuration of the kana-kanji conversion device of this embodiment is as shown in FIG. 1 as in the above-mentioned first embodiment, and the outline of the processing is also the same as in the first embodiment. . That is, the kana-kanji conversion device according to the present embodiment has an input unit 1, a word search unit 2, an independent word dictionary 11, an adjunct word dictionary 12,
Phrase candidate generation unit 3, connection table 13, sequence candidate selection unit 4, syntax determination unit 5, qualification relation determination unit 6, qualification relation rule 1
4, the output unit 7 is provided.

As shown in FIG. 1, the input unit 1 is connected to the word search unit 2, the word search unit 2 is connected to the phrase candidate generation unit 3, and the phrase candidate generation unit 3 is connected to the sequence candidate selection unit 4. The sequence candidate selection unit 4 is connected to the syntax determination unit 5 and the output unit 7. Further, the modification relationship determination unit 6
Is connected to the syntax determination unit 5. The word search unit 2 includes an independent word dictionary 11 and an auxiliary word dictionary 12, the phrase candidate generation unit 3 includes a connection table 13, and the qualification relation determination unit 6 includes a qualification relation rule 14.

Next, with reference to FIG. 2, the overall flow of processing in the kana-kanji conversion device of the present embodiment will be briefly described. The main part of the kana-kanji conversion process in this embodiment is roughly divided into two processes (steps S21 and S22) shown in FIG.

In step S20, a kana character string is input from the input unit 1 such as a keyboard.

At step S21 (first conversion process),
With respect to the inputted kana character string, a process of recognizing the range of the bunsetsu is performed by using information about the connectivity between the independent word and the attached word, information about the connectivity between the attached word and the attached word, and the like. This first
The conversion process of is performed by the word search unit 2 and the phrase candidate generation unit 3 illustrated in FIG. The first processing is, for example, Japanese Patent Application Laid-Open No. 60-72014, Japanese Patent Application Laid-Open No. 3-2.
The processing described in detail in 65061 may be used.

At step S22 (second conversion processing),
A process of determining the first candidate (or the first candidate to a desired candidate) is performed for each candidate that has been identified as a phrase by the first conversion process. The second conversion process is performed by the sequence candidate selection unit 4, the syntax determination unit 5, and the modification relation determination unit 6 shown in FIG.

In step S23, the first candidate and the like determined by the second conversion process are output to the output unit 7 such as a CRT by the sequence candidate selection unit 4 shown in FIG.

After such processing is performed, the user
A predetermined operation such as confirming the first candidate output to the output unit 7 as the conversion result can be performed.

Next, the first conversion process and the second conversion process which form the main part of the kana-kanji conversion process in this embodiment will be described.

First, the first process will be described.

The word search unit 2 refers to the independent word dictionary 11 and the auxiliary word dictionary 12 to extract word candidates.

As shown in FIG. 29, the independent word dictionary 11 includes reading (301), kanji kana notation (302), part-of-speech (303), adnominal modifier (304), and continuation for each word of the independent word. Modified by body shape (305), modified by case particle "no" (306), essential flag (3)
07) and the word number (308) are stored.

Regarding the modifiability by the adnominal (304), the modifiability by the adnomial form (305), and the modifiability by the case particle "no" (306), 0 means that it has no property.
0-1 means that there is that property. here,
The smaller the number, the lower the likelihood of occurrence when the word has the modified property. For example,
The word number 032 “overseas” means that it is less likely to be modified by adnominal, adnominal, and case particle “no”. Regarding the indispensable flag (307), 1 means that modification by any one of the three modifiable properties 304 to 306 is indispensable. The number 0 to 1 means the strength of indispensability, and the smaller the number, the weaker the indispensability.
That is, the smaller the number, the smaller the degree of having to undergo any modification among the modifiability given a value other than 0. For example, for the word number 020 "Matter", the modifier by the adnominal,
This means that it is indispensable to be modified by the adnominal form or by the case particle "no". It means that the word number 037 "breath" should be almost modified by the adnominal or the case particle "no".

As shown in FIG. 4, the adjunct word dictionary 12 stores the adjunct word reading 41 and the grammatical information 42 of the word.

The word candidates extracted by the word search unit 2 are sent to the phrase candidate generation unit 3. The phrase candidate generating unit 3 refers to the connection table 13 to generate a phrase candidate from the given word candidate, and sends the result to the sequence candidate selecting unit 4.
As shown in FIG. 5, the connection table 13 stores independent word and adjunct word, and adjunct word and adjunct word connection information. In FIG. 5, adjuncts 001, adjuncts 002, ... Correspond to adjuncts such as “ga” and “o”. Figure 5
The value 1 indicates that the connection is possible, and the value 0 indicates that the connection is not possible. The term “endability” indicates whether or not one phrase can be terminated by the independent word or its adjunct word. 1 indicates that it can be terminated, and 0 indicates that it cannot be terminated. There is.

Here, examples of inputting each kana shown in (a) to (e) of FIG. 30 are "that day", "daishoo", "sankainoikoha", "kekokuchitaino", and "soon after". ", The independent word dictionary 11 shown in FIG. 3, the auxiliary word dictionary 12 shown in FIG. 4, and the connection table 13 shown in FIG. 31. FIG. 31 shows an example of the structure of a sequence candidate generated by performing
(A), FIG. 31 (b), FIG. 32 (a), FIG. 32 (b),
FIG. 32 (c) shows each.

The structure of these sequence candidates is the sequence number (2
701), syntax judgment (2702), clause number (270
3), word number (2704), part-of-speech (2705), linked object modification information (2706), modification grammar information (2707),
It is composed of a modifier clause (2708). Here, the related object modification information (2706) is 304 to 307 in FIG.
It corresponds to.

Each item will be described with reference to FIG. 31A as an example.

The sequence number is a sequence candidate number. The sequence candidate with the sequence number 0 is “the abnormal” (second candidate in FIG. 30A), and the sequence candidate with the sequence number 1 is “the abnormal”.
This corresponds to (first candidate in FIG. 30A).

The syntax judgment is information indicating the syntactic validity of the series candidate, and 0 is set as an initial value.
If this value is -1, it means that there is a clause that cannot be modified. In the case of 0, it means other possibility, that is, the clauses forming the sequence candidate have the modification destination in the sequence candidate (or may be inside or outside the sequence candidate).

The phrase number is the number of the phrase candidate. The phrase "that" is phrase number 0, the phrase "or more" is phrase number 1,
The phrase “abnormal” corresponds to the phrase number 2.

The part of speech is a part of speech of an independent word here,
For example, nouns and verbs.

The adnominal modified flag is information relating to the information on the modifiability by the adnominal. A value of 0 means that there is no modification, and a value of 1 means that there is modification. In the present embodiment, values such as adnominals, verbs, and adverbs whose values are always fixed to 1 or 0 are not set, and the modification relation rules described later are not defined.

The modified grammatical information is the morphological or syntactical grammatical information of the last word forming a phrase, for example,
Nouns, verb adnominal forms, adjective conjunction forms, case particles "wo"
For example, "no". For example, the clause number 0 in FIG.
In the case of "the", the modified grammar information is an adnominal. In addition, the phrase grammar information of "or more" of the clause number 1 is a noun.

The modifier clause represents the clause number of the clause candidate modified by the word. In addition, -1 is set when the clause to modify cannot exist. However, 0 is set when there is a possibility that the modification destination exists regardless of whether the modification candidate is the phrase number of the candidate phrase, whether it is inside or outside the sequence candidate.

The syntax judgment and the value of the modifier clause are judged and entered by the processing by the syntax judgment unit 5 and the modification relation judgment unit 6 in the second conversion processing described next.

Next, the second conversion process will be described in detail.

The sequence candidate selection unit 4 sends the sequence candidates sent from the phrase candidate generation unit 3 to the syntax judgment unit 5. Figure 9
Is a flowchart showing the flow of this processing. In step S901, the number of sequence candidates is set to N. In step S902, j representing a series candidate is set to 0. In step S903, the sequence j is sent to the syntax determination unit 5 first. This process is performed for all the series candidates (step S90).
4, S905).

As will be described later, the syntax judgment unit 5 to which the sequence candidate is given is a syntax judgment process (process for judging whether there is a clause that cannot have a modification destination, based on a modification relation rule defined in advance. )I do. The value (0 or -1) of the syntax determination is returned to the sequence candidate selection unit 4 for each sequence candidate.

The sequence candidate selection unit 4 selects the first candidate (or the first candidate to the desired candidate) from the sequence candidates based on the result of the syntax determination processing by the syntax determination unit 5 and outputs it to the output unit 7. send. FIG. 10 is a flowchart showing the flow of this processing. In step S1001, the number of sequence candidates is set to N
Set to. In step S1002, j representing a series candidate is set to 0. In step S1003, it is determined whether the syntax determination item of the sequence candidate is 1, and in step S1004 it is 0. If it is 1, step S1005
Is stored in the memory 1, and if 0, step S1006.
Save in memory 2 with. Otherwise (ie -1
In the case of) is stored in the memory 3 in step S1007.
This process is performed for all the streams (step S100).
8, S1009), and then in step S1010, the data is sent to the output unit 7 in the order of the memory 1, the memory 2, and the memory 3.

Here, the above process will be specifically described by taking the input shown in FIG. 30 (a) as an example.

As described above, in the first conversion process,
For example, the processing result by the bunsetsu candidate generation unit 3 obtained for the input “that item” in FIG.
It is (a). Sequence candidate selection unit 4 given this result
Sends the sequence candidates to the syntax determination unit 5 in the order of “more than that” of sequence number 0 and “abnormal” of sequence number 1.

The syntax judging section 5 firstly determines the modification relation judging section 6
Using, the qualifying clause of each bunsetsu is obtained by analyzing the qualifying relation of the bunsetsu for each series candidate. For example, FIG.
In the case of the input example of (e), FIG. 33 (a), FIG. 33 (b),
As shown in FIGS. 34 (a), 34 (b), and 34 (c), the value of the modifier clause (clause number or 0 or -1)
Is required. For example, in “more than” of sequence number 0 in FIG. 33 (a), “that” of phrase number 0 cannot modify “more than” of phrase number 1, so the phrase of phrase number 0 is modified. The clause is -1.

Next, the syntax judgment section 5 carries out the syntax judgment of each series candidate based on these results, and obtains the value of the syntax judgment. For example, in the case of the input examples of FIGS. 30 (a) to 30 (e), FIGS. 33 (a), 33 (b), 34 (a), 34
As shown in (b) and FIG. 34 (c), the syntax determination value "-1" is returned for sequence number 0, and "0" is returned for sequence number 1.

The sequence candidate selection unit 4 determines the first candidate or the like based on the value of the syntax determination of each clause candidate. In this case, to the output unit 7 in the order of the sequence candidate “the abnormal” having the sequence number 1 having the syntax determination value of 0 and the sequence candidate “more than that” having the sequence number 0 having the syntax determination value of −1. send.

Thus, as shown in FIG. 30A, the sequence candidate “30s” with the sequence number 1 is the conversion candidate 1, and the sequence candidate “30 maximum” with the sequence number 0 is the conversion candidate. It is output as 2. Similarly, FIG. 30 (b)
As shown in (e), "compensation for selection", "after participation", "after all delay""soon to be of Hara" as conversion candidate 1, "select size", "after participation", " Eventually, "delayed" and "soon to be original" are output as conversion candidates 2.

The processing of the syntax determining section 5 (and the modification relation determining section 6) will be described in more detail below.

The syntax judgment unit 5 judges the syntactic validity of each series candidate. FIG. 13 is a flowchart showing the flow of the processing. In step S1101, the total number of clauses of the sequence candidate is set to B, and in step S1102, b representing a clause is set to 0. In step S1103, it is checked whether or not a modifier clause exists in the clause b (whether the term of the modifier clause of the sequence candidate is -1). In the case of −1, in step S1106, the item (170
Set -1 to 2). This process is performed for all the clauses except the rightmost clause in the sequence (step S110).
4, S1105), and when completed, returns to the sequence candidate selection unit 4.

Here, the input of FIG. 30A will be specifically described as an example. In the case of this example, the value entered in the item (2708) of the modifier clause of FIG. 33A is in the state by the processing by the modifier relation determining unit 6 described below. Therefore, in the case of the sequence number 0, the item of the modifier clause is -1, so -1 is entered in the item (2702) of syntax judgment. In the case of the sequence number 1, nothing is written in the syntax judgment, so that it remains 0.

Next, the processing of the modification relation judging section 6 will be described. Here, a process of determining the presence or absence of a qualifying relation between phrases is performed. FIG. 14 is a flowchart showing the flow of the processing. In step S1201, the total number of clauses in the series candidate is set to B, and in step S1202, b representing the clause in the series candidate is set to 0. In step S1203, the processing for determining the modification destination of the phrase b is performed. This determination process is performed for all phrase candidates other than the rightmost phrase (steps S1204 and S1205). For example, the sequence candidate of sequence number 0 obtained from the input shown in FIG.
Is composed of the phrase number 0 "that" and the phrase number 1 "or more", so the modification destination of the phrase number 0 "that" is obtained.

The process of checking the modification relation for each clause in step S1203 will be described below. FIG. 15 is a flowchart showing the flow of the processing.

In step S1300, K is initialized to -1. In step S1301, the total number of modification relation determination rules defining the relationship between the modified clause and the modified clause is set in R, and in step S1302, r representing the modification relation rule is set. Set to 0.

FIG. 35 is an example of the modification relation rule, which is the part of speech of the last word of the phrase to be modified (1401), the condition to be satisfied by the modified part of speech (1402), the part of speech of the modified phrase (1403), and the target part. Conditions that the modified part of speech must meet (140
4), the applicable range of the rule (1405).

Rule (1) states that "the case particle" ga "modifies an adjective (until the adjective appears)" and rule (2) states that "adnominal modifies a noun (use (Independent words before words appear) ", rule (3) is
Rule (4) says that the adnomial ending modifies a noun that has a modifiability due to the adnominal form (up to the independent word before the adverb appears), "the case particle" no " Rule (5) says that "modify a noun that has a qualifying property by" (up to an independent word before the idiom appears) does not qualify a noun whose qualified essential flag is not 1. Rule (6) means that "adverbs modify verbs (until a verb appears)".

In step S1303, whether the phrase b is the modified part of speech of the rule r, and in step S1304 whether the phrase b satisfies the part of speech of the rule r, the modification grammar of the sequence candidate information as shown in FIGS. Check by referencing information (2707). If the condition is satisfied, the process proceeds to step S1305. If the conditions are not met, the next qualifying relation rule is applied. This process is repeated until all rules are applied (steps S1313 and S13).
14).

In step S1305, b + 1 is set to j representing the phrase to be modified. In step S1306, whether the phrase j is the modified part of speech of the rule r, and step S13
At 07, it is checked whether the phrase j satisfies the part-of-speech condition of the rule r.

If the two conditions are satisfied, step S
At 1308, the clause j is set in the modified clause (708) of the clause b.

If either condition is not satisfied,
The next clause in the sequence candidate is checked as a modification destination (step S1309). In this case, if there is no next clause, 1 is set to k indicating that there is a clause to be modified in the subsequent input. If there is a next clause, it is checked in step S1311 if the applicable range of the rule r is satisfied. If so, the processing from step S1306 is repeated. If not satisfied, the next rule is applied (steps S1313 and S1314).

If none of the rules match, the flow advances to step S1315 to check whether k is 1. If k is 1, 0 is set in the qualified clause (2708) of clause b (step S1316). If k is -1, -1 is set in the qualified clause of clause b (step S1317).

Here, regarding the above processing, FIG.
This will be specifically described by taking as an example the case where the modification relation is obtained for the input of.

First, the case of checking the modification relation of the phrase number 0 "that" will be described.

First, the above modification relation rule (1) will be investigated. Check whether the phrase “that” is the modified part-of-speech “case particle” in rule (1). Since "that" is a continuum,
This rule is not applied and the next rule is tried.

In Rule (2), "that" satisfies the modifier part-of-speech "adjoint part", so it is next checked whether the phrase "above" on the right side satisfies the modified part-of-speech "noun". Although the part-of-speech of "or above" is "noun", it does not satisfy the part-of-speech condition "adjective with adjunct modifier" (270 in FIG.
6), it can be seen that there is no modification relationship with "the".

Then, application of each rule is tried one after another. In this case, since it does not match the rule after this, "that" has no modification target.

Therefore, the modifier clause of the clause is -1.
Set.

Next, similar processing is performed for sequence number (1). In this case, since the modification relation rule (2) is matched, 2 is set in the modification clause (2708) of the clause “that”.

Then, as described above, the syntax determination is performed using these pieces of information, and the sequence candidate selection unit 4 first selects "the abnormal" of the sequence number (1),
The conversion candidate 1 will be displayed on the output unit 7.

Similarly, in the case of FIG. 30B, the sequence candidate (0) does not have a modification destination in “select”, but the sequence candidate (1) has a modification destination. The candidate (1) is selected first and displayed on the output unit 7.

In the case of FIG. 30C, the sequence candidate (0)
Is a candidate for series (1), although "participating" has no modification destination
Has a modification destination, so that the series candidate (1) is selected first and displayed on the output unit 7.

In the case of FIG. 30D, the sequence candidate (0)
Indicates that the "zone" in which the modification is essential is not modified, but the "delay" of the series candidate (1) is not indispensable, so that the series candidate (1) is selected first and the output unit 7 is selected. Will be displayed.

In the case of FIG. 30E, the sequence candidate (0)
, The original, which is essential to be modified, is not modified, but the “belly” of the candidate series (1) is not essential to be modified. Therefore, the candidate series (1) is selected first and the output unit 7 is selected. Will be displayed.

As described above, according to the present embodiment, it is possible to avoid conversion of an independent word that is not modified by a continuum into an erroneous expression modified by a continuum, and obtain high conversion accuracy. it can.

(Fifth Embodiment) A fifth embodiment of the present invention will be described.

A block diagram showing the configuration of the kana-kanji conversion device of this embodiment is as shown in FIG. 1 as in the first embodiment described above, and the outline of the processing is also the same as in the first embodiment. . Here, differences from the first embodiment will be mainly described.

In the present embodiment, the independent word dictionary 11 has the structure shown in FIG.
As shown in Fig. 6, the reading (31
1), kanji kana notation (312), part-of-speech (313), modifier for noun (314), adnominal modifier with case particle "no" (315), essential flag (316), word number (317) memorized Has been done.

Regarding the modifiability to the noun (314) and the adnominal modifier with the case particle "no" (315), 1 means that the property exists, and 0 means that the property does not exist. As for the indispensable flag (316), 1 means that modification by any one of the two modifiability 314 and 315 that is 1 is indispensable. Further, when the numbers are 0 to 1, the explanation is the same as in the fourth embodiment. For example, with respect to "spirit," it means that adnominal modification with the case particle "no" is indispensable. Further, with respect to "middle grade", it means that any modification of the noun and the adnominal modification with the case particle "no" is indispensable.

For each kana input example shown in FIG. 37, as shown in FIG.
6, the independent word dictionary 11 shown in FIG. 6, the auxiliary word dictionary 12 shown in FIG.
FIG. 38 shows the structure of the sequence candidates generated using the connection table 13 shown in FIG. The structure of this sequence candidate is sequence number (2711), syntax judgment (2712), clause number (2713), word number (2714), part of speech (271).
5), linked information (2716), modification grammar information (2
717) and a modified clause (2718).
Here, the linked modification information (2716) is 31 in FIG.
4 to 316.

FIG. 40 shows an example of the modification relation rule of this embodiment. This rule is that the part of speech 1411 of the last word of the phrase to be modified and the condition 141 that the modifying part of speech must satisfy.
2, the part of speech 1413 of the phrase to be modified, the condition 1414 to be satisfied by the modified part of speech, and the application range 1415 of the rule.

Rule (1) is that "a noun that has a modifier for a noun modifies the noun (up to the next clause of the modifier part of speech)".
That is, rule (2) states that "a case particle" no "modifies a noun for a noun with a case particle" no "that has a noun-modifying property (up to the next clause of the modifier part-of-speech)". ,
Rule (3) means that the case particle “ga” (if the modifier essential flag is not 1 in the independent word, modifies the verb (until the adjective appears)).

The decoration relation rule shown in FIG. 40 is applied to the series candidates in FIG. 38 to judge the decoration relation, and FIG.
The syntax is determined as shown in.

Here, the case of obtaining the modification relationship will be described with reference to the example of FIG. First, a case will be described in which the modification relation of the phrase number 0 “spirit” is investigated.

First, the above modification relation rule (1) will be investigated. Phrase "Kei" is the modified part-of-speech "Noun" of rule (1)
Check if The modified part-of-speech of "Kagega" is the case particle "ga", so it does not match this rule. The following rule is tried to be applied, but since "Keikiga" does not satisfy the modified part-of-speech "case particle" no "" of rule (2), the next rule is tried to be applied. "Kagega" does not satisfy the part-of-speech condition "the indispensable flag of the independent word is not 1" (Fig. 38 (a)).
2716), the following rule is tried to be applied, but it does not match the rules after this, and eventually "Kagega" has no modification target. Therefore, the qualifying clause of the clause is −
Set 1. Similar processing is performed for the sequence number (1). In this case, since it matches the modification relation rule (3), 1 is set in the modification phrase (2718) of the phrase "Kagekaga".

A syntax judgment is performed using this information, and the sequence candidate selection unit 4 selects the sequence number (1) “I can see the machine image” first and displays it on the output unit 7.

Similarly, as shown in FIGS. 37 (b) and 37 (c), "occurrence in the Middle East" and "membership fee of the main meeting" are selected and displayed on the output unit 7.

As described above, according to this embodiment, it is possible to avoid the conversion of an adnominal modifier into an erroneous expression that modifies a word that is not subject to modification by the adnominal modifier, and achieves high conversion accuracy. Obtainable.

(Sixth Embodiment) A sixth embodiment of the present invention will be described.

A block diagram showing the configuration of the kana-kanji conversion device of the present embodiment is as shown in FIG. 1 like the first embodiment described above, and the outline of the processing is also the same as that of the first embodiment. . Here, differences from the first embodiment will be mainly described.

In the present embodiment, the independent word dictionary 11 is stored in FIG.
As shown in 1, the reading (32
1), kanji kana notation (322), part of speech (323), adverb noun phrase configuration information (324), and word number (325) are stored. Here, regarding the adverb noun phrase configuration information,
1 means that the property exists, and 0 means that it does not exist.

For each kana input example shown in FIG. 42, as shown in FIG.
1. The independent word dictionary 11 shown in FIG. 1 and the auxiliary word dictionary 12 shown in FIG.
FIG. 43 shows the structure of sequence candidates generated using the connection table 13 shown in FIG. The structure of this sequence candidate is sequence number (2721), syntax judgment (2722), clause number (2723), word number (2724), part of speech (272).
5), adverb noun phrase construction information (2726), modified grammar information (2727), and modified clause (2728). Here, the adverb noun phrase configuration information (2726) is as shown in FIG.
1 corresponds to 324.

FIG. 45 shows an example of the modification relation rule of this embodiment. This rule includes the part of speech 1421 of the last word of the phrase to be modified and the condition 142 that the modified part of speech must satisfy.
2. The part of speech 1423 of the phrase to be modified, the condition 1424 to be satisfied by the modified part of speech, and the application range 1425 of the rule.

Rule (1) means that "adjuncts modify a noun (until there is an adverb that has no noun phrase constituents)." The same applies to rules (2) and (3).

The modification relation rule shown in FIG. 45 is applied to the series candidates shown in FIG. 43 to determine the modification relationship.
The syntax is determined as shown in. Here, FIG.
In the example of (a), the case of obtaining the modification relationship will be described. First, the case of checking the modification relation of the phrase number 0 “that” will be described.

The above modification relation rule (1) will be investigated. Check whether the phrase "that" is the modified part-of-speech "adjective" of rule (1). Although the modified part-of-speech of "that" is "adjunct", it does not satisfy the condition "until an adverb having no noun phrase composition exists" in the application range (272 in FIG. 43 (a)).
6) Attempt to apply the next rule, but since it does not match any other rule, “that” has no modification destination after all. Therefore, -1 is set to the qualified clause of the clause. Similar processing is performed for the sequence number (1). In this case, since it matches the modification relation rule (1), 2 is set in the modification clause (2728) of the clause “that”.

A syntax decision is made using this information, and the sequence candidate selection unit 4 makes the sequence number (1) "the imitation".
Will be selected first and displayed on the output unit 7.

Similarly, as shown in FIG. 37B, "the request" is selected and displayed on the output unit 7.

As described above, according to the present embodiment, it is possible to avoid conversion of an adverb that does not form a noun phrase into an erroneous expression that forms a noun phrase, and obtain high conversion accuracy.

Here, in each of the above-described embodiments, a modification relation rule is described as having a modification relation, but on the contrary, a rule having no modification relation is described, and when the rule matches, It is also possible to determine that there is no modification destination.

It is also possible to perform a full-scale parsing. On the contrary, it is also possible to judge very simply by the pattern of the part of speech.

[0207] Further, when creating a series candidate, it is not necessary to separately have the clauses shared with other series, but it is possible to have them in a shared form.

When selecting a series candidate, naturally, other information such as frequency information and co-occurrence information can be used.

Further, in each of the embodiments, the information regarding the modifiability and the modifiability is described in the independent word dictionary, but the information is not necessarily in the dictionary and may be provided as another table.

Further, in each of the embodiments, description has been made on the case where the information regarding the modifying property and the modifying property and the information regarding the indispensability thereof are binary information of 1 and 0, but the case of 1 to 0 can be similarly used. . For example, “overseas” in FIG. 29 is used such that the priority is lowered when it is modified by an adnominal, an adnominal form, or a case particle “no”. If "overseas" in FIG. 29 is modified by the particle "no", the value of the syntax judgment is 0 instead of -1, but the priority is lower than other candidates whose syntax judgment is 0. Used for. If the indispensability flag is 0 to 1, the modifier is given a positive value, and if it is not subjected to any modification, it depends on the strength of the indispensability (the higher the number, the stronger). To lower the priority. For example, when “breath” in FIG. 29 is not modified by any of the adnominal, the adnominal form, and the case particle “no”, the value of the syntactic judgment is set to 0, but the syntactic judgment is set to 0 rather than other candidates. It is used to drop the priority greatly.

In each embodiment, the information regarding the modifying property and the modifying property is expressed as a flag.
It may be expressed as one part of speech. For example, ""
It is also essential to use a part of speech (eg, up-and-coming) that is indispensable to be used by modifying the adnominal body via.

As described above, the present invention is not limited to the above-described embodiments, but various modifications can be carried out without departing from the scope of the invention.

[0213]

According to the first aspect of the present invention, the part of speech of the first independent word which becomes the modified clause, the part of speech of the second independent word which becomes the modified clause, and the first independent word and the second independent word. Since the priority of bunsetsu candidates is determined based on the qualification relation rule that defines the condition that qualification relations are established between words, it is possible to check the qualification relations between bunsetsu candidates in detail, which was previously overlooked. It is possible to avoid conversion to an erroneous expression and obtain high conversion accuracy.

According to the second and third aspects of the invention, the modifier information indicating the presence or absence of the modifier by a number that independently forms a clause without an adjunct to the independent word, and the modifier between the number and the independent word. Since the priority of bunsetsu candidates is determined based on the qualifying relational rules that define the conditions under which the relationship is established, avoiding the conversion of an independent word that is not qualified by a number into an incorrect expression that is modified by the number. You can

According to the fourth aspect of the present invention, the modifier information indicating whether or not each common noun is modifiable with respect to a proper noun when it constitutes a phrase independently without accompanying words, and a modifier between the common and proper nouns. Since the priority of bunsetsu candidates is determined based on the modification relation information that defines the conditions for establishing the relationship, common nouns that have no modifier to the proper noun modify the proper noun to avoid incorrect conversion. can do.

According to the fifth aspect of the present invention, a verb or an adjective's adjunct form or final form +
Modifier information indicating the presence or absence of the modifier by the case particle "ni" or the adnominal form or the ending form of the auxiliary verb + the case particle "ni", and the adjunct form or the ending form of the verb or adjective + the case particle ”Or“ the adnominal or final form of the auxiliary verb + the case particle “ni” ”and the qualifying relation rule that defines the condition that the qualifying relationship is established between the verb or adjective, the priority of the clause candidates is determined. A verb or adjective that is not modified by either the adnominal adjunct or the adjective adjunct + the case particle "ni" or by the adverb adjunct or the adjective "ni" is wrong It is possible to avoid conversion to expressions.

According to the sixth aspect of the present invention, the qualification information indicating whether or not the noun has the qualification of the noun to the noun, and the condition for establishing the qualifying relation between the noun and the noun and the case particle “no”. Since the priority order of the bunsetsu candidates is determined based on the modified relational rules described above, it is possible to avoid conversion of an independent word having no modifiability by a noun to an erroneous expression modified by the noun.

According to the seventh aspect of the invention, the modifier information indicating the adnominal modifier of each noun and the noun or case particle "no"
Since the priority of bunsetsu candidates is determined based on the qualifying relation rule that defines the condition that the qualifying relation is established between the noun and the noun,
It is possible to avoid conversion of an adnominal modifier into an erroneous expression that modifies a word that is not qualified by the adnominal modifier.

According to the eighth invention, the phrase candidate is based on the noun phrase constituent element information relating to whether or not each adverb can be an element of a noun phrase and the modifier relation rule defining the condition that the adverb is an element of the noun phrase. Since the priority order of is determined, it is possible to avoid conversion of an adverb that does not form a noun phrase into an erroneous expression that forms a noun phrase.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a kana-kanji conversion device according to first to sixth embodiments of the present invention.

FIG. 2 is a flowchart showing an outline of processing of the kana-kanji conversion device shown in FIG.

FIG. 3 is a diagram showing an example of information written in an independent word dictionary according to the first embodiment.

FIG. 4 is a diagram showing an example of information written in an auxiliary word dictionary according to first to sixth embodiments.

FIG. 5 is a diagram showing an example of information described in connection tables according to the first to sixth embodiments.

FIG. 6 is a diagram showing an example of conversion candidates for an input example in the first embodiment.

FIG. 7 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generating unit in the embodiment.

FIG. 8 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generating unit in the embodiment.

FIG. 9 is a flowchart showing the flow of processing in the sequence candidate selection unit according to the first to sixth embodiments.

FIG. 10 is a flowchart showing the flow of processing in the sequence candidate selection unit according to the first to sixth embodiments.

FIG. 11 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determination unit in the first embodiment.

FIG. 12 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determining unit in the embodiment.

FIG. 13 is a flowchart showing a processing flow of a syntax determination unit according to first to sixth embodiments.

FIG. 14 is a flowchart showing a processing flow in a modification relation determination processing unit according to first to sixth embodiments.

FIG. 15 is a flowchart showing a flow of a process of applying a modification relation rule according to the first to sixth embodiments.

FIG. 16 is a diagram showing an example of a modification relation rule in the first embodiment.

FIG. 17 is a diagram showing an example of information written in the independent word dictionary in the second embodiment.

FIG. 18 is a diagram showing an example of conversion candidates for an input example in the same embodiment.

FIG. 19 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generating unit in the embodiment.

FIG. 20 is a diagram showing an example of a modification relation rule in the embodiment.

FIG. 21 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determination unit in the example.

FIG. 22 is a diagram showing an example of information described in the independent word dictionary in the third embodiment.

FIG. 23 is a diagram showing an example of conversion candidates for an input example in the same embodiment.

FIG. 24 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generating unit in the embodiment.

FIG. 25 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generating unit in the embodiment.

FIG. 26 is a diagram showing an example of a modification relation rule in the embodiment.

FIG. 27 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax judgment unit in the embodiment.

FIG. 28 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax judgment unit in the embodiment.

FIG. 29 is a diagram showing an example of information written in the independent word dictionary in the fourth embodiment.

FIG. 30 is a diagram showing an example of conversion candidates for an input example in the same embodiment.

FIG. 31 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generation unit in the embodiment.

FIG. 32 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generation unit in the embodiment.

FIG. 33 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determination unit in the example.

FIG. 34 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determination unit in the example.

FIG. 35 is a diagram showing an example of a modification relation rule in the embodiment.

FIG. 36 is a diagram showing an example of information described in the independent word dictionary in the fifth embodiment.

FIG. 37 is a diagram showing an example of conversion candidates for an input example in the same embodiment.

FIG. 38 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generation unit in the embodiment.

FIG. 39 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax determination unit in the example.

FIG. 40 is a diagram showing an example of a modification relation rule in the embodiment.

FIG. 41 is a diagram showing an example of information written in the independent word dictionary in the sixth embodiment.

FIG. 42 is a view showing an example of conversion candidates for an input example in the same example.

FIG. 43 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the phrase candidate generation unit in the embodiment.

FIG. 44 is a diagram showing an example of a structure of a sequence candidate which is a processing result in the syntax judgment unit in the embodiment.

FIG. 45 is a diagram showing an example of a modification relation rule in the embodiment.

FIG. 46 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

FIG. 47 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

FIG. 48 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

FIG. 49 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

FIG. 50 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

FIG. 51 is a diagram for explaining erroneous conversion in conventional Kana-Kanji conversion.

[Explanation of symbols]

1 ... Input part, 2 ... Word search part, 3 ... Phrase candidate generation part, 4
... clause candidate selection unit, 5 ... syntax determination unit, 6 ... modifier relation determination unit, 7 ... output unit, 11 ... independent word dictionary, 12 ... adjunct word dictionary, 13 ... connection table, 14 ... modifier relation rule

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Tatsuya Uehara 1 Komukai Toshiba-cho, Sachi-ku, Kawasaki-shi, Kanagawa Toshiba Research & Development Center (72) Inventor Yumi Mizutani Komukai-Toshiba, Kawasaki-shi, Kanagawa Town No. 1 Toshiba Corporation Research & Development Center

Claims (8)

[Claims] 1. A word corresponding to an input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words, and an auxiliary word dictionary that stores readings, notations and grammatical information for adjunct words, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence / absence of a qualifying relation is determined by the part of speech of the first independent word that becomes the qualifying clause, the part of speech of the second independent word that becomes the qualifying clause, and the qualifying relation between the first independent word and the second independent word. A kana-kanji conversion method characterized by making a determination based on a qualifying relation rule that defines a condition to be established, and determining the priority of the phrase candidates based on the presence or absence of the determined qualifying relation. 2. A word corresponding to an input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words, and an adjunct dictionary that stores readings, notations and grammatical information for adjuncts, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence / absence of a qualifying relation is defined by the qualification information indicating the qualification by a number that independently forms a clause without an adjunct to the independent word and the condition for the qualifying relation to be established between the number and the independent word. A kana-kanji conversion method characterized by making a determination based on a qualifying relation rule and deciding the priority of the phrase candidates based on the presence or absence of the determined qualifying relation. 3. An independent word dictionary in which reading, notation and part-of-speech information are stored for an independent word, an adjunct dictionary in which reading, notation and grammatical information for an adjunct word are stored, and an independent word without an adjunct word And a modified relation rule storage means for storing modified relation rules that define the conditions under which a modified relation is established between a number and an independent word. A word search means for searching the word corresponding to the inputted kana from the independent word dictionary and the auxiliary word dictionary, between the independent word and the auxiliary word searched by the word search means, and the auxiliary word and the auxiliary word Clause candidate generating means for determining connectivity between the clauses and generating clause candidates, and a clause symbol generated by the clause candidate generating means based on information in the modified storage means and the modification relation rule storing means. And a priority order determination unit that determines the priority order of the phrase candidates based on the presence or absence of the modification relationship determined by the determination unit. Kana to Kanji conversion method. 4. A word corresponding to the input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words, and an auxiliary word dictionary that stores readings, notations and grammatical information for adjunct words, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence or absence of a qualifying relation, the qualification information indicating whether or not each common noun has a qualifying property with respect to a proper noun when it forms a bunsetsu independently without accompanying words, and the condition that the qualifying relation is established between the common noun and the proper noun. The kana-kanji conversion method is characterized in that the priority order of the phrase candidates is determined based on the presence or absence of the determined modification relationship. 5. A word corresponding to the input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words, and an auxiliary word dictionary that stores readings, notations and grammatical information for adjuncts, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. Whether or not there is a qualifying relationship is determined by the verb or adjective's "adjective or verb or adjective form + ending particle" ni "" or "adjective adjunct or ending + case particle" ni "" Qualification information and the adnominal adjunct or verb or adjective + case particle "ni" or the adjunct or adjunct verb + case particle "ni" and a verb or adjective Kana conversion method determined based on standing to qualify the relationship rules define the conditions, based on the presence or absence of the determined modified relationship, and determines the priority of the phrase candidates. 6. A word corresponding to the input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words and an auxiliary word dictionary that stores readings, notations and grammatical information for adjunct words, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence or absence of a qualifying relation is defined as a qualifying relation rule that specifies the qualifying information that indicates whether or not there is a qualifying relation to the noun, and the condition that a qualifying relation is established between the noun and the noun and the case particle "no". A kana-kanji conversion method, characterized in that the priority order of the phrase candidates is determined based on the presence or absence of the determined modification relationship. 7. A word corresponding to the input kana is searched from an independent word dictionary that stores readings and notations and part-of-speech information for independent words and an auxiliary word dictionary that stores readings, notations and grammatical information for adjunct words, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence or absence of a qualifying relationship is determined based on the qualifying information that indicates the adnominal qualifying property of each noun, and the qualifying relation rule that defines the condition for the qualifying relation to be established between the noun or the case particle "no" and the noun. A kana-kanji conversion method, characterized in that the priority of the phrase candidates is determined based on the presence or absence of the determined modification relation. 8. A word corresponding to the inputted kana is searched from an independent word dictionary in which reading, notation and part-of-speech information are stored for independent words and an auxiliary word dictionary in which reading, notation and grammatical information are stored for adjunct words, Based on the part-of-speech information and the grammatical information, the connectivity between the retrieved independent word and the adjunct word and the adjunct word and the adjunct word is determined to generate a bunsetsu candidate, and between the generated bunsetsu candidates. The presence or absence of a qualifying relationship is judged based on the noun phrase constituent element information regarding whether or not each adverb can be an element of a noun phrase, and the qualifying relation rule that prescribes the conditions under which an adverb becomes an element of a noun phrase. A kana-kanji conversion method, characterized in that the priority of the phrase candidates is determined based on the presence or absence of a relationship.