JP2017078978A - Natural language processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a natural language processor capable of converting document data, such as news articles, into expressions that are subjective and easily convey true intention as if a human type robot, a personified character, and the like had feelings in the human type robot, the personified character, and the like.SOLUTION: In a natural language processing part 20, 5W1H elements are determined by analyzing document data. In a scenario creation processing part 30, an adverbial usage calculation part 31 calculates adverbial usages corresponding to a prescribed adverbial usage calculation reference by referring to an adverbial usage DB 50. An adverbial usage replacement part 32 replaces prescribed 5W1H elements of the document data by adverbial usages calculated by an adverbial usage calculation part 31. A document output part 40 outputs new document data with the 5W1H elements replaced by the adverbial usage.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a natural language processing apparatus that executes natural language processing of document data such as a news article in a humanoid robot or anthropomorphic character.

  In recent years, with the improvement of natural language processing technology such as morphological analysis and syntactic analysis for document data, human-type robots and anthropomorphic characters are used to read out document data such as news articles. Yes. For example, as represented by Pepper (registered trademark) of SOFTBANK CORP., It has been noticed that document data such as news articles is collected from the Internet and read out in response to a human request.

  However, in conventional humanoid robots and anthropomorphic characters, document data such as news articles are called “reporting practices”, which are standard formats such as “what”, “where” and “what” With 5W1H, reading was repeated monotonously and not subjectively, and received a mechanical impression with no emotion.

  In this regard, in past natural language processing techniques, it has been disclosed to perform language processing in accordance with various rules instead of reading out document data as it is (see Patent Documents 1 to 4).

JP 2008-152637 A JP 2007-310829 A JP 2002-183117 A JP-A-7-175808

  However, any natural language processing technology still has a mechanical impression without feelings.

  The present invention has been made in view of the above-described technical background, and it is subjective in human-like robots, anthropomorphic characters, etc. as if the document data such as news articles have emotions. It is an object of the present invention to provide a natural language processing device that can be converted into an expression that is easy to convey with real intention.

In order to achieve the above object, the present invention analyzes a document input unit to which document data is input and the document data input by the document input unit, so that at least what (What), who (Who), A natural language processing unit for determining one or more 5W1H elements of when, where, why, and why, and adverbial usage calculation for calculating adverbial usage The adverbial usage database stored while the criteria and the adverbial usage corresponding to the adverbial usage calculation criterion are associated with each other, and the document data on which the 5W1H element determination is performed by the natural language processing unit, By referring to the adverbial usage database, an adverbial usage calculation unit that calculates adverbial usage corresponding to a predetermined adverbial usage calculation standard, and the document data An adverbial usage replacement unit that replaces an element of 5W1H with an adverbial usage calculated by the adverbial usage calculation unit, and a new 5W1H element replaced with an adverbic usage by the adverbial usage replacement unit And a document output unit for outputting document data.
In addition, in the present invention, in addition to adverbs in Japanese grammar (mainly a predicate (a verb / adjective / adjective verb), a modifier), as well as adverbs in English grammar, nouns, pronouns The phrase, clause, or entire sentence is explained, and the usage related to the part-of-speech, phrase, or clause as a modifier is defined as “adverbial usage”.

  According to this, in order to calculate the adverbial usage based on the contents of the document data and to replace the 5W1H element of the document data with the adverbial usage, in a humanoid robot, anthropomorphized character, etc. It is possible to convert document data such as articles into an expression that is easy to convey subjective and sincere feeling as if it has emotion.

  The adverbial usage database stores 5W1H elements to be substituted for adverbial usage while being associated with adverbial usage calculation criteria, and the adverbial usage calculation unit uses the natural language processing unit to store 5W1H elements. By referring to the adverbial usage database, the adverbial usage corresponding to a predetermined adverbial usage calculation standard and the 5W1H element to be replaced with the adverbial usage are obtained. And the adverbial usage replacement unit replaces the 5W1H element of the document data with the adverbial usage on the basis of the adverbial usage calculated by the adverbial usage calculation unit and the 5W1H element calculation result. Is preferred.

  Further, the adverbial usage database stores the adverbial usage insertion position in the document data while being associated with the adverbial usage calculation criteria, and the adverbial usage calculation unit determines the 5W1H element by the natural language processing unit. By referring to the adverbial usage database, the adverbial usage corresponding to the predetermined adverbial usage calculation standard, the 5W1H element to be replaced with the adverbial usage, and the document data The adverbial usage replacement unit calculates the adverbial usage calculated by the adverbial usage calculation unit, the element of 5W1H, and the calculation result of the insertion position of the adverbial usage. Based on this, it is preferable that the 5W1H element of the document data is replaced with the adverbial usage and then inserted into the insertion position of the adverbial usage in the document data.

  Further, the adverbial usage database is stored with the criteria related to the user's personality being associated with the adverbial usage calculation criteria, and the adverbial usage calculation unit determines the 5W1H element by the natural language processing unit. By referring to the adverbial usage database for the document data, an adverbial usage corresponding to a predetermined adverbial usage calculation criterion may be calculated according to a criterion related to the personality of the user.

  In addition, the adverbial usage calculation standard relates to a counter word, and the adverbial usage calculation unit refers to the adverbial usage database for the document data for which the natural language processing unit has determined the 5W1H element. Accordingly, an adverbial usage corresponding to the adverbial usage calculation standard for a predetermined counter-word may be calculated.

  The adverbial usage calculation standard relates to a connection pattern, and the adverbial usage calculation unit refers to the adverbial usage database for the document data for which the natural language processing unit has determined the 5W1H element. By doing so, the adverbial usage corresponding to the adverbial usage calculation standard for the predetermined connection pattern may be calculated.

  The adverbial usage calculation standard relates to exaggerated expressions, and the adverbial usage calculation unit refers to the adverbial usage database for the document data for which the natural language processing unit has determined the 5W1H element. By doing so, an adverbial usage corresponding to the adverbial usage calculation standard for a predetermined exaggerated expression may be calculated.

  According to the present invention, in order to calculate the adverbial usage based on the content of the document data and replace the 5W1H element of the document data with the adverbial usage, in a humanoid robot, anthropomorphic character, etc. It is possible to convert document data such as news articles into a subjective and true expression that feels as if it has emotion.

It is a figure which shows schematic structure of this apparatus. It is a figure explaining analysis of document data by a natural language processing part. It is a figure explaining substitution of adverbial usage of document data concerning Example 1. FIG. It is a figure which shows the correspondence of the adverbial usage calculation standard which concerns on Example 1. FIG. It is a figure explaining substitution of adverbial usage of document data concerning Example 2. FIG. It is a figure which shows the correspondence of the adverbial usage calculation standard which concerns on Example 2. FIG. It is a figure which shows the correspondence of the personality reference | standard which concerns on Example 2. FIG. It is a figure explaining substitution of adverbial usage of document data concerning Example 3. It is a figure which shows the correspondence of the adverbial usage calculation standard which concerns on Example 3. FIG. It is a figure which shows the correspondence of the word which concerns on Example 3, and a word group. It is a figure explaining substitution of the adverbial usage of the document data concerning Example 4. It is a figure which shows the correspondence of the adverbial usage calculation standard which concerns on Example 4. FIG. It is a figure explaining substitution of adverbial usage of document data concerning Example 5. It is a figure which shows the correspondence of the adverbial usage calculation standard which concerns on Example 5. FIG.

  Next, an embodiment of a natural language processing apparatus (hereinafter referred to as the present apparatus 1) according to the present invention will be described with reference to FIG.

  As shown in FIG. 1, the apparatus 1 includes a document input unit 10, a natural language processing unit 20 provided on the output side of the document input unit 10, and a scenario generation provided on the output side of the natural language processing unit 20. A processing unit 30 and a document output unit 40 provided on the output side of the scenario generation processing unit 30 are provided.

  The document input unit 10 receives document data such as news articles distributed from the Internet or read by a user.

  The natural language processing unit 20 includes a morphological analysis unit 21 provided on the output side of the document input unit 10, a syntax analysis unit 22 provided on the output side of the morpheme analysis unit 21, and an output side of the syntax analysis unit 22. And a 5W1H determination unit 23 provided.

  The morpheme analysis unit 21 performs morpheme analysis of each sentence of the document data input by the document input unit 10. This morphological analysis uses grammatical knowledge (collection of grammar rules) and a dictionary (a word list with information such as parts of speech) as information sources, and linguistics called morphemes are sentences written in natural language. The work of dividing each part of speech into the smallest parts of speech that have the meaning above is shown. For example, as shown in FIG. 2, when morphological analysis is performed on the sentence “The first snow in Kumejima was recorded in the afternoon of February 17, 1977.”, “1977” “Year” “2”, “Month”, “17”, “Day”...

  The syntax analysis unit 22 performs a syntax analysis of each sentence of the document data subjected to the morphological analysis by the morpheme analysis unit 21. This parsing shows the work of making words that have meaning and purpose called phrases by using parts of speech divided into minimum units. For example, as shown in FIG. 2, when syntax analysis is performed on a sentence that has been morphologically analyzed as described above, ““ February 17, 1977 ”“ Afternoon ”” “In Kumejima” “ It is a group of words with meaning and purpose, such as “First in history”, “Naru”, “Yukiga”, and “Recorded”.

  The 5W1H determination unit 23 performs 5W1H determination on each sentence of the document data analyzed by the syntax analysis unit 22. This determination of 5W1H means what (What), who (Who), what (When), where (Where), why (Why) for each phrase that is a group of words having meaning and purpose in each sentence. ), What (how) element is determined. In the present embodiment, five general elements are determined as 5W1H, but elements such as “who”, “how much”, and “how many” are also determined. Also good. For example, as shown in FIG. 2, when a 5W1H determination is performed on the above-mentioned parsed sentence, WHEN “after noon on February 17, 1977”, WHERE “in Kumejima”, WHAT “first in observation history” It is determined that “the snow that becomes” is “recorded”.

  The scenario generation processing unit 30 includes an adverbial usage calculation unit 31 provided on the output side of the 5W1H determination unit 23, and an adverbial usage replacement unit 32 provided on the output side of the adverbial usage calculation unit 31. An adverbial usage database (hereinafter referred to as adverbial usage DB 50) is connected in a communicable state.

  This adverbial usage DB 50 includes adverbial usage calculation criteria for calculating adverbial usage, adverbial usage corresponding to the adverbial usage calculation criteria, and elements of 5W1H to be replaced with the adverbial usage, The insertion position of the adverbial usage in the document data is stored while being associated with each other.

  The adverbial usage calculation unit 31 refers to the adverbial usage DB 50 with respect to the document data for which the 5W1H determination unit 23 has determined the 5W1H element, and thereby corresponds to a predetermined adverbial usage calculation standard. The usage, the 5W1H element to be replaced with the adverb usage, and the insertion position of the adverb usage in the document data are calculated. Specific methods for calculating adverbial usage will be specifically described in Examples 1 to 5.

  In addition, in the present invention, in addition to adverbs in Japanese grammar (mainly a predicate (a verb / adjective / adjective verb), a modifier), as well as adverbs in English grammar, nouns, pronouns The phrase, clause, or entire sentence is explained, and the usage related to the part-of-speech, phrase, or clause as a modifier is defined as “adverbial usage”.

  The adverbial usage replacement unit 32 calculates the 5W1H element of the document data based on the adverbial usage calculated by the adverbial usage calculation unit 31, the 5W1H element, and the calculation result of the insertion position of the adverbial usage. Is replaced with the adverbial usage and inserted at the insertion position of the adverbial usage in the document data. The replacement method of the adverbial usage will be specifically described in Examples 1 to 5.

  The document output unit 40 outputs, in a predetermined format, new document data in which the 5W1H element is replaced with the adverbial usage by the adverbial usage replacement unit 32. When outputting this new document data, for example, it is possible to output it in a voice format by a humanoid robot or a virtual character.

  In the present embodiment, a case where an element of “What” in document data is replaced with an adverbial usage will be described with reference to FIGS. 3 and 4.

  As shown in FIG. 4 (a), the adverbial usage DB 50 includes an adverbial usage calculation standard for calculating an adverbial usage, an adverbial usage corresponding to the adverbial usage calculation standard, and the adverbial usage. The 5W1H element to be replaced and the insertion position of the adverbial usage in the document data are stored while being associated with each other.

  This adverbial usage calculation standard is a calculation standard for the element of “HOW” of 5W1H, such as “when HOW contains an apology”, “when HOW is positive”, and “when HOW is negative”.

  The adverbial usage corresponding to these adverbial usage calculation standards relates to, for example, apology, gratitude, grief, etc. As shown in FIG. ”,” “Thank you” for appreciation, “Unfortunately” for grief, and “I ’m sorry,” respectively. In addition, the concrete expression of such an adverbic usage respond | corresponds similarly in Example 2-5.

  Further, the element of 5W1H to be replaced with these adverbial usages is set to “WHEN”.

  Also, the insertion position of the adverbial usage in the document data is set to “sentence”.

  Thus, when the document data A1 shown in FIG. 3 is input to the document input unit 10, in the natural language processing unit 20, the morpheme analysis unit 21 performs morphological analysis as shown in the document data A2 in FIG. After the syntax analysis unit 22 performs the syntax analysis as shown in the document data A3 in FIG. 3, the 5W1H determination unit 23 performs the determination of 5W1H as shown in the document data A4 in FIG.

  Next, in the scenario generation processing unit 30, the adverbial usage calculation unit 31 refers to the adverbial usage DB 50 for the document data A4 in FIG. Calculate usage. In the document data A4 of the present embodiment, since an apology is included in the HOW, the adverbial usage (apology) corresponding to the adverbial usage calculation criterion “HOW includes an apology” is calculated. Also, the 5W1H element “WHEN” and the insertion position “sentence” to be replaced with this adverbial usage (apology) are calculated.

  Then, the adverbial usage replacement unit 32 calculates the 5W1H element of the document data based on the adverbial usage calculated by the adverbial usage calculation unit 31, the 5W1H element, and the calculation result of the insertion position of the adverbial usage. Is replaced with the adverbial usage and then inserted at the insertion position of the adverbial usage in the document data. In the document data A4 of this embodiment, as shown in the document data A5 of FIG. 3, “June 1” of the element “WHEN” of 5W1H is used as an adverbial usage (apology) ““ Do not cause trouble. Replace with "and" and insert it at the beginning of the sentence.

  After that, in the document output unit 40, the element “WHEN” of 5W1H “June 1” has been replaced with the adverbial usage (apology) “Make no inconvenience”. The document data A5 shown in FIG. 3 is output in a predetermined format.

  In the present embodiment, a case where the 5W1H element of the document data is replaced with an adverbial usage according to the personality criterion of the user will be described with reference to FIGS.

  As shown in FIG. 6, the adverbial usage DB 50 includes adverbial usage calculation criteria for calculating adverbial usage, criteria related to user personality, adverbial usage corresponding to the adverbial usage calculation criteria, The 5W1H element to be replaced by the adverbial usage and the insertion position of the adverbial usage in the document data are stored while being associated with each other.

  The user personality standards indicate the user's thinking tendency. For example, the political thoughts are determined to be right or left, and the overseas soldiers are considered hawks or pigeons. In addition, the criteria regarding the personality of the user may be determined in advance by the user's selection or the like, or may be determined based on the contents of the document data. In particular, when determining personality standards based on the contents of document data, as shown in FIG. 7, a discrimination is set for each user's thinking tendency, and in the case of one thinking tendency, plus, the other political tendency In this case, as a minus, it is possible to define a standard for the user's personality after quantifying the thinking tendency.

  Thus, when the document data B1 shown in FIG. 5 is input to the document input unit 10, in the natural language processing unit 20, the morpheme analysis unit 21 performs morpheme analysis as shown in the document data B2 in FIG. After the analysis unit 22 performs syntax analysis as shown in the document data B3 in FIG. 5, the 5W1H determination unit 23 performs determination of 5W1H as shown in the document data B4 in FIG.

  Next, in the scenario generation processing unit 30, the adverbial usage calculation unit 31 refers to the adverbial usage DB 50 for the document data B4 in FIG. The adverbial usage corresponding to the standard usage calculation standard is calculated. In the document data B4 of this embodiment, since the description of the constitutional amendment or constitutional amendment is in “WHAT”, when the user's personality standard is “politics is right”, the user's personality standard is “politics is right” Accordingly, the adverbial usage (sorry) corresponding to the adverbial usage calculation standard “WHAT has descriptions of constitutional amendments and constitutional amendments” is calculated. On the other hand, if the user's personality is “politics is left”, it corresponds to the adverbial usage calculation standard “WHAT has a description of constitutional or constitutional amendments in WHAT” according to the standard “politics is left” regarding the user ’s personality Calculate adverbial usage (sorry). Also, the 5W1H element “WHEN” and the insertion position “sentence” to be replaced with the adverbial usage (apology) are calculated.

  The adverbial usage replacement unit 32 then converts the 5W1H element of the document data into the adverb based on the adverbial usage by the adverbial usage calculation unit 31 and the calculation result of the 5W1H element and the insertion position of the adverbial usage. After replacing with the usage, it is inserted at the insertion position of the adverb usage in the document data. In the document data B4 of the present embodiment, as shown in the document data B5 in FIG. 5, the “Waiting for 15 days” of the element “WHEN” of the 5W1H is used as an adverbial usage (a long-awaited) or “adverbial” Replaced with usage (sorry) “regret but” and insert it at the beginning of the sentence.

  After that, the document output unit 40 is that the 5W1H element “WHEN” “Last 15 days” is an adverbic usage (satisfied) “waiting” or an adverbic usage (sorry) “regret, The document data A5 shown in FIG. 5 is output in a predetermined format.

  In this embodiment, the case where the 5W1H element of the document data is replaced with the adverbial usage based on the adverbial usage calculation standard regarding the counter word will be described with reference to FIGS.

  As shown in FIG. 9, the adverbial usage DB 50 includes adverbial usage calculation standards for adversaries for calculating adverbial usage, adverbial usages corresponding to the adverbial usage calculation criteria, and adverbial usages. The 5W1H element to be replaced and the insertion position of the adverbial usage in the document data are stored while being associated with each other.

  The adverbial usage criteria for this adversary correspond to groups of opposite words that have opposite meanings, such as “warm” and “cold”, “high” and “low”, “early” and “slow”. Both are associated so that identifications such as “warmth / warm”, “high / low”, “late”, etc. correspond to the word groups relating to the respective words. Further, as shown in FIG. 10, the word group related to this counter word is set so that one or a plurality of words belong to each word group. In this embodiment, the adverbial usage calculation standard corresponding to the word group related to the word is used, but the adverbial usage calculation standard corresponding to the word related to the word may be used.

  Thus, when the document data C1 shown in FIG. 8 is input to the document input unit 10, in the natural language processing unit 20, the morpheme analysis unit 21 performs a predetermined morpheme analysis, and the syntax analysis unit 22 performs a predetermined syntax analysis. Then, the 5W1H determination unit 23 determines 5W1H as shown in the document data C2 in FIG.

  Next, in the scenario generation processing unit 30, the adverbial usage calculation unit 31 refers to the adverbial usage DB 50 with respect to the document data C2 in FIG. Calculate adverbial usage. In the document data C2 of this embodiment, the word “Okinawa” to which the word group “warm” belongs and the word “snow snow” to the word group “cold” are included in 5W1H. Adverbial usage (startle) corresponding to (cold) is calculated. Also, the 5W1H element “WHEN” and the insertion position “sentence” to be replaced with the adverbial usage (amazing) are calculated.

  Then, the adverbial usage replacement unit 32 uses the adverbial usage calculation unit 31 to calculate the adverbial usage, the element of 5W1H, and the calculation result of the insertion position of the adverbial usage as the adverb. After replacing with the usage, it is inserted at the insertion position of the adverb usage in the document data. In the document data C2 of this embodiment, as shown in the document data C3 of FIG. 8, the “last year” of the 5W1H element “WHEN” is replaced with “unbelievable” of the adverbial usage (amazing). Insert at the beginning of the sentence above.

  After that, the document output unit 40 replaces the document data C3 shown in FIG. 8 in which “last year” of the element “WHEN” of 5W1H is replaced with “unbelievable” of the adverbial usage (amazing). Output in a predetermined format.

  In this embodiment, when each word group is set in advance for a predetermined word, an associated word group such as the word group “red” may be set for the word “苺”.

  In this embodiment, a case where the 5W1H element of the document data is replaced with the adverbial usage based on the adverbial usage calculation standard regarding the connection pattern will be described with reference to FIG. 11 and FIG.

  As shown in FIG. 12, the adverbial usage DB 50 includes adverbial usage calculation standards related to connection patterns for calculating adverbial usage, adverbial usages corresponding to the adverbial usage calculation criteria, and adverbial usages. The 5W1H element to be replaced and the insertion position of the adverbial usage in the document data are stored while being associated with each other.

  The adverbial usage calculation criteria for this connection pattern are as follows: connection type such as assumption, reverse connection, forward connection, etc., “If A, A”, “A, but B”, “A because B” (A and B are clauses / phrases) And a connection pattern such as a word is associated with each other so as to correspond to each other.

  Thus, when the document data D1 shown in FIG. 11 is input to the document input unit 10, the morphological analysis unit 21 performs morpheme analysis in the natural language processing unit 20 as shown in the document data D2 in FIG. After the analysis unit 22 performs syntax analysis as shown in the document data D3 in FIG. 11, the 5W1H determination unit 23 performs determination of 5W1H as shown in the document data D4 in FIG.

  Next, in the scenario generation processing unit 30, the adverbial usage calculation unit 31 refers to the adverbial usage DB 50 with respect to the document data D4 in FIG. Calculate adverbial usage. In the document data D4 of this embodiment, since the connection data (assumed) “if, if A” is included in the document data, the adverbial usage corresponding to the adverbial usage calculation standard for the connection pattern (assumed) ( Assumption). Also, the 5W1H element “WHEN” and the insertion position “sentence” to be replaced with the adverbial usage (amazing) are calculated.

  Then, the adverbial usage replacement unit 32 uses the adverbial usage calculation unit 31 to calculate the adverbial usage, the element of 5W1H, and the calculation result of the insertion position of the adverbial usage as the adverb. After replacing with the usage, it is inserted at the insertion position of the adverb usage in the document data. In the document data D4 of the present embodiment, as shown in the document data D5 of FIG. 11, “December 8, 1941” of the element “WHEN” of 5W1H may be “unbelievable”. Insert it at the beginning of the sentence after replacing "". At this time, the word “if” related to the connection pattern may be omitted.

  After that, in the document output unit 40, the element “WHEN” of 5W1H “December 8, 1941” is replaced with the adverbial usage (assuming may be incredible). 11 is output in a predetermined format.

  In the present embodiment, a case where the 5W1H element of the document data is replaced with an adverbial usage based on the adverbial usage calculation standard regarding the exaggerated expression will be described with reference to FIGS. 13 and 14.

  As shown in FIG. 14, the adverbial usage DB 50 includes adverbial usage calculation criteria for exaggerated expressions for calculating adverbial usage, adverbial usage corresponding to the adverbial usage calculation criteria, and adverbial usage. The 5W1H element to be replaced and the insertion position of the adverbial usage in the document data are stored while being associated with each other.

  The adverbial usage calculation standard regarding this exaggeration corresponds to the exaggerated expressions such as simple, unique, absolute etc. and exaggerated patterns such as “easy to HOW”, “one in the world before WHAT”, “absolutely HOW” Associated to be.

  Thus, when the document data E1 shown in FIG. 13 is input to the document input unit 10, in the natural language processing unit 20, the morpheme analysis unit 21 performs morpheme analysis as shown in the document data E2 in FIG. After the analysis unit 22 performs syntax analysis as shown in the document data E3 in FIG. 13, the 5W1H determination unit 23 performs determination of 5W1H as shown in the document data E4 in FIG.

  Next, in the scenario generation processing unit 30, the adverbial usage calculation unit 31 refers to the adverbial usage DB 50 for the document data E4 in FIG. Calculate the usage. In the document data E4 of the present embodiment, the exaggeration of “easy” is included in the HOW, so the adverbial usage calculation standard (simple) relating to the exaggerated expression (simple) adverbial usage corresponding to “easy to HOW” (amazing) "Unbelievable" is calculated. Also, the 5W1H element “WHEN” and the insertion position “sentence” to be replaced with the adverbial usage (amazing) are calculated.

  Then, the adverbial usage replacement unit 32 uses the adverbial usage calculation unit 31 to calculate the adverbial usage, the element of 5W1H, and the calculation result of the insertion position of the adverbial usage as the adverb. After replacing with the usage, it is inserted at the insertion position of the adverb usage in the document data. In the document data E4 of the present embodiment, as shown in the document data E5 of FIG. 13, “everyday” of the element “WHEN” of 5W1H is replaced with “unbelievable” of the adverbial usage (amazing). Insert at the beginning of the sentence. At this time, the particle “mo” may be added to the end of “rebounder” of the element “WHO” of 5W1H.

  After that, the document output unit 40 predetermines the document data E5 shown in FIG. 13 in which “everyday” of the element “WHEN” of 5W1H is replaced with “unbelievable” of the adverbial usage (amazing). Output in the form

  In the present embodiment, when the adverbial usage calculation unit 31 calculates an adverbial usage corresponding to a predetermined adverbial usage calculation standard, the replacement target and insertion position of the adverbial usage are also calculated. However, they may not be calculated together. In this case, it is conceivable that the 5W1H element to be replaced by the adverb usage is preset as, for example, “WHEN” only, or the insertion position is preset as the position of the original 5W1H element.

  Further, the case where only one 5W1H element of the document data is replaced with the adverbial usage has been described. However, a plurality of 5W1H elements may be replaced with the adverbial usage corresponding to each.

  Further, the adverbial usage calculation standard and the adverbial usage are not limited to those in the first to fifth embodiments, but may be other adverbial usage calculation standards and adverbial usage.

  As mentioned above, although embodiment of this invention was described with reference to drawings, this invention is not limited to the thing of embodiment shown in figure. Various modifications and variations can be made to the illustrated embodiment within the same range or equivalent range as the present invention.

DESCRIPTION OF SYMBOLS 10 ... Document input part 20 ... Natural language processing part 21 ... Morphological analysis part 22 ... Syntax analysis part 23 ... 5W1H determination part 30 ... Scenario generation part 31 ... Adverbial usage Calculation unit 32 ... adverb usage replacement unit 40 ... document output unit 50 ... adverb usage DB

Claims (7)

A document input unit for inputting document data;
By analyzing the document data input by the document input unit, at least one of what, who, what, where, why, and how A natural language processing unit for determining a plurality of 5W1H elements;
An adverbial usage database for storing adverbial usage calculation criteria for calculating adverbial usage and adverbial usages corresponding to the adverbial usage calculation criteria in association with each other;
Adverbial usage calculation for calculating adverbial usage corresponding to a predetermined adverbial usage calculation standard by referring to the adverbial usage database for the document data for which the 5W1H element has been determined by the natural language processing unit And
An adverbial usage replacement unit that replaces a predetermined 5W1H element of the document data with an adverbial usage calculated by the adverbial usage calculation unit;
A natural language processing apparatus comprising: a document output unit that outputs new document data in which predetermined 5W1H elements are replaced by adverbial usage by the adverbial usage replacement unit. The adverbial usage database stores 5W1H elements to be substituted for adverbial usage while being associated with adverbial usage calculation criteria,
The adverbial usage calculation unit refers to the adverbial usage database for the document data for which the 5W1H element has been determined by the natural language processing unit, so that the adverbial usage calculation unit corresponds to a predetermined adverbial usage calculation standard. Calculate the usage and the 5W1H element to be replaced by the adverbial usage,
The adverbial usage replacement unit replaces the 5W1H element of the document data with the adverbial usage based on the adverbial usage calculated by the adverbial usage calculation unit and the calculation result of the 5W1H element. The natural language processing device described in 1. In the adverbial usage database, the insertion position of the adverbial usage in the document data is stored while being associated with the adverbial usage calculation criterion,
The adverbial usage calculation unit refers to the adverbial usage database for the document data for which the 5W1H element has been determined by the natural language processing unit, so that the adverbial usage calculation unit corresponds to a predetermined adverbial usage calculation standard. Calculating the usage, the 5W1H element to be replaced by the adverbial usage, and the insertion position of the adverbial usage in the document data;
The adverbial usage replacement unit replaces the 5W1H element of the document data with the adverbial usage calculated by the adverbial usage calculation unit, the 5W1H element, and the calculation result of the insertion position of the adverbial usage. The natural language processing apparatus according to claim 1, wherein the natural language processing device is inserted into the insertion position of the adverbial usage in the document data after being replaced with the adverbial usage. The adverbial usage database is stored with criteria relating to the user's personality being associated with adverbial usage criteria,
The adverbial usage calculation unit refers to the adverbial usage database with respect to the document data for which the 5W1H element has been determined by the natural language processing unit, so that a predetermined adverb The natural language processing apparatus according to any one of claims 1 to 3, wherein an adverbial usage corresponding to a standard usage calculation standard is calculated. The adverbial usage calculation standard relates to the opposite word,
The adverbial usage calculation unit corresponds to the adverbial usage calculation standard for a predetermined opposite word by referring to the adverbial usage database for the document data for which the 5W1H element has been determined by the natural language processing unit. The natural language processing apparatus according to any one of claims 1 to 4, which calculates adverbial usage. The adverbial usage calculation criteria relate to connection patterns;
The adverbial usage calculation unit corresponds to the adverbial usage calculation standard for a predetermined connection pattern by referring to the adverbial usage database for the document data for which the natural language processing unit has determined the 5W1H element. The natural language processing apparatus according to claim 1, wherein the adverbial usage is calculated. The adverbial usage calculation standard relates to exaggerated expressions,
The adverbial usage calculation unit corresponds to the adverbial usage calculation standard for a predetermined exaggerated expression by referring to the adverbial usage database for the document data for which the natural language processing unit has determined the 5W1H element. The natural language processing apparatus according to claim 1, wherein the adverbial usage is calculated.