JP2017068879A - Style conversion device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To highly accurately convert polite forms into casual forms.SOLUTION: An input unit 10 accepts an input sentence. A form replacement unit 26 converts polite forms included in the input sentence into casual forms on the basis of predetermined rules for converting a processing object morpheme being a polite form into a casual form according to at least one of parts of speech or conjugated forms of a morpheme appearing one or two before the processing object morpheme being a polite form, notation of a subsequent morpheme or a subsequent morpheme string subsequently appearing on the processing object morpheme, and the absence of a subsequent morpheme, or the fact whether the subsequent morpheme is a sentence end symbol or not.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a style conversion device, method, and program.

  Non-Patent Document 1 describes a method for unifying the style of a sentence into “a key” (normal) by means of a statistical method using a language model created from newspaper articles for the transcription of lecture speech. Yes. However, it has been reported that the conversion to “a key” (normal) has low accuracy (that is, there are many cases where the expression to be converted cannot be converted correctly). Furthermore, it is stated that many of the parts that could not be converted were expressions that needed to change the verb usage.

Kazuya Shimooka, Tatsuya Kawahara, Hiroshi Okuno, “Shaping the Style Using Statistical Methods for Transcription of Lectures”, Information Processing Society of Japan Report on Natural Language Processing, 2002 (44), p.81-88, 2002.

  The conventional technology does not take into account the use of verbs that change depending on the morpheme before and after, so it generates sentences that are ineligible as Japanese by replacing morphemes (columns), and some that include "mas" There was a problem that the expression of was not converted.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a stylistic conversion device, method, and program capable of accurately converting a honorific expression into a regular expression.

  In order to achieve the above object, the style conversion device of the present invention includes an input unit that receives an input sentence, and a part of speech or a utilization type of a morpheme that appears one or two times before a processing target morpheme that is a sympathetic expression. And the following morpheme or subsequent morpheme string appearing after the processing target morpheme, the absence of the subsequent morpheme, or whether the subsequent morpheme is a sentence ending symbol. An expression replacement means for converting a honorific expression included in the input sentence to a normal expression based on a predetermined rule for converting a processing target morpheme that is a body expression into a normal expression. Has been.

  The style conversion method of the present invention is a style conversion method in a style conversion apparatus including an input means and an expression replacement means, wherein the input means accepts an input sentence, and the expression replacement means is a honorific expression. Part of speech or inflection type of a morpheme that appears one or two times before a processing target morpheme and a notation of a subsequent morpheme or a subsequent morpheme string that appears following the processing target morpheme, or the absence of the subsequent morpheme, or In accordance with at least one of whether or not the subsequent morpheme is a sentence ending symbol, the input sentence is converted to the input sentence based on a predetermined rule for converting the processing target morpheme that is the honorific expression into a normal expression. And converting the contained honorific expression into a regular expression.

  In the expression replacement means of the present invention, the part of speech of a morpheme that appears one or two times before the processing target morpheme is a verb class, and appears before the processing target morpheme and the processing target morpheme Included in the input sentence based on the rules for converting the processing target morpheme that is a sympathetic expression into a regular expression depending on whether or not “is” is omitted between It is possible to convert the expression of honored body into a regular expression.

  The expression replacement means according to the present invention is configured so that the part of speech of a morpheme that appears one or two times before the processing target morpheme corresponds to a noun class, an adjective class, or a verb class. Based on the rule for converting a processing target morpheme that is a body expression into a normal expression, the honorific expression included in the input sentence can be converted into a normal expression.

  The expression replacement means of the present invention converts a processing target morpheme that is “is”, “deshi”, “decho”, “masu”, “masashi”, “maso”, or “no” into a normal expression. On the basis of the rule for the above, it is possible to convert the honorific expression included in the input sentence into a regular expression.

  The program of this invention is a program for functioning a computer as each means of the style conversion apparatus of this invention.

  As described above, according to the stylistic conversion device, method, and program of the present invention, the part-of-speech or inflection type of the morpheme that appears one or two times before the processing target morpheme that is the honorific expression, and the processing A processing target morpheme that is a sympathetic representation, depending on at least one of the following morpheme or subsequent morpheme sequence notation that follows the target morpheme, the absence of the subsequent morpheme, or whether the subsequent morpheme is a sentence ending symbol Based on a predetermined rule for converting to normal expression, by converting the honorific expression contained in the input sentence to the normal expression, it is possible to accurately convert the honorific expression to the normal expression. The effect of being able to be obtained is obtained.

It is a block diagram which shows the functional structural example of the style conversion apparatus which concerns on this Embodiment. It is an image figure which shows an example of a part of verb utilization table | surface. It is a flowchart which shows the expression substitution process routine in this Embodiment. It is a flowchart which shows the expression substitution process routine in this Embodiment. It is a flowchart which shows the expression substitution process routine in this Embodiment.

<Overview>
First, an outline of an embodiment of the present invention will be described.

  The problem to be solved by the embodiment of the present invention is to rewrite a sympathetic expression contained in a sentence written in Japanese into a regular expression. In the embodiment of the present invention, even if the expression is an honorific expression that occurs in an environment that cannot be dealt with by simple substitution of a character string, the expression is converted to a normal expression while suppressing generation of a sentence that is ineligible as Japanese. Realize that.

  In the embodiment of the present invention, when replacing the honorific expression with the regular expression, (A) the type of part of speech of the morpheme that appears one or two before the processing target morpheme, and (B) the processing target morpheme By using conversion rules based on four conditions: notation, (C) notation of subsequent morpheme or succeeding morpheme string, and (D) presence / absence of tails, it is necessary to change the utilization form of verbs that could not be handled by conventional research Even if there is a certain expression, the expression “is / mass” (“is”, “mas”, and their use forms) is changed to “da” (or an empty string) while suppressing the conversion to an expression that is ineligible as Japanese. ), And can be converted to the proper usage of verbs.

For example, it is possible to convert sentences written with respectful expressions such as “Mt. Fuji is beautiful” and “I will climb Mt. Fuji” into sentences with a regular expression such as “Mt. Fuji is beautiful” and “Climb Mt. Fuji”. is there.
In the following description, “sentence” (text) refers to a sentence written in Japanese. Here, the sentence written in Japanese may be anything written in Japanese, such as writing to a blog / SNS, speech recognition result of speech, text chat, and the like.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<System configuration of style conversion device>
In the present embodiment, a sentence written in Japanese is received as an input sentence, and the expression “is / mas” included in the input sentence is appropriately used as “da” (or empty character) and verbs. Replace with form (utilization ending). Stylistic conversion device 100 according to the present embodiment is configured by a computer including a CPU, a RAM, and a ROM that stores a program for executing a style conversion processing routine described later. As shown in FIG. 1, this computer can be functionally represented by a configuration including an input unit 10, a calculation unit 20, and an output unit 30. Hereinafter, each part is explained in full detail.

  The input unit 10 receives a sentence to be converted to a sentence style as an input sentence.

  The computing unit 20 converts the honorific expression included in the input sentence received by the input unit 10 into a regular expression. The calculation unit 20 includes a morpheme analysis unit 22, a verb utilization table database 24, and an expression replacement unit 26.

  The morpheme analysis unit 22 performs morpheme analysis on the input sentence received by the input unit 10. Specifically, the morpheme analysis unit 22 performs morpheme analysis on the input sentence, and acquires a morpheme boundary, a part of speech, and a utilization type of verbs.

For example, when “I climb Mt. Fuji” is accepted as an input sentence, the morphological analysis unit 22 performs morphological analysis on the input sentence, and the morpheme boundary “Mt. Fuji / ni / climb / mass” indicates the part of speech and verb class of each morpheme. The utilization type “noun / particle / verb: La line five steps / auxiliary verb” is acquired as a morphological analysis result.
Here, any morpheme analyzer used in the morpheme analyzer 22 may be used as long as the morpheme boundary, the part of speech, and the verb type can be acquired.

  The verb utilization table database 24 stores a verb utilization table. FIG. 2 shows an example of a part of the verb utilization table. As an example of a part of the verb utilization table, FIG. 2 shows a one-stage utilization ending list that is a utilization type of the verb “see”, and a utilization ending list that is a utilization form of the verb “writing”. , And a list of utilization endings of the five-stage use of the La row, which is a utilization type of the verb “climb”. As shown in FIG. 2, the verb utilization table includes a utilization ending list for each utilization type of verbs. Here, the verb class represents a verb, an auxiliary verb utilizing a verb type, or an auxiliary verb, as will be described later. In addition, the usage endings corresponding to the usage forms of verbs (former form (A-dan), form (O-dan), continuous form, continuous use form (sound form), end form, and command form)) are used in the inflection list. Stored.

Here, the green shape (stage A) refers to a green shape whose vowel is a. This is a form connected to a morpheme such as “no”, and one example is “do not climb”. The green form (o stage) refers to a green form whose vowel is o. This is a form connected to a morpheme such as “U”, and “Let ’s climb” is an example.
In addition, the continuous form (sound form) refers to a continuous form that has been turned into a sound stool, a sound stool or a sound-repellent stool. This is a form of connection to “te (de)” and “ta (da)”, and one example is “climb”.

  For example, as shown in FIG. 2 above, when the verb usage type is “La line 5 stage usage”, the corresponding usage ending list includes “ra”, “ro”, “ri”, “tsu”. , “RU”, and “RE” are stored as utilization endings.

Based on the morpheme analysis result obtained by the morpheme analyzer 22, the expression replacement unit 26 processes the part-of-speech or inflection type of the morpheme that appears one or two times before the processing target morpheme that is the honored expression. A processing target morpheme that is a sympathetic representation, depending on at least one of the following morpheme or subsequent morpheme sequence notation that follows the target morpheme, the absence of the subsequent morpheme, or whether the subsequent morpheme is a sentence ending symbol Using a plurality of predetermined rules for converting the symbol to the normal expression, the honorific expression included in the input sentence is converted to the normal expression.
The processing target morpheme indicates a morpheme corresponding to the expression “is / mass” included in the input sentence as described later.

  Specifically, the expression replacement unit 26 (A) is one or two before the processing target morpheme, based on the morpheme notation, part of speech, and verb utilization type acquired by the morpheme analysis unit 22. Types of part of speech of morphemes that appear, (B) notation of processing morphemes, (C) notation of subsequent morphemes or subsequent morpheme strings, and (D) whether or not tails are omitted. To perform the replacement process for the “is-mas” expression contained in the input sentence. The expression replacement unit 26 includes a utilization ending acquisition process and a replacement process using 14 types of rules.

  (A) Type of part of speech of morpheme that appears one or two before, (C) Notation of succeeding morpheme or succeeding morpheme string, (D) Presence or absence of tail omission is (B) Processing target morpheme After the “mass” expression is detected, it is grasped by examining the morphemes that appear before and after the morpheme to be processed. (A) Type of part of speech of morpheme that appears one or two before, (B) notation of processing target morpheme, (C) notation of subsequent morpheme or subsequent morpheme string, and (D) presence or absence of tail omission explain.

(A) Type of part of speech of morpheme that appears one or two times before The reference is made to the type of part of speech of the morpheme that appears one time before the morpheme to be processed ("is / mas" expression). When the morpheme that appears immediately before the processing target morpheme is the connection particle “te” or “de”, the morpheme that appears two times before the processing target morpheme (the one immediately before “te” or “de”) Refer to the type of part of speech of morpheme)). In the present embodiment, the type of part-of-speech of the morpheme that appears immediately before is classified into the following three types (a) to (c). In addition, verbs are further subdivided according to utilization types (such as five-stage utilization and one-stage utilization). In the present embodiment, a case will be described as an example in which the part-of-speech and utilization type of a morpheme that appears one or two before the processing target morpheme is used.

(A) Nouns As nouns, nouns (including pronouns, formal nouns, proper expressions (including person names, place names, date expressions, time expressions, etc.)) and adjective verbs are included.

(A) Adjectives Examples of adjectives include adjectives and auxiliary verbs (for example, “no”, “tai”, etc.) that make use of adjective forms.

(C) Verbs As verbs, verbs, auxiliary verbs that make use of the verb type (for example, “being” or “to be”), and auxiliary verbs (for example, “being” or “being”) are included.

  Therefore, the expression replacement unit 26 is a process that is a sympathetic expression depending on whether the type of part of speech of the morpheme that appears one or two before corresponds to a noun class, an adjective class, or a verb class. Based on a plurality of predetermined rules for converting the target morpheme into the normal expression, the honorary expression included in the input sentence is converted into the normal expression.

(B) Notation of processing target morpheme The processing target morpheme is expressed as “is”, “deshi”, “decho”, “masu”, “masashi”, “maso”, “no”. It shall be pointed out either.
In this embodiment, the detection of the morpheme to be processed (detection of the utilization form of “is” and “mas”) is performed based on the “basic form” of the morpheme output as part of the morpheme analysis result. The utilization form of “is” and “masu” may be detected by matching character strings against morpheme or morpheme string notation.
Accordingly, the expression replacement unit 26 converts the processing target morpheme that is “is”, “deshi”, “decho”, “masu”, “masashi”, “maso”, or “no” into a normal expression. Based on a plurality of predetermined rules, the honorific expression included in the input sentence is converted into a regular expression.

(C) Notation of subsequent morpheme or subsequent morpheme string The subsequent morpheme represents a morpheme that appears after the morpheme to be processed. The subsequent morpheme string represents a morpheme string that appears subsequent to the processing target morpheme. In the present embodiment, the subsequent morpheme is represented by “ta” (completion auxiliary verb), “te” (connection particle), “no *” (morpheme whose first character is “no”), “n”. Or a questionable final particle (“ka”, “kashira”, “kana”). Further, the notation of the subsequent morpheme sequence is “n / was” obtained by adding “do” to “n”.
Further, in the present embodiment, the honorific expression included in the input sentence is converted into a regular expression when there is no subsequent morpheme (end of sentence) or when the subsequent morpheme is a sentence end symbol.
Therefore, the expression replacement unit 26 always represents the processing target morpheme that is a honorific expression depending on whether the subsequent morpheme or the subsequent morpheme string is written, there is no subsequent morpheme, or whether the subsequent morpheme is a sentence ending symbol. On the basis of a plurality of predetermined rules for converting to the expression, the honorific expression included in the input sentence is converted into a regular expression.

(D) Presence or absence of tail omission Tail omission is the presence or absence of the abbreviation “teru” between “ver” and “mas” (and its usage), or “ It shall indicate whether or not the abbreviation “de” is present. For example, if the morpheme that appears immediately before the morpheme to be processed is “te” or “de”, it is determined that the tail is omitted. When the morpheme that appears immediately before the morpheme to be processed is not “te” or “de”, it is determined that the tail is not omitted. As an example where the tail is omitted, for example, "I'm climbing" and "I smell".
In the expression replacement unit 26, the part of speech of the morpheme that appears one or two times before the processing target morpheme is a verb class, and between the processing target morpheme and the morpheme that appears one time before the processing target morpheme Depending on whether or not “is” is omitted, based on a predetermined rule for converting a processing target morpheme that is a honorific expression into a regular expression, the honorific expression included in the input sentence is Convert to normal expression.

  (A) to (D) in the flowcharts of FIGS. 3 to 5 to be described later are (A) the type of part of speech of the morpheme that appears one or two before the processing target morpheme, and (B) the processing target morpheme. It shows which condition corresponds to the notation, (C) notation of subsequent morpheme or succeeding morpheme string, and (D) presence or absence of tail omission.

  In the present embodiment, the expression replacement unit 26 is based on the morpheme analysis results obtained by the morpheme analysis unit 22 and has a plurality of rules defined according to the four conditions (A) to (D) described above. As follows, the honorific expression included in the input sentence is converted into the regular expression using the rules shown in the following (1) to (14).

(1) Rule to delete “is” The expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is an adjective and the processing target morpheme is “is”. Sometimes, delete the target morpheme “is” from the input sentence.
For example, when the input sentence “Mt. Fuji is beautiful” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha / beautiful /) obtained by the morpheme analysis unit 22 and the part of speech (noun / particle / particle). Based on the adjective / auxiliary verb), the processing target morpheme “is” is deleted from the input sentence, and the sentence “Mt. Fuji is beautiful” is output.

In addition, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a noun, the processing target morpheme is “is”, and the processing target morpheme is “is”. Deletes the target morpheme “is” from the input sentence.
For example, when the input sentence “Is Mount Fuji is a mountain” is input, the expression replacement unit 26 deletes the processing target morpheme “is” from the input sentence, and the sentence “Mt. Fuji is a mountain” is deleted. Is output.
However, the form of nouns or adjectives + "ka" (+ end-of-sentence symbols (.?! Etc.) or end-of-sentence), such as "mountain" or "beautiful", is a hard utterance in conversation. To give an impression, “ka” may be followed by a question mark (“?”), Or “ka” may be replaced with a question mark.

(2) Rule for Substituting “Doshi” with “Daro” The expression replacement unit 26 has an adjective class for the part of speech of the morpheme that appears immediately before the morpheme to be processed, and the notation of the morpheme to be processed is “ When it is, it replaces the target morpheme “Doshi” with “Daro”.
For example, when the input sentence “Mt. Fuji is beautiful” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha / clean / de / u) obtained by the morpheme analysis unit 22 and the part of speech (noun). / Subject / adjective verb / auxiliary verb / auxiliary verb), the sentence “Doshi” to be processed in the input sentence is replaced with “Daro”, and “Doshi” is replaced with “Daro” "Wax" is output.

  In addition, the expression replacement unit 26 performs the processing target morpheme “syo” when the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a noun and the notation of the processing target morpheme is “sho”. "Is replaced with" Daro ".

(3) Rule that deletes “deshi” and replaces “te” with “de” The expression replacement unit 26 is a noun class for the part of speech of the morpheme that appears before the morpheme to be processed. And when the notation of the target morpheme is “deshi” and the notation of the subsequent morpheme is “te”, the subsequent morpheme “te” is replaced with “de” and the target morpheme “deshi” is deleted. To do.
For example, when the input sentence “Mt. Fuji is beautiful” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha ////) and the part of speech obtained by the morpheme analysis unit 22. Based on (noun / particle / adjective verb / auxiliary verb / particle), the subsequent morpheme “te” is replaced with “de”, the target morpheme “deshi” is deleted, and “te” is replaced with “de”. , The sentence “Mt. Fuji is beautiful” is output with “deshi” removed.

(4) Rules for Substituting “Doshi” with “Da” The expression replacement unit 26 indicates that the type of part-of-speech of the morpheme that appears immediately before the processing target morpheme is a noun, and the processing target morpheme is notated. When it is “deshi” and the notation of the subsequent morpheme is “ta”, the processing target morpheme “deshi” is replaced with “da”.
For example, when the input sentence “Mt. Fuji was beautiful” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha // do / do) obtained by the morpheme analysis unit 22 and the part of speech ( Based on nouns / particles / adjective verbs / auxiliary verbs / auxiliary verbs), the sentence “Mt. Fuji was beautiful”, where the target morpheme “Dashi” was replaced with “Da” and “Doshi” was replaced with “Da”. Is output.

(5) Rule for Substituting “Da” with “Da” The expression replacement unit 26 indicates that the type of part-of-speech of the morpheme that appears immediately before the processing target morpheme is a noun, and the processing target morpheme is represented by “ And the subsequent morpheme notation is not “no *” (the morpheme whose first character is “no”), the processing target morpheme “is” is replaced with “da”.
For example, when the input sentence “Mt. Fuji is beautiful” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha / clean /) obtained by the morpheme analysis unit 22 and the part of speech (noun / particle / particle). Based on the adjective verb / auxiliary verb), the processing target morpheme “is” is replaced with “da”, and the sentence “Mt. Fuji is beautiful” in which “da” is replaced with “da” is output.

(6) Rule for replacing “is” with “na” The expression replacement unit 26 indicates that the type of part-of-speech of the morpheme that appears immediately before the processing target morpheme is a noun, and the processing target morpheme is represented by “ When the notation of the subsequent morpheme is “no *” (the morpheme whose first character is “no”), the processing target morpheme “is” is replaced with “na”.
For example, when the input sentence “Mt. Fuji is a volcano” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / ha / volcano / is / so) obtained by the morpheme analysis unit 22 and the part of speech (noun). Based on / particle / noun / auxiliary verb / particle), the sentence “is” to be processed is replaced with “na”, and the sentence “Mt. Fuji is a volcano” with “na” replaced with “na” is output.

(7) Rule that replaces the inflection ending with the continuous form (sound form), replaces “ta” with “da”, and deletes “masashi”. The expression replacement unit 26 performs processing when “tail omitted” is “none”. The type of part-of-speech of the morpheme that appears immediately before the target morpheme is a verb class, and the usage type of the verb class is a 5-line utilization, 5-line utilization, 5-line utilization, or 5-line utilization When the morpheme used in the fifth stage is “Masashi” and the subsequent morpheme is “Ta”, the ending of the morpheme that appears immediately before the target morpheme (Sound form), the subsequent morpheme “ta” is replaced with “da”, and the processing target morpheme “masashi” is deleted.
The expression replacement unit 26 reads a utilization ending list corresponding to the utilization type of the verb class from the verb utilization table of the verb utilization table database 24, and uses the continuation form (sound form) of the continuous form in the utilization ending list. Is replaced with a morpheme ending that appears immediately before the processing target morpheme.

  For example, when the input sentence “I screamed at Mt. Fuji” is input, the expression replacement unit 26, the morpheme boundary obtained by the morpheme analysis unit 22 (Mt. Fuji / de / scream / better / ta), the part of speech and the verb Based on the utilization type of a class (noun / particle / verb: ba line 5 dan / auxiliary verb / auxiliary verb), the utilization ending of the morpheme that appears immediately before the morpheme to be processed is replaced with the continuous form (stool form), Subsequent morpheme “ta” is replaced with “da”, and processing target morpheme “masashi” is deleted. Then, the expression replacement unit 26 outputs a sentence “screamed at Mt. Fuji” in which the inflection ending is replaced with the continuous form (sound form), “ta” is replaced with “da”, and “masi” is deleted.

(8) A rule that replaces the inflection ending with the continuous form (sound form) and deletes “masashi” The expression replacement unit 26 is one before the morpheme to be processed when “tail omitted” is “none” for the verb class When the type of part of speech of a morpheme that appears in a verb is a verb class and the notation of the processing target morpheme is “better”, the ending of the morpheme that appears immediately before the processing target morpheme is used as a continuous form (sound form ) And delete “Mashi”.
When “tail omission” is “none” for the verb class, a utilization ending list corresponding to the utilization type of the verb class is read from the verb utilization table of the verb utilization table database 24 and the consecutive forms in the utilization ending list are read. The utilization ending of (sound form) is acquired and replaced with the utilization ending of the morpheme that appears immediately before the processing target morpheme.
On the other hand, when “tail omitted” is “present”, the expression replacing unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme has “ ”Is deleted.

  For example, when the input sentence “I climbed Mt. Fuji” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / Ni / Climb / Masashi / Ta) obtained by the morpheme analysis unit 22 and the part of speech and Based on the usage type of the verb class (noun / particle / verb: La line five steps / auxiliary verb / auxiliary verb), the usage ending is replaced with the continuous form (sound form), and the processing target morpheme “masashi” is deleted. Then, the expression replacement unit 26 outputs the sentence “Climbed Mt. Fuji” in which the inflection ending is replaced with the continuous form (sound form) and “Mashi” is deleted.

  In addition, when “tail omission” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the morpheme to be processed is a verb class, and the usage type of the verb class is “Gas”. 5 rows of 5 rows, 5 rows of rows, 5 rows of rows, or 5 rows of rows, the morpheme to be processed is “masashi”, and the subsequent morpheme is “ta”. Sometimes, the processing target morpheme “masashi” is deleted.

  In addition, when “tail omission” is “present”, the expression replacement unit 26 uses a verb class for the part of speech of the morpheme that appears immediately before the morpheme to be processed, and the usage type of the verb class When using 5 dan, 5 banks, 5 banks, or 5 banks, and the morpheme to be processed is “Masashi” and the subsequent morpheme is “te” In addition, the processing target morpheme “masashi” is deleted.

(9) Rule that replaces the inflection ending with the continuous form (sound form), replaces “te” with “de”, and deletes “masashi”. The expression replacement unit 26 performs processing when “tail omitted” is “none”. The type of part-of-speech of the morpheme that appears immediately before the target morpheme is a verb class, and the usage type of the verb class is a 5-line utilization, 5-line utilization, 5-line utilization, or 5-line utilization If the notation of the target morpheme is “Mashi” and the notation of the subsequent morpheme is “te”, the ending of the morpheme that appears immediately before the target morpheme ), The subsequent morpheme “te” is replaced with “de”, and the processing target morpheme “masashi” is deleted.
The expression replacement unit 26 reads a utilization ending list corresponding to the utilization type of the verb class from the verb utilization table of the verb utilization table database 24, and uses the continuation form (sound form) of the continuous form in the utilization ending list. Is replaced with a morpheme ending that appears immediately before the processing target morpheme.

  For example, when the input sentence “scream at Mt. Fuji” is input, the expression replacement unit 26, the morpheme boundary (Mt. Fuji / de / scream / better / te) obtained by the morpheme analysis unit 22, the part of speech and the verb Based on the utilization type of a class (noun / particle / verb: ba line 5 dan / auxiliary verb / particle), replace the utilization ending of the morpheme that appears immediately before the morpheme to be processed with the continuous form (sound form), Subsequent morpheme “te” is replaced with “de”, and processing target morpheme “masashi” is deleted. Then, the expression replacement unit 26 outputs the sentence “Screaming at Mt. Fuji” in which the inflection ending is replaced with the continuous form (sound form), “te” is replaced with “de”, and “masi” is deleted.

(10) Rule that replaces the ending form with the final form and deletes “mas” When the “tail omission” is “none” for the verb class, the expression replacement unit 26 is a morpheme that appears before the morpheme to be processed. When the type of part-of-speech is a verb class and the morpheme of the processing target is "mas", the ending morpheme of the morpheme that appears immediately before the processing target morpheme is replaced with the final form, Delete "mas".
When “tail omitted” is “none” for verbs, the usage ending list corresponding to the usage type of the verbs is read from the verb usage table in the verb usage table database 24, and the end of the usage in the usage ending list. The utilization ending of the shape is acquired and replaced with the utilization ending of the morpheme that appears immediately before the processing target morpheme.
On the other hand, when “tail omitted” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme is “notified”. ”Is read from the verb usage table in the verb usage table database 24 and the final usage ending in the usage ending list is inserted after“ te ”or“ de ”. Then, the target morpheme “mas” is deleted.

  For example, when the input sentence “I climb Mt. Fuji” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / Ni / Climb / Make) obtained by the morpheme analysis unit 22 and the part of speech and verb class. Based on the conjugation type (noun / participant / verb: La line 5 dan / auxiliary verb), the morpheme ending in the morpheme that appears immediately before the target morpheme is replaced with the final form, and the target morpheme “mas” is deleted To do. Then, the expression replacing unit 26 outputs the sentence “Climbing Mt. Fuji” in which the ending is replaced with the final form and “mas” is deleted.

(11) A rule that replaces the inflection ending with a blank form (o-dan) and deletes “maso” The expression replacement unit 26 is one before the morpheme to be processed when “tail omitted” is “none” for the verb class When the type of part of speech of a morpheme that appears in a verb is a verb class and the notation of the processing target morpheme is “Mashou”, the ending of the morpheme that appears immediately before the processing target morpheme ) And delete “Masho”.
When “tail omission” is “none” for a verb class, a utilization ending list corresponding to the utilization type of the verb class is read from the verb utilization table of the verb utilization table database 24, and the occurrence ending list in the utilization ending list is not yet recorded. The utilization ending of the form (o stage) is acquired and replaced with the utilization ending of the morpheme that appears immediately before the morpheme to be processed.
On the other hand, when “tail omitted” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme is represented by “Mashiyo”. ”Is read from the verb usage table in the verb usage table database 24 and the usage endings of the green form (o-dan) in the usage ending list are read by“ te ”or“ de ” ”Is inserted after and“ Masho ”is deleted.

  For example, when the input sentence “Let's climb Mt. Fuji” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / Ni / Climb / Masho / U) obtained by the morpheme analysis unit 22 and the part of speech and Based on the verb usage (noun / participant / verb: La line 5dan / auxiliary verb / auxiliary verb), replace the morphological inflection ending in the previous morpheme with the morpheme (odan). The processing target morpheme “Masho” is deleted. Then, the expression replacing unit 26 outputs a sentence “Let's climb Mt. Fuji” in which the ending ending is replaced with a blank form (o-dan) and “Mashou” is deleted.

(12) Rules for replacing the ending utterance with the command form and deleting “no” When the “tail omission” is “none” for the verb class, the expression replacement unit 26 appears before the morpheme to be processed. If the type of part of speech is a verb class and the morpheme to be processed is "No" and there is no subsequent morpheme or the subsequent morpheme is a sentence ending symbol class (.?! Etc.) The ending of the morpheme that appears immediately before the morpheme is replaced with an instruction form, and the processing target morpheme “N” is deleted.
When “tail omitted” is “none” for a verb class, a utilization ending list corresponding to the utilization type of the verb class is read from the verb utilization table of the verb utilization table database 24, and the instructions in the utilization ending list are read. The utilization ending of the shape is acquired and replaced with the utilization ending of the morpheme that appears immediately before the processing target morpheme.
On the other hand, when “tail omitted” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme is not “notified”. ”And there is no subsequent morpheme, or the subsequent morpheme is an end-of-sentence symbol (.?!, Etc.), the utilization ending list of the one-stage utilization is read from the verb utilization table in the verb utilization table database 24, The usage ending of the command form in the usage ending list is inserted after “te” or “de”, and the processing target morpheme “n” is deleted.

  For example, when the input sentence “Please return” is input, the expression replacement unit 26 determines the morpheme boundary (O / Return / Please / Ne) obtained by the morpheme analysis unit 22 and the part of speech and verb class. Based on the conjugation type (prefix / verb: La line 5 dan / auxiliary verb: la line 5 dan / auxiliary verb), replace the morphological inflection ending of the morpheme that appears before the morpheme to be processed with the imperative form. Delete the morpheme "Mase". Then, the expression replacement unit 26 outputs the sentence “Please return”, in which the ending is replaced with the command form, and “se” is deleted.

  In addition, there are things that use the imperative ending in the colloquial style, such as “Nasaru”, “Isurusuru”, “Suru”, “Osaru”, etc., but in the colloquial style, “I” is used as an imperative inflection. In this case, “i” is used instead of “re” in this embodiment.

(13) A rule that replaces the inflection ending with the original form (stage A), replaces “n” with “not”, and deletes “not”. The expression replacement unit 26 sets “tail omitted” to “none” for verbs. The type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, the processing target morpheme is “not”, and the subsequent morpheme is “n”, and When the subsequent morpheme is not followed by “do” (when the subsequent morpheme string is not “n / do”), the morpheme endings that appear immediately before the morpheme to be processed are changed to the previous morpheme. Replace the subsequent morpheme “n” with “not”, and delete the processing target morpheme “n”.
When “tail omission” is “none” for a verb class, a utilization ending list corresponding to the utilization type of the verb class is read from the verb utilization table of the verb utilization table database 24, and the occurrence ending list in the utilization ending list is not yet recorded. The utilization ending of the shape (stage) is acquired and replaced with the utilization ending of the morpheme that appears immediately before the processing target morpheme.
On the other hand, when “tail omitted” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme is not “notified”. , And the subsequent morpheme is “n”, and the subsequent morpheme is not followed by “do” (when the subsequent morpheme string is not “n / do”) The list is read from the verb usage table of the verb usage table database 24, and the usage ending of the green form (A column) in the usage ending list is inserted after “te” or “de”, and the subsequent morpheme “n” is inserted. Replace with “None” and delete the processing target morpheme “N”.

  For example, when the input sentence “I do not climb Mt. Fuji” is input, the expression replacement unit 26 uses the morpheme boundary (Mt. Fuji / Ni / Climb / Ne / Nen) obtained by the morpheme analysis unit 22 and the part of speech and Based on the verb type (noun / participant / verb: La line five steps / auxiliary verb / auxiliary verb), replace the morphological inflection ending of the morpheme that appears immediately before the morpheme to be processed with the blank form (A step). The subsequent morpheme “n” is replaced with “not”, and the processing target morpheme “don” is deleted. Then, the expression replacement unit 26 outputs the sentence “Don't climb Mt. Fuji” in which the inflection ending is replaced with a blank form (A-stage), “N” is replaced with “None”, and “N” is deleted.

  Even if the process of replacing the subsequent morpheme “n” with “not” is not performed, there is no grammatical problem in Japanese (eg, climbing Mt. Fuji). However, because it seems unnatural as a colloquial of modern standard language (Tokyo dialect), in this embodiment, the process of replacing the subsequent morpheme “n” with “not” is performed, and the ending utterance is changed to the unfinished form (A ).

(14) A rule that replaces the inflection ending with the original form (A), replaces “Denshi” with “None”, and deletes “Not”. The expression replacement unit 26 sets “tail omission” to “none” for verbs. In this case, the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, the processing target morpheme is “not”, and the subsequent morpheme is “n”, And when “D” followed the subsequent morpheme (when the subsequent morpheme string is “N / D”), the ending morpheme of the morpheme that appears immediately before the target morpheme is changed to the previous morpheme. ), “Denshi” is replaced with “None”, and the processing target morpheme “N” is deleted.
When “tail omission” is “none” for a verb class, a utilization ending list corresponding to the utilization type of the verb class is read from the verb utilization table of the verb utilization table database 24, and the occurrence ending list in the utilization ending list is not yet recorded. The utilization ending of the shape (stage) is acquired and replaced with the utilization ending of the morpheme that appears immediately before the processing target morpheme.
On the other hand, when “tail omitted” is “present”, the expression replacement unit 26 indicates that the type of part of speech of the morpheme that appears immediately before the processing target morpheme is a verb class, and the processing target morpheme is not “notified”. , And the subsequent morpheme is “n”, and the subsequent morpheme is followed by “do” (when the subsequent morpheme string is “n / do”). The ending list is read from the verb usage table in the verb usage table database 24, and the usage ending of the green form (A column) in the usage ending list is inserted after “te” or “de”, and “denshi” is inserted. Replace with “None” and delete the target morpheme “N”.

  For example, when the input sentence “I did not climb Mt. Fuji” is input, the expression replacement unit 26 uses the morpheme boundary obtained by the morpheme analysis unit 22 (Mt. Fuji / n / climb / don / don / de / And morphological inflections that appear immediately before the morpheme to be processed, based on the use type of part of speech and verbs (noun / participant / verb: La line five steps / auxiliary verb / auxiliary verb / auxiliary verb / auxiliary verb) Is replaced with a blank form (stage A), “Denshi” is replaced with “None”, and the morpheme to be processed “Not” is deleted. Then, the expression replacement unit 26 outputs the sentence “I did not climb Mt. Fuji” in which the ending ending was replaced with a blank form (A column), “Denshi” was replaced with “None”, and “Not” was deleted. To do.

  In the rules (1) to (14) above, when the morpheme that appears one before the processing target morpheme is the connection particle “te” or “de”, the morpheme that appears one time before the processing target morpheme Instead, refer to the type of part of speech of the morpheme that appears two times before the processing target morpheme (the morpheme immediately before “te” or “de”), and the morpheme that appears two times before the processing target morpheme Apply rules according to part of speech.

  The output unit 30 outputs the sentence output by the expression replacement unit 26 as a result.

<Operation of style conversion device>
Next, the operation of the style conversion device 100 according to the present embodiment will be described. When a sentence to be converted is input to the style conversion apparatus 100, the expression substitution processing routine shown in FIGS. 3 to 5 is executed by the style conversion apparatus 100.

  First, in step S <b> 100, the input sentence 10 is received as an input sentence by the input unit 10.

  In step S102, the morpheme analysis unit 22 performs morpheme analysis on the input sentence received in step S100, and acquires morpheme boundaries, parts of speech, and utilization types of verbs.

In step S104, the expression replacement unit 26 determines the type of part of speech of the morpheme that appears immediately before the processing target morpheme of the input sentence based on the morpheme analysis result acquired in step S102. When the type of part of speech of the morpheme that appears immediately before the processing target morpheme of the input sentence is an adjective class, the process proceeds to step S106. When the type of part of speech of the morpheme that appears immediately before the processing target morpheme of the input sentence is a noun class, the process proceeds to step S112. When the type of part of speech of the morpheme that appears immediately before the processing target morpheme of the input sentence is a verb class, the process proceeds to step S128.
In addition, when the morpheme that appears immediately before the processing target morpheme is the connection particle “te” or “de”, instead of the morpheme that appears before the processing target morpheme, it is two before the processing target morpheme. Referring to the morphemes appearing in, it is determined whether the type of part of speech of the morpheme that appears two times before the processing target morpheme is an adjective class, a noun class, or a verb class.

  In step S106, the representation replacement unit 26 determines the notation of the processing target morpheme based on the morpheme analysis result acquired in step S102. If the notation of the morpheme to be processed is “I”, the process proceeds to step S108. On the other hand, if the notation of the processing target morpheme is “de”, the process proceeds to step S110.

  In step S108, the expression replacement unit 26 deletes the processing target morpheme “is” from the input sentence.

  In step S 110, the expression replacement unit 26 replaces the processing target morpheme “desh” included in the input sentence with “daro”.

  In step S112, the expression replacement unit 26 determines whether or not a questionable final particle (such as “ka”) follows the processing target morpheme based on the morpheme analysis result acquired in step S102. When a question final particle (such as “ka”) follows the processing target morpheme, the process proceeds to step S108. On the other hand, if no question final particle follows the processing target morpheme, the process proceeds to step S114.

  In step S114, the representation replacement unit 26 determines the notation of the processing target morpheme based on the morpheme analysis result acquired in step S102. If the notation of the processing target morpheme is “de”, the process proceeds to step S110. If the notation of the processing target morpheme is “deshi”, the process proceeds to step S116. When the notation of the processing target morpheme is “is”, the process proceeds to step S122.

  In step S116, the representation replacement unit 26 determines the notation of the subsequent morpheme based on the morpheme analysis result acquired in step S102. If the notation of the subsequent morpheme is “te”, the process proceeds to step S118. On the other hand, if the notation of the processing target morpheme is “ta”, the process proceeds to step S120.

  In step S118, the expression replacement unit 26 deletes the processing target morpheme “deshi” included in the input sentence, and replaces the subsequent morpheme “te” with “de”.

  In step S120, the expression replacement unit 26 replaces the processing target morpheme “deshi” included in the input sentence with “da”.

  In step S122, the expression replacement unit 26 determines the notation of the subsequent morpheme based on the morpheme analysis result acquired in step S102. If the subsequent morpheme is not “*”, the process proceeds to step S124. On the other hand, if the notation of the processing target morpheme is “no *”, the process proceeds to step S126.

  In step S124, the expression replacement unit 26 replaces the processing target morpheme “is” included in the input sentence with “da”.

  In step S 126, the expression replacement unit 26 replaces the processing target morpheme “is” included in the input sentence with “na”.

  In step S128, the expression replacement unit 26 determines whether or not “tail” is omitted between the verb class and the processing target morpheme, based on the morpheme analysis result acquired in step S102. If “tail” is omitted, the process proceeds to step S130. On the other hand, if “tail” is not omitted, the process proceeds to step S132.

  In step S130, the expression replacement unit 26 reads the utilization ending list of the one-stage utilization from the verb utilization table of the verb utilization table database 24, and acquires the utilization endings in the utilization ending list of the one-stage utilization.

  In step S132, the expression replacement unit 26 reads a utilization ending list corresponding to the utilization type of the verb class from the verb utilization table of the verb utilization table database 24, and acquires the utilization ending in the utilization ending list.

  In step S134, the representation replacement unit 26 determines the notation of the processing target morpheme based on the morpheme analysis result acquired in step S102. If the notation of the processing target morpheme is “better”, the process proceeds to step S136. If the notation of the morpheme to be processed is “mas”, the process proceeds to step S148. If the notation of the morpheme to be processed is “Masho”, the process proceeds to step S150. If the notation of the processing target morpheme is “not”, the process proceeds to step S152.

  In step S136, the expression replacement unit 26 determines the notation of the subsequent morpheme based on the morpheme analysis result acquired in step S102. If the notation of the subsequent morpheme is “ta”, the process proceeds to step S138. If the notation of the subsequent morpheme is “te”, the process proceeds to step S144.

  In step S138, the expression replacement unit 26 determines the usage type of the verb class based on the morphological analysis result acquired in step S102. When the usage type of the verb class is five-stage utilization, five-stage utilization, five-stage utilization, five-stage utilization, or five-stage utilization, the process proceeds to step S139. On the other hand, if the usage type of the verb class is not the fifth row utilization, the fifth row utilization, the fifth row utilization, or the fifth row utilization of the row, the process proceeds to step S142.

  In step S139, the expression replacement unit 26 determines whether or not “tail” is omitted between the verb class and the processing target morpheme, based on the morpheme analysis result acquired in step S102. If “tail” is omitted, the process proceeds to step S142. On the other hand, if “tail” is not omitted, the process proceeds to step S140.

  In step S140, based on the utilization ending of the continuous form (sound form) in the utilization ending list acquired in step S130 or S132 by the expression replacement unit 26, the utilization ending of the input sentence is converted to the continuous form (sound form). Then, the subsequent morpheme “ta” is replaced with “da”, and the processing target morpheme “masashi” is deleted.

  In step S142, based on the utilization ending of the continuous form (sound form) in the utilization ending list acquired in step S130 or step S132 by the expression replacement unit 26, the utilization ending of the input sentence is converted to the continuous form (sound form). Replace and delete the morpheme “masashi” to be processed.

  In step S144, the expression replacement unit 26 determines the verb utilization type based on the morphological analysis result acquired in step S102. If the usage type of the verb class is five-stage utilization, five-stage utilization, five-stage utilization, five-stage utilization, or five-stage utilization, the process proceeds to step S145. On the other hand, if the usage type of the verb class is not the fifth row utilization, the fifth row utilization, the fifth row utilization, or the fifth row utilization of the row, the process proceeds to step S142.

  In step S145, the expression replacement unit 26 determines whether or not “tail” is omitted between the verb class and the processing target morpheme, based on the morpheme analysis result acquired in step S102. If “tail” is omitted, the process proceeds to step S142. On the other hand, if “tail” is not omitted, the process proceeds to step S146.

  In step S146, the expression replacement unit 26 converts the used endings of the input sentence to the continuous form (sound form) based on the used endings of the continuous form (sound form) in the use ending list acquired in step S130 or S132. Replace the subsequent morpheme “te” with “de”, and delete the processing target morpheme “masashi”.

  In step S148, the expression replacement unit 26 replaces the utilized endings of the input sentence with the ending forms based on the ending forms of the ending forms in the utilization ending list acquired in step S130 or step S132, and processes the morpheme to be processed. "Is deleted.

  In step S150, the expression replacement unit 26 changes the used ending of the input sentence to the uncorrected form (o stage) based on the used ending of the unusable form (o stage) in the used ending list acquired in step S130 or step S132. Replace and delete the processing target morpheme “Mashi”.

  In step S152, the expression replacement unit 26 determines the subsequent morpheme based on the morpheme analysis result acquired in step S102. If there is no subsequent morpheme, or if the subsequent morpheme is a sentence ending symbol (.?!, Etc.), the process proceeds to step S154. On the other hand, if the notation of the processing target morpheme is “n”, the process proceeds to step S156.

  In step S154, the expression replacement unit 26 replaces the utilization ending of the input sentence with the instruction form based on the utilization ending of the instruction form in the utilization ending list acquired in step S130 or step S132, and processes the morpheme to be processed. "Is deleted.

  In step S156, the expression replacement unit 26 determines whether or not “do” follows the subsequent morpheme. If “do” follows the subsequent morpheme, the process proceeds to step S160. On the other hand, if “do” does not follow the subsequent morpheme, the process proceeds to step S158.

  In step S158, the expression replacement unit 26 converts the used endings of the input sentence to the undeformed form (step A) based on the used endings of the undescripted form (step A) in the list of used endings acquired in step S130 or step S132. Replace the subsequent morpheme “n” with “not”, and delete the processing target morpheme “n”.

  In step S160, the expression replacement unit 26 converts the used endings of the input sentence to the uncorrected form (stage A) based on the used endings in the used form (stage A) in the list of used endings acquired in step S130 or S132. Replace “Denshi” with “None”, and delete the processing target morpheme “Nen”.

  In step S162, the sentence converted in any of the steps S108, S110, S118, S120, S124, S126, S140, S142, S146, S148, S150, S154, S158, and S160 is output as a result. Then, the expression replacement processing routine is completed.

  As described above, according to the stylistic conversion device according to the present embodiment, the part-of-speech or inflection type of the morpheme that appears immediately before the processing target morpheme that is the honorific expression, and the processing target morpheme follow. The morpheme to be processed is a regular expression according to at least one of the following morpheme or subsequent morpheme string notation, absence of a subsequent morpheme, or whether the subsequent morpheme is a sentence ending symbol. Based on a predetermined rule for conversion, the honorific expression included in the input sentence is converted into the normal expression, so that the honorific expression can be accurately converted into the normal expression.

Further, according to the stylistic conversion device according to the present embodiment, it is possible to automatically unify a document in which normal expression and respected expression are mixed into normal expression.
If the stylistic conversion device according to the present embodiment is applied to document proofreading, it is possible to support unifying the style of the document. Moreover, if the stylistic conversion device according to the present embodiment is applied to utterance generation of a computer (dialogue system) that interacts with humans, the system utterance database can be unified all the time, and the system's way of speaking ( Can be made consistent.

  Note that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the gist of the present invention.

  Moreover, although the style conversion apparatus 100 demonstrated the case where the verb utilization table database 24 was provided, the verb utilization table database 24 is provided in the external device of the style conversion apparatus 100, for example, and the style conversion apparatus 100 is connected with an external device. You may make it refer to the verb utilization table | surface database 24 by communicating using a communication means.

  The above-described style conversion device 100 has a computer system inside, but the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.

  In the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium.

DESCRIPTION OF SYMBOLS 10 Input part 20 Operation part 22 Morphological analysis part 24 Verb utilization table database 26 Expression substitution part 30 Output part 100 Stylistic conversion apparatus

Claims (3)

An input means for receiving an input sentence;
Part-of-speech or inflection type of a morpheme that appears one or two times before a processing target morpheme that is a honorific expression, and a notation of a subsequent morpheme or subsequent morpheme string that appears following the processing target morpheme, the subsequent morpheme Based on a predetermined rule for converting the processing target morpheme that is the honorific expression into a normal expression according to at least one of whether or not the subsequent morpheme is a sentence ending symbol And an expression replacement means for converting the honorific expression included in the input sentence into a regular expression, and the rule is:
(1) Rule to delete “is” (2) Rule to replace “Doshi” with “Daro” (3) Rule to delete “Doshi” and replace “te” with “de” (4) “ Rules for replacing "Dashi" with "Da" (5) Rules for replacing "Da" with "Da" (6) Rules for substituting "Da" with "Na" (7) Consecutive endings (sound form) ), A rule that replaces “ta” with “da”, and deletes “masashi” (8) a rule that replaces the inflection ending with the consecutive form (voice form), and a rule that deletes “masi” (9) the inflection ending A rule that replaces “te” with “de” and deletes “masashi” in the continuous form (sound flight form) (10) A rule that replaces the ending ending with the ending form and deletes “mas” (11) ending ending Is a rule that replaces the original form (o-dan) and deletes “Mashi” (12) A rule that replaces the inflection ending with an instruction form and deletes “not” (13) Rule that replaces the inflection ending with the original form (A column), replaces “N” with “None”, and deletes “N” (14) The ending ending item with the original form (A column), “No” A stylistic conversion device including a rule that replaces “” and deletes “no”. A stylistic conversion method in a stylistic conversion device including an input means and an expression replacement means,
The input means accepting an input sentence;
The part-of-speech or inflection type of the morpheme that appears one or two times before the processing target morpheme that is the honorific expression, and the subsequent morpheme or subsequent morpheme sequence that appears after the processing target morpheme In accordance with at least one of the following notation, the absence of the succeeding morpheme, or whether the succeeding morpheme is a sentence ending symbol or not. Based on the determined rule, converting the honorific expression included in the input sentence into a regular expression,
The rule is
(1) Rule to delete “is” (2) Rule to replace “Doshi” with “Daro” (3) Rule to delete “Doshi” and replace “te” with “de” (4) “ Rules for replacing "Dashi" with "Da" (5) Rules for replacing "Da" with "Da" (6) Rules for substituting "Da" with "Na" (7) Consecutive endings (sound form) ), A rule that replaces “ta” with “da”, and deletes “masashi” (8) a rule that replaces the inflection ending with the consecutive form (voice form), and a rule that deletes “masi” (9) the inflection ending A rule that replaces “te” with “de” and deletes “masashi” in the continuous form (sound flight form) (10) A rule that replaces the ending ending with the ending form and deletes “mas” (11) ending ending Is a rule that replaces the original form (o-dan) and deletes “Mashi” (12) A rule that replaces the inflection ending with an instruction form and deletes “not” (13) Rule that replaces the inflection ending with the original form (A column), replaces “N” with “None”, and deletes “N” (14) The ending ending item with the original form (A column), “No” The method of stylistic conversion is characterized by including a rule that replaces “” and deletes “No”.   The program for functioning a computer as each means of the style conversion apparatus of Claim 1.

Images