JP2017054434A - Contradiction creation device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To create a new contradiction situation pair.SOLUTION: A relation extraction unit 20 extracts a relation of a noun pair for contradiction by checking a contradiction situation pair database 18 and a noun relation database 16 and creates a relation list in which the relation of the extracted noun pair is stored. Regarding a contradiction situation pair obtained from the contradiction situation pair database 18, if the relation of a noun pair included in the contradiction situation pair is included in the relation list 22, a noun pair replacement unit 28 acquires a noun pair having the same relation with the relation of the noun pair from the noun relation database 16 and creates a contradiction situation pair by replacing the noun pair of the contradiction situation pair with the acquired noun pair. A single noun replacement unit 30 extracts a coordinate term of a shared noun that the acquired contradiction situation pair has from the noun relation database 16 and creates the contradiction situation pair by replacing the shared noun of the contradiction situation pair with the acquired coordinate term.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a contradiction creation device, method, and program, and more particularly, to a contradiction creation device, method, and program for creating a new contradiction situation pair.

  In the field of language generation in natural language processing, it is important to generate sentences with consistent contents. However, it is not easy to determine whether a computer-generated sentence has consistent contents because it requires common sense knowledge.

  For example, inconsistency may be included in the generated sentence as a case where consistency is lacking. For example, the sentence “I have never been to Paris but climbed the Eiffel Tower” contains a contradiction. This is because “I have never been to Paris” and “I climbed the Eiffel Tower” are contradictory.

  In order to avoid the generation of a sentence containing such a contradiction, it is necessary to check using knowledge of the contradiction.

  For example, if the computer has the knowledge that "I have never been to Paris" and "I climbed the Eiffel Tower" are contradictory, a sentence like the previous example will not be generated. .

  As a methodology for acquiring such contradiction knowledge, there are a method of automatically acquiring from a corpus (see Non-Patent Document 1) and a method of using crowdsourcing (a method for requesting data creation from a user on the Internet) (Non-Patent Document 1). 2).

Chikara Hashimoto, Kentaro Torisawa, Stijn De Saeger, Jong-Hoon Oh, and Jun'ichi Kazama.Excitatory or inhibitory: A new semantic orientation extracts contradiction and causality from the web.In Proceedings of EMNLP2012, pp. 619-630, 2012. Takamine, Hajime Morita, Daisuke Kawahara, Ikuo Kurohashi, Ryuichiro Higashinaka, Yoshihiro Matsuo. Analysis and classification of conflicts between situations using crowdsourcing, 21st Annual Meeting of the Language Processing Society, pp.305-308,2015 .

  Even if the contradiction is automatically extracted from the corpus or the crowdsourcing is used, the amount of contradiction knowledge that can be acquired is limited. Common sense knowledge (including contradiction knowledge) provided by humans is considered enormous, and increasing the amount of contradiction knowledge is an important issue.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a contradiction creation apparatus, method, and program capable of creating a new contradiction pair.

  In order to achieve the above object, the contradiction creation device according to the first invention includes a contradiction pair database storing a contradiction pair, which is a pair of sentences representing contradictory situations, and a relationship between the noun pair and the noun pair. By comparing the stored noun relationship database, the relationship between the noun pairs related to the contradiction is extracted, the relationship extraction unit that creates the relationship list storing the relationship of the extracted noun pairs, and the contradiction situation pair database obtained above For the contradiction pair, when the relationship of the noun pair included in the contradiction pair is included in the relationship list, the noun pair having the same relationship as the relationship of the noun pair is acquired from the noun relationship database, and the acquired A noun pair replacement unit that creates a contradiction pair by replacing the noun pair of the contradiction pair by a noun pair. It has been.

  A contradiction creation method according to a second invention is a contradiction creation method in a contradiction creation device including a relationship extraction unit and a noun pair replacement unit, wherein the relationship extraction unit is a pair of sentences representing a conflicting situation. By comparing the contradiction pair database storing pairs with the noun pair and the noun relationship database storing the relationship between the noun pairs, the relationship between the noun pairs related to the contradiction is extracted, and the relationship between the extracted noun pairs is stored. Creating a relationship list, and when the noun pair replacement unit includes the contradiction pair obtained from the contradiction pair database, the noun pair included in the contradiction pair is included in the relationship list, the noun A noun pair having the same relationship as the pair relationship is acquired from the noun relationship database, and the name of the contradiction pair is acquired by the acquired noun pair. To create a conflict situation pair by replacing the pair.

  According to 1st invention and 2nd invention, the relationship of the noun pair regarding a contradiction is extracted and the relationship list is created, and the relationship of the noun pair included in the contradiction pair obtained from the database is included in the relationship list. In this case, a new contradiction pair can be created by acquiring a noun pair having the same relationship as that of the noun pair from the noun relation database and replacing the noun pair of the contradiction pair with the acquired noun pair.

  The contradiction creating apparatus according to the third invention is configured to obtain a contradiction pair having a common noun common to the antecedent and the consequent from the contradiction pair database storing the contradiction pair that is a pair of sentences representing the contradictory situation. Extracting a noun pair and a noun pair and a noun relation database storing relations of the noun pair, and extracting a common noun synonym of the contradiction pair acquired by the contradiction pair acquiring unit And a single noun replacement unit that creates a contradiction pair by replacing the common noun of the contradiction pair with the same isotope.

  A contradiction creation method according to a fourth invention is a contradiction creation method in a contradiction creation device including a contradiction pair acquisition unit and a single noun substitution unit, wherein the contradiction pair acquisition unit A contradiction pair having a common noun common to the antecedent and the consequent is obtained from the contradiction pair database storing the contradiction pair that is a pair, and the single noun replacement unit includes the noun pair and the noun pair. From the noun relation database storing the relationship, extracting the common noun isotopes of the contradiction pair acquired by the contradiction pair acquisition unit, replacing the common noun of the contradiction pair with the extracted isotope, Create a contradiction pair.

  According to the third invention and the fourth invention, a contradiction pair having a common noun is acquired from the contradiction pair database, and a common noun synonym of the acquired contradiction pair is extracted from the noun relation database. A new contradiction pair can be created by replacing the common noun of the contradiction pair with the acquired isotope.

  A program according to a fifth aspect of the invention is a program for causing a computer to function as each part constituting the contradiction creation device.

  As described above, according to the contradiction creation apparatus, method, and program of the present invention, an effect that a new contradiction situation pair can be created is obtained.

It is a block diagram which shows the structure of the contradiction creation apparatus which concerns on embodiment of this invention. It is a flowchart which shows the content of the contradiction creation process routine in embodiment of this invention. It is a figure which shows an experimental result.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment of the present invention, a case where the contradiction pair included in the contradiction pair database is expanded will be described as an example.

<Outline of the present embodiment>
When a contradiction pair database exists, a new contradiction pair is obtained by extending the contradiction pair in the database using a noun relation database obtained from the knowledge base.

  Here, the conflict pair database is a list of conflict pairs. The contradictory situation pair is a sentence representing two contradictory situations, and the sentence representing each situation is represented by a predicate term structure. The predicate term structure is composed of one predicate and one or more terms (noun phrases). An example of a contradiction pair is shown below.

<Go to France, go to the Eiffel Tower>
<Do not schedule trips, apply for overseas tours>

  The noun relation database indicates two nouns and the relation between the two nouns, and is constructed from an existing knowledge base.

  The existing knowledge base is, for example, Freebase (registered trademark) (https://www.freebase.com/). Freebase (registered trademark) stores a large amount of data that expresses two things and their relationships in triplicate. Freebase (registered trademark) includes the following as nouns and their relations, and obtains them as a noun relation database.

(Eiffel Tower, contained-by, France)
(Hakone, contained-by, Kanagawa)

  Here, contained-by means “included”, meaning that the Eiffel Tower is included in France. In the second example, Hakone is included in Kanagawa Prefecture. A contradiction pair can be newly created using such a noun relation database representing the two nouns and their relations. For example, the contradictory pair <Go to France, Go to the Eiffel Tower> includes the terms France and the Eiffel Tower respectively. And these are contained-by according to the noun relation database. Therefore, assuming that it is replaced by Hakone and Kanagawa Prefecture, which are also in a contained-by relationship, a new contradiction pair can be created: <Do not go to Kanagawa Prefecture, go to Hakone>. In addition, although the example which paid attention only to the contained-by relationship was shown here, you may use the other relationship contained in a noun relationship database.

  Since the content included in the newly created contradiction pair is automatically created, it may contain noise such as grammar and semantic errors. Therefore, the appropriateness of the content is checked depending on whether or not the content included in the contradictory state pair is present more than a threshold value in the large text corpus.

<System configuration>
A contradiction creation device 10 according to an embodiment of the present invention includes a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Read) that stores a contradiction creation program for executing a contradiction creation processing routine described later. Only memory). The CPU functions as the contradiction creation device 10 when the CPU reads and executes the contradiction creation program from the ROM which is an internal storage device.

  As shown in FIG. 1, the contradiction creation device 10 includes a knowledge base group 12, a noun relationship acquisition unit 14, a noun relationship database 16, a contradiction situation pair database 18, a relationship extraction unit 20, a contradiction situation pair acquisition unit 24, a contradiction situation pair. A creation unit 26, a check unit 32, a large-scale text corpus 34, and a database addition unit 36 are provided.

  The knowledge base group 12 includes a first knowledge base 12A, a second knowledge base 12B, and a third knowledge base 12C.

  In the present embodiment, Freebase (registered trademark) is used as the first knowledge base 12A, Wikipedia (registered trademark) is used as the second knowledge base 12B, and a Japanese dictionary is used as the third knowledge base 12C. Any knowledge base other than these knowledge bases can be used as long as it can obtain nouns and their relationships. For example, NTT (registered trademark) Japanese vocabulary system may be used.

  The noun relationship acquisition unit 14 acquires the noun pair and the relationship between the noun pair from the first knowledge base 12 </ b> A to the third knowledge base 12 </ b> C and stores them in the noun relationship database 16.

  Freebase (registered trademark) basically stores two nouns and their relationships as triples, so all these triples can be extracted. For Wikipedia (registered trademark) and Japanese language dictionaries, it is possible to obtain nouns that are in a higher-level relationship based on dictionary headings and their definition sentences. For example, the following methods of Non-Patent Document 3 and Non-Patent Document 4 may be used. For example, in the Japanese dictionary, if the headline “Tricycle” is defined as “One of the vehicles”, it is understood that the broader term of the tricycle is a vehicle, and “Tricycle” and “Vehicle” are in a lower-level relationship. Noun relations can be acquired.

[Non-Patent Document 3]: Asuka Sumida, Naoki Yoshinaga, Kentaro Torizawa, Extracting upper and lower relations from Wikipedia article structure, natural language processing, vol.16 (3), pp.3-24, 2008.

[Non-patent document 4]: Yasuki Daifuku, Daisuke Kawahara, Ikuo Kurohashi, Automatic construction of a thesaurus by integrated use of large-scale corpus and Japanese dictionary, 10th Annual Conference of the Language Processing Society, 2004.

  Examples of data included in the noun relation database 16 extracted from Freebase (registered trademark), Wikipedia (registered trademark), and Japanese dictionary are as follows.

(Eiffel Tower, contained-by, France)
(Hakone, contained-by, Kanagawa)
(Golf, sports-equipment, golf club)
(Baseball, sports-equipment, bat)
(Overseas tour, is-a, travel)
(Domestic tour, is-a, travel)
(TOEIC test, is-a, test)
(Tennis, is-a, ball game)
(Golf, is-a, ball game)

  The first four are from Freebase (R). As mentioned above, contained-by represents the relationship “contained in” and sports-equipment represents the relationship between the sport and the tool used in that sport. For example, “golf” and “golf club” have a sports-equipment relationship. The following is-a represents the broader relation (the left side is the narrower term and the right side is the broader term) For example, an overseas tour is a subordinate term for travel.

  In addition, a lower word set having an is-a relationship with the same broader word (a relationship between the lower words is referred to as a “same word”) is also registered in the noun relation database 16. For example, “overseas tour”, “domestic tour”, etc. for “travel”, and “tennis”, “golf”, etc. for “ball games” are obtained. The synonyms are stored in the noun relation database 16 as follows.

(Overseas tour, same language, domestic tour)
(Tennis, isotope, golf)

  The contradiction situation pair database 18 stores a plurality of contradiction situation pairs, which are pairs of sentences representing contradictory situations.

  The contradiction situation pair database 18 may be created manually or may be constructed by the method of Non-Patent Document 1 or Non-Patent Document 2.

  The relationship extraction unit 20 collates the contradiction pair database 18 and the noun relationship database 16, extracts the relationship between the noun pairs related to the contradiction that appears in the contradiction pair database 18 a number of times greater than or equal to the threshold, and determines the relationship between the extracted noun pairs. The stored relation list 22 is created.

  Specifically, first, a noun pair is extracted from the term part in each contradiction (predicate term structure) pair stored in the contradiction pair database 18. Here, each noun of a noun pair is extracted from a different predicate term structure, and the noun includes a compound noun (a sequence of two or more nouns).

  Next, it is checked whether or not these noun pairs exist in the noun relation database 16, and if they exist, the relation in which they exist is searched and tabulated.

  For example, the above-mentioned contradiction pair database 18 extracts the noun pairs (France, Eiffel Tower), (Travel, Overseas Tour), (Schedule, Overseas Tour), of which (France, Eiffel Tower), (Travel, It can be seen that there are two types of overseas tours) in the noun relation database 16. Therefore, the relationship between contained-by and is-a is counted up once. As described above, the noun relation database 16 is collated with respect to the noun pairs in all the contradiction pair included in the contradiction pair database 18 to count which relations can be collated.

  The matching is performed using the representative notation output from the morphological analyzer JUMAN in order to deal with fluctuations in the notation. Of course, if there is a dictionary including fluctuation information, it is used.

  Finally, check whether the number of occurrences of the found noun pair relationship exceeds a specified threshold value (for example, 10), and if so, the noun pair relationship is considered an effective relationship for creating a contradiction. Add to the relationship list 22. The process here can be said to be a process for examining what relationship nouns are likely to appear in a contradiction pair.

  The contradiction event pair acquisition unit 24 extracts one or more contradiction event pairs having a common noun common to the term part in the antecedent and consequent from the prepared contradiction event pair database 18 and replaces a single noun replacement described later. Send to part 30.

  In addition, the contradiction event pair acquisition unit 24 extracts one or more contradiction event pairs having no shared noun from the previously prepared contradiction event pair database 18 and sends them to the noun pair replacement unit 28 described later.

  The contradictory situation pairs extracted here are processed one by one by the noun pair replacement unit 28 and the single noun replacement unit 30 in the subsequent stage.

  The following is an example of a contradiction pair that is included in the contradiction pair database 18 and does not have a common noun.

<Go to France, go to the Eiffel Tower>
<Do not schedule trips, apply for overseas tours>

  The contradiction situation pair creation unit 26 includes a noun pair replacement unit 28 and a single noun replacement unit 30.

  When the relationship list 22 includes the relationship of the noun pairs included in the contradiction pair for each of the contradiction pairs having no shared noun obtained from the contradiction pair database 18, the noun pair replacement unit 28 includes: A noun pair having the same relationship as the noun pair relationship is acquired from the noun relationship database 16, and the contradiction pair is created by replacing the noun pair of the contradiction pair with the acquired noun pair.

  For example, if <not go to France, go to the Eiffel Tower> is a processing target, first, a noun pair is extracted from here. The noun pair is (France, Eiffel Tower). Then, the noun relation database 16 is referred to look for the relation between France and the Eiffel Tower. You can see that it is a contained-by relationship. Since this relationship is included in the relationship list 22 obtained by the relationship extraction unit 20, another noun pair having the same relationship is acquired from the noun relationship database 16 and replaced with the noun pair of the contradiction pair. Create a contradiction pair. For example, from this example, the following new contradiction pair is generated:

<Don't go to Kanagawa Prefecture, go to Hakone>

  Here, the relationship between Kanagawa Prefecture and Hakone is contained-by, so we have replaced France with Kanagawa and the Eiffel Tower with Hakone. By replacing in this way, a new contradiction pair can be created.

  The newly created contradiction pair is sent to the check unit 32.

  The single noun replacement unit 30 extracts and obtains, from each noun relation database 16, the common noun synonym of the contradiction pair for each contradiction pair having the common noun obtained from the contradiction pair database 18. The contradiction pair is created by replacing the common noun of the contradiction pair with the same isotope.

  For example, consider the following inconsistency pair:

<Do not check the zoo location, go to the zoo>
<I'm not interested in tennis, I want to play tennis>

  From here, the single noun replacement unit 30 creates the following new contradiction pair.

<Do not check the location of the botanical garden, go to the botanical garden>
<I'm not interested in golf, I want to play golf>

  Here, “zoo” and “botanical garden” are synonymous terms, and “tennis” and “golf” are both synonymous terms with “ball game” in the national language dictionary.

  The newly created contradiction pair is sent to the check unit 32.

  For each newly created contradiction pair, for each sentence representing a situation included in the contradiction pair, the check unit 32 has a frequency of the sentence representing the situation in the large-scale text corpus 34 equal to or greater than a threshold value. If the frequency of each sentence representing a situation included in the contradiction pair is equal to or greater than a threshold, it is determined that the contradiction pair is appropriate, and is sent to the database adding unit 36.

  Specifically, the number of occurrences in the large text corpus 34 is independently checked for each of the antecedent (left sentence) and the subsequent (right sentence) of the newly created contradiction pair, If it appears more than 10 times), it is judged that this contradiction pair is appropriate and adopted as a new contradiction pair. Appearance is checked by searching for phrases in a predicate term structure expression representing the situation using a general Web search engine or a search index created from large-scale text. The search may be performed by normalizing the predicate term structure expression to the JUMAN representative notation in order to absorb the fluctuation of the notation.

  Note that the processing step of the check unit 32 may be skipped based on the balance between the amount and accuracy of the contradiction pair to be acquired. In that case, the contradiction pair newly created by the contradiction pair creation unit 26 is sent to the database addition unit 36 as it is.

  The database adding unit 36 adds the contradiction pair sent from the check unit 32 to the contradiction pair database 18.

<Operation of contradiction creation device>
Next, the operation of the contradiction creation device 10 according to the present embodiment will be described. In the contradiction creation apparatus 10, the contradiction creation processing routine shown in FIG. 2 is executed.

  First, in step S100, each combination of a noun pair and a relationship between the noun pair is acquired from the first knowledge base 12A to the third knowledge base 12C and stored in the noun relation database 16.

  In step S102, the contradiction pair database 18 and the noun relation database 16 are collated, the relation of the noun pairs related to the contradiction that appears in the contradiction pair database 18 more than a threshold value is extracted, and the relationship of the extracted noun pairs is determined. The stored relation list 22 is created.

  In step S104, a contradiction pair having no shared noun is acquired from the contradiction pair database 18. In step S106, the relationship between the noun pairs included in the contradiction pair obtained in step S104 and having no shared noun is the same as the relationship between the noun pairs when the relationship list 22 includes the relationship. The noun pair having “” is acquired from the noun relation database 16.

  In step S108, a contradiction pair is created by replacing the noun pair of the contradiction pair acquired in step S104 with the noun pair acquired in step S106.

  In step S110, for each sentence representing the situation included in the contradiction pair created in step S108, it is checked whether the frequency of the sentence representing the situation in the large-scale text corpus 34 is equal to or higher than a threshold value. It is determined whether the frequency of each sentence representing a situation included in the situation pair is equal to or higher than a threshold value. When any frequency of the sentence representing the situation included in the contradiction pair is equal to or higher than the threshold value, it is determined that the created contradiction pair is appropriate, and the process proceeds to step S112.

  On the other hand, if the frequency of at least one of the sentences representing the situations included in the contradiction pair is less than the threshold, it is determined that the created contradiction pair is not appropriate, and the process proceeds to step S114.

  In step S112, the contradiction pair created in step S108 is added to the contradiction pair database 18.

  In step S112, it is determined whether or not the processing of steps S104 to S112 has been executed for all of the contradiction pairs having no shared noun stored in the contradiction pair database 18, and the steps S104 to S112 are performed. If there is a contradiction pair that does not execute the process and does not have a shared noun, the process returns to step S104 to acquire the contradiction pair.

  On the other hand, when the processes of steps S104 to S112 are executed for all of the contradiction pair having no shared noun stored in the contradiction pair database 18, the process proceeds to step S116.

  In step S116, a contradiction pair having a common noun is acquired from the contradiction pair database 18. In step S118, the noun relation database 16 acquires the common noun synonyms of the contradiction pair acquired in step S116.

  In step S120, a contradiction pair is created by replacing the common noun of the contradiction pair acquired in step S116 with the isotope acquired in step S118.

  In step S122, for each sentence representing the situation included in the contradiction situation pair created in step S120, it is checked whether the frequency of the sentence representing the situation in the large-scale text corpus 34 is equal to or higher than a threshold value. It is determined whether the frequency of each sentence representing a situation included in the situation pair is equal to or higher than a threshold value. When any frequency of the sentence representing the situation included in the contradiction situation pair is equal to or higher than the threshold, it is determined that the created contradiction pair is appropriate, and the process proceeds to step S124.

  On the other hand, if the frequency of at least one of the sentences representing the situations included in the contradiction pair is less than the threshold, it is determined that the created contradiction pair is not appropriate, and the process proceeds to step S126.

  In step S124, the contradiction pair created in step S120 is added to the contradiction pair database 18.

  In step S126, it is determined whether or not the processing of steps S116 to S124 has been executed for all of the conflicting pairs having shared nouns stored in the conflicting pair database 18, and the processing of steps S116 to S124 is performed. If there is a contradiction pair having a common noun that has not been executed, the process returns to step S116 to obtain the contradiction pair.

  On the other hand, when the processes in steps S116 to S124 are executed for all the contradictory pairs having a common noun stored in the contradictory pair database 18, the contradiction creating process routine is terminated.

<Expansion experiment of contradiction pair database>
An example of a contradiction pair that can be acquired as a result of executing the technique based on the embodiment of the present invention is as shown in FIG. All of FIG. 3 passed through the check unit 32, and they were all correct contradiction pairs.

  Further, as an example excluded by the check unit 32, there is <I do not like city, I am addicted to a government-designated city>. This is generated by the relationship of (decree-designated city, is-a, city) from <I do not like sushi, addicted to conveyor belt sushi>, but "added to the government-designated city" appears in the large text corpus 34 Since the number of times was small, it was excluded by the check unit 32 and was not acquired correctly.

  As described above, according to the contradiction creating apparatus according to the present embodiment, the noun pair included in the contradiction pair obtained from the contradiction pair is created by extracting the relationship between the noun pairs related to the contradiction and creating the relation list. If the relationship of a pair is included in the relationship list, a new inconsistency can be obtained by acquiring a noun pair having the same relationship as the relationship of the noun pair from the noun relationship database and replacing the noun pair of the inconsistency pair with the acquired noun pair. Pairs can be created.

  In addition, a contradiction pair having a common noun is obtained from the contradiction pair database, a synonym of the common noun possessed by the obtained contradiction pair is extracted from the noun relation database, and the contradiction pair is shared by the obtained isotope. By replacing nouns, you can create new contradictory pairs.

  Further, by expanding the contradiction situation pair database, in the language generation system, generation of utterances without contradiction is realized, and sentences that are easy to understand for the user are generated by the computer. As a result, human intellectual activities can be promoted.

  The present invention is not limited to the above embodiment, and various modifications and applications can be made without departing from the gist of the present invention.

  For example, in the present specification, the embodiment in which the program is installed in advance has been described. However, the program stored in an external storage device or recording medium is read as needed, and downloaded and executed via the Internet. You may make it do. In addition, the program can be provided by being stored in a computer-readable recording medium.

DESCRIPTION OF SYMBOLS 10 Contradiction creation apparatus 12 Knowledge base group 14 Noun relation acquisition part 16 Noun relation database 18 Contradiction pair database 20 Relation extraction part 22 Relation list 24 Contradiction pair acquisition part 26 Contradiction pair creation part 28 Noun pair replacement part 30 Single noun Replacement unit 32 Check unit 34 Large text corpus 36 Database addition unit

Claims (7)

By comparing the contradiction pair database storing the contradiction pair, which is a pair of sentences representing the contradictory situation, and the noun relation database storing the relationship between the noun pair and the noun pair, the relationship of the noun pair related to the contradiction is obtained. A relationship extraction unit that extracts and creates a relationship list that stores relationships of extracted noun pairs;
For the contradiction pair obtained from the contradiction pair database, when the relationship of the noun pair included in the contradiction pair is included in the relationship list, a noun pair having the same relationship as the relationship of the noun pair is selected as the noun. A noun pair replacement unit that acquires from a relational database and creates a contradiction pair by replacing the noun pair of the contradiction pair by the acquired noun pair;
Contradiction creation device including A contradiction pair acquisition unit that acquires a contradiction pair having a common noun in the antecedent and the consequent from the contradiction pair database storing a contradiction pair that is a pair of sentences representing a contradictory situation;
From the noun relation database storing the relationship between the noun pair and the noun pair, a common noun synonym of the contradiction pair acquired by the contradiction pair acquiring unit is extracted, and the contradiction pair A single noun replacement that creates a contradiction pair by replacing a shared noun,
Contradiction creation device including   For each sentence representing the situation included in the created contradiction pair, it is checked whether the frequency of the sentence representing the situation in the large-scale text corpus is equal to or higher than a threshold value. The contradiction creation device according to claim 1 or 2, further comprising a check unit that determines that the contradiction pair is appropriate if the frequency of each of the included sentences is equal to or greater than a threshold value. A contradiction creation method in a contradiction creation device including a relationship extraction unit and a noun pair replacement unit,
By comparing the contradiction pair database storing the contradiction pair that is a pair of sentences representing the contradictory situation and the noun relation database storing the relationship between the noun pair and the noun pair, To extract the relationship of the noun pairs for and create a relationship list that stores the relationships of the extracted noun pairs,
The noun pair replacement unit, for the contradiction pair obtained from the contradiction pair database, if the relationship of the noun pair included in the contradiction pair is included in the relationship list, the same relationship as the relationship of the noun pair A contradiction creating method of creating a contradiction pair by obtaining a noun pair having a noun pair from the noun relation database and replacing the noun pair of the contradiction pair with the acquired noun pair. A contradiction creation method in a contradiction creation device including a contradiction situation pair acquisition unit and a single noun substitution unit,
The contradiction pair acquisition unit acquires a contradiction pair having a common noun common to the antecedent and the consequent from the contradiction pair database storing the contradiction pair that is a pair of sentences representing the contradictory situation,
The single noun replacement unit extracts the noun pairs and the noun relation database that stores the relationship between the noun pairs, and extracts the same nouns of the common nouns possessed by the contradiction pair acquired by the contradiction pair acquisition unit. A contradiction creation method for creating a contradiction pair by replacing a common noun of the contradiction pair with a synonym.   The check unit checks whether the frequency of the sentence representing the situation in the large-scale text corpus is greater than or equal to a threshold for each sentence representing the situation included in the created contradiction pair. The contradiction creation method according to claim 4 or 5, further comprising: determining that the contradiction pair is appropriate if the frequency of each sentence representing the situation included in the contradiction pair is equal to or greater than a threshold.   The program for functioning a computer as each part which comprises the contradiction creation apparatus of any one of Claims 1-3.