JP2012234106A - Automatic question creating device and creating method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically create multiple-choice questions in high level of completion in wide areas.SOLUTION: An automatic question creating device is configured to automatically create fill-in-the-blank type multiple-choice questions using item collection database in which data related to multiple items is stored. The automatic question creating device acquires respective pieces of information on a title of items, an explanation sentence, and a category from a predetermined medium to create the item collection database; tags predetermined words in a randomly set question sentence; sets one of tagged words as a right answer; extracts wrong answer candidates for the right answer from words that are titles of items by referring to the item collection database; and sets at least one of the extracted wrong answer candidates as a wrong answer.

Description

  The present invention relates to an automatic problem creation apparatus and a creation method for automatically creating various multiple-choice questions that ask knowledge in an arbitrary field such as history, geography, current affairs, fashion, law, and language.

  Conventionally, as one of the methods for testing various knowledge, there is a multiple choice problem in which the respondent selects the option that the respondent thinks is correct from the multiple options corresponding to the question, and determines whether the answer is correct or incorrect. .

  Creating a multiple-choice question requires a great deal of effort for the question creator to determine an incorrect answer, and the creation burden is large. Under such circumstances, for example, Patent Document 1 discloses a problem creation device that automatically creates a multiple-choice question regarding English words using a computer. This question creation device prepares a data table that associates Japanese translation with each English word, selects one English word that should be a multiple choice question from this data table, and corresponds to the selected English word Based on the Japanese translation, the wrong answer is determined from the Japanese translation for each other English word other than the English word.

JP 2004-85734 A

  However, in the case of a multiple choice problem that asks knowledge in an arbitrary field such as history, geography, current affairs, fashion, law, and national language, for example, the technique of the problem creation device disclosed in Patent Document 1 cannot be used as it is. Have difficulty.

  The present invention has been made in view of the above problems, and automatically creates multiple-choice questions in various fields with a high degree of completeness, and the problem creator automatically saves labor for creating problems. An object is to provide an apparatus and a production method.

  The present invention that achieves the above object is an electronic computer, and uses an item database that stores the title of the item, a commentary that explains the content of the item, and a category to which the item belongs, respectively. An automatic problem creation device for automatically creating a burying-type multiple-choice question, which obtains information on the title of the item, the commentary, and the category from a predetermined medium, Create the item database to be stored in a system and store it in the item database preparation unit to be stored in the storage device of the electronic computer, the problem sentence setting unit for setting the question sentence to be questioned, and the problem sentence setting unit A tagging processing unit for tagging a predetermined word in a questioned sentence, and a correct answer setting for setting one of the words tagged by the tagging processing unit as a correct answer And an error candidate surveying unit that extracts an incorrect answer candidate for the correct answer set by the correct answer setting unit from the words that are the titles of the items in the item database. And an error answer setting unit that sets at least one of the error answer candidates extracted by the error answer candidate investigation unit as an error answer.

  A part-of-speech structure that creates a part-of-speech structure database that stores a result of analyzing a part-of-speech structure of each word that is the title of the item stored in the item database, and stores the result in a storage system. Preferably, a database preparation unit is further provided, and the erroneous answer setting unit preferably refers to the part-of-speech structure database and sets an erroneous answer candidate in which the word set as the correct answer and the part-of-speech structure match as an incorrect answer.

  In addition, it further includes a category hierarchy database preparation unit that acquires information about the vertical relationship of the categories from the predetermined medium, creates a category hierarchy database that stores the acquired information in a predetermined system, and stores the information in the storage device, It is preferable that the incorrect answer candidate investigating unit refers to the category hierarchy database and extracts, as an incorrect answer candidate, a word whose category hierarchy matches a word set as a correct answer.

  Further, it is preferable that the erroneous answer candidate investigation unit selects a category closest to the context of the question sentence, and extracts a word that is a title of an item belonging to the selected category as an incorrect answer candidate.

  Moreover, it is preferable that the tagging processing unit performs tagging on a word that is a title of an item in the item database.

  Moreover, the appearance which stores the result which investigated the appearance frequency with respect to the said commentary sentence stored in the said item database of each word used as the title of the said item stored in the said item database in a predetermined system An appearance frequency database preparation unit that creates a frequency database and stores the frequency database in the storage device; an appearance frequency survey unit that refers to the appearance frequency database and investigates the appearance frequency of the words tagged by the tagging processing unit; The correct answer setting unit preferably sets the word as a correct answer based on the appearance frequency when there are a plurality of tagged words.

  In addition, the present invention that achieves the above object uses an item database that stores a title of the item, a commentary that explains the content of the item, and a category to which the item belongs, respectively, for a plurality of items. An automatic problem creation method for automatically creating a hole-filling multiple-choice problem by a computer, wherein the central processing unit of the electronic computer obtains information on the title, commentary, and category of the item from a predetermined medium. Acquire, create the item database that stores the acquired information in a predetermined system, store it in the storage device of the electronic computer, set a question sentence to be a question, and set a predetermined word in the set question sentence Tagging, setting one of the tagged words as a correct answer, and setting a wrong answer candidate for the set correct answer as the item in the item database. Of extracts from the word that is the title it is characterized by executing a process of setting the wrong answer at least one of the extracted the wrong answer candidate.

  According to the present invention, it is possible to automatically create multiple-choice questions in various fields with a high degree of completeness, and it is possible to achieve an excellent effect that it is possible to save time for the problem creator to create a problem.

It is a block diagram which shows typically the whole hardware structure of the problem automatic preparation apparatus which concerns on one Embodiment of this invention. It is a figure which shows typically one of the items (article) in Wikipedia from which the automatic problem preparation apparatus which concerns on one Embodiment of this invention acquires information. It is a figure which shows typically a part of category hierarchy prescribed | regulated in Wikipedia from which the automatic problem preparation apparatus which concerns on one Embodiment of this invention acquires information. It is a figure which shows typically one of the category commentary pages in Wikipedia from which the automatic problem preparation apparatus which concerns on one Embodiment of this invention acquires information. It is a block diagram which shows typically the whole software structure of the problem automatic preparation apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the part-of-speech structure of the item which the problem automatic preparation apparatus concerning one Embodiment of this invention analyzed. It is a figure which shows the category hierarchy used for extraction of an incorrect answer candidate by the automatic question preparation apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the procedure (question automatic creation method) of the problem creation by the problem automatic creation apparatus which concerns on one Embodiment of this invention, and the completed problem.

  An embodiment of an automatic problem creation apparatus and creation method of the present invention will be described in detail with reference to the drawings. In this embodiment, the case where the open-type multiple-choice questions in various fields are automatically created by using “Free Encyclopedia, Wikipedia (Japanese version)” (referred to as article content), which is an open content. Will be described.

  FIG. 1 shows an overall hardware configuration of an automatic problem creation apparatus 1 according to the present embodiment. The automatic problem creation apparatus 1 is composed of an electronic computer such as a computer, a central processing unit 2 such as a CPU, an input device 3 such as a keyboard and a mouse, an output device 4 such as a display, a ROM, a RAM, and a hard disk. The storage devices 5 are provided as physical components, and the devices 2 to 6 are connected by a bus 6. Further, the automatic problem creation apparatus 1 can be connected to a public communication network such as the Internet via a communication interface.

  The storage device 5 stores programs and databases. Specifically, a program 10 in which a procedure for automatically creating a problem is defined, and an item database that stores information on a plurality of items. 21, a category hierarchy database 22 that stores information on the category hierarchy that defines the hierarchical relationship of the category (classification) to which each item of the item database 21 belongs, and information on the part of speech structure of each item in the item database 21. A part-of-speech structure database 23 and an appearance frequency database 24 that stores information on the appearance frequency of each item in the item database 21 are stored.

  Information on a plurality of items in the item database 21 is preferably obtained from a predetermined medium (here, Wikipedia). There are a plurality of articles on Wikipedia that can be accessed by the automatic problem creation apparatus 1 using the Internet. For example, as of April 1, 2011, there are more than 740000 articles on the Japanese version of Wikipedia. An “article” in Wikipedia is a page on which information as an encyclopedia is described, and is also referred to as an “item” conventionally used in an encyclopedia. Hereinafter, this “article” will also be referred to as “item”.

  In Wikipedia, for each item, the title of the item, a commentary explaining the content of the item, and the category to which the item belongs are described, and all of this information is registered in the item database 21. As an example of an item on Wikipedia, the item relating to “Junichiro Koizumi” is illustrated in FIG. 2 and includes the title “Junichiro Koizumi” and “Junichiro Koizumi, 1942 (Showa 17) 1 May 8-) is a former Japanese politician. Advisor to the Center for International Public Policy. He has served as a member of the House of Representatives, Minister of Health and Welfare (69th, 70th and 81st generation), Minister of Posts and Telecommunications (55th generation), Prime Minister (87th, 88th and 89th generation). Is described. Further, it is described that this item belongs to the following ten categories. "Japanese Prime Minister", "Japanese Minister Experienced Person", "Japanese Foreign Minister", "Representative of the House of Representatives", "Liberal Democratic National Assembly", "Kanagawa Prefecture Selected Politician", "Joiichiro Koizumi" “Person from Kanagawa Prefecture”, “1942”, “living person”. Information in this category is displayed at the bottom of the corresponding page on Wikipedia.

  Further, in Wikipedia, the vertical relationship of the category (classification) to which each item belongs is defined as a category hierarchy. FIG. 3 schematically shows a part of the category hierarchy. For example, the categories of “society” and “technology” are located below the “main category”, and “economy” and “ “Living” category, “Bankruptcy” and “Accounting” categories under “Economy” category, “Weapons” and “Space development” categories under “Technology” category ing. An item (article) related to each category belongs to each category. Note that the “major category” in Wikipedia is defined as a collection of only the most major categories that should be the starting point of a search using the category function.

  FIG. 4 illustrates an explanation page of the “Minister” category in Wikipedia. As can be seen from FIG. 4, the information included in the explanation page of the “Minister” category includes the category title, the category commentary, the list of “items” belonging to this ministerial category, the top level of the category, The category hierarchy information meaning the lower (sub) is included. In this item list, titles of “items” belonging to the “Minister” category are listed in alphabetical order.

  Next, the overall software configuration of the automatic problem creation apparatus 1 will be described. As shown in FIG. 5, the automatic problem creation apparatus 1 includes, as functional components, an item collection database preparation unit 30, a category hierarchy database preparation unit 31, a part-of-speech structure database preparation unit 32, and an appearance frequency database preparation unit 33. A question sentence setting unit 34, a tagging processing unit 35, an appearance frequency examining unit 36, an incorrect answer candidate examining unit 37, a correct answer setting unit 38, an incorrect answer setting unit 39, a problem generating unit 41, A flow control unit 43 is provided. Each of these units 31 to 43 is a function realized by the central processing unit 2 executing the program 10 stored in the storage device 5.

  The item database preparing unit 30 is a means for preparing an item database 21 in which data of a plurality of items is stored in a predetermined system. The item collection database preparation unit 30 acquires text data of a plurality of items published on the Internet as “Free Encyclopedia, Wikipedia Japanese Version” via the Internet. Then, an item collection database 21 storing the acquired item data (title, description, category) in a predetermined system is created and stored in the storage device 5.

  The category hierarchy database preparation unit 31 creates a category hierarchy database 22 that defines the vertical relationship of the categories of each item stored in the item database 21 and stores it in the storage device 5. Data stored in the category hierarchy database 22 is also acquired from Wikipedia.

  The part-of-speech structure database preparation unit 32 is a means for preparing the part-of-speech structure database 23 by analyzing the part-of-speech structure of words in all the titles, explanations, and categories of the item database 21. The part-of-speech structure database preparation unit 32 sequentially extracts the titles of all items stored in the item database 21 and sequentially analyzes the part-of-speech structure of the titles. In addition to the title, commentary sentences and categories may be extracted in order, and the part of speech structure database may be prepared by analyzing these part of speech structures. For example, as shown in FIG. 6, the word “society” is analyzed as “society (nouns, proper nouns, organizations)”, the word “economy” is analyzed as “economics (nouns, general)”, and “ The word “Junichiro Koizumi” is analyzed as “Koizumi (noun, person name, surname)” and “Junichiro (noun, person name, first name)”. It is preferable to use appropriate analysis software (for example, an open source morphological analysis engine such as Mecab) for the analysis. Then, the part-of-speech structure database 23 is created by associating the word that is the title of the item stored in the item database 21 with the analyzed part-of-speech information of the word and stores it in the storage device 5.

  The appearance frequency database preparation unit 33 is a means for preparing an appearance frequency database 24 that stores the appearance frequency of the title of each item for all items stored in the item database 21. First, the appearance frequency database preparation unit 33 refers to the item database 21 and the category hierarchy database 22, and for each word that is the title of the item, the word is included in all the explanatory sentences belonging to the category. Investigate how often it appears and how often it appears. Then, a word appearance probability table in each category is created. That is, a word appearance probability table X1 related to one category C1, a word appearance probability table X2 related to another category C2, a word appearance probability table X3 related to another category C3, and the like are created. Then, an appearance frequency database 24 that stores the appearance frequency of each word in all categories and the word appearance probability table for each category in a predetermined system is stored in the storage device 5. Note that the appearance frequency of words of nouns that are not titles or words other than nouns may be investigated and reflected in each appearance probability table in the appearance frequency database 24.

  The question sentence setting unit 34 is a means for setting an arbitrary sentence as a question sentence and storing it in the storage device 5. The sentence used as the question sentence may be acquired from another predetermined medium such as current news, and may be stored in the storage device 5 or may be appropriately input by the user each time. Of course, it may be extracted from the item database 21. If extracted from the item database 21, the user may specify a category in advance with reference to the category hierarchy database 22, and narrow down the items belonging to this category to read out sentences as question sentences. The category specification is preferably executed by the question sentence setting unit 34 displaying a screen prompting the specification on the output unit 4 such as a display, and the user inputting and selecting a desired category via the input device 3.

  In the present embodiment, the question sentence setting unit 34 exemplifies a case where one sentence block serving as a question candidate is set, but the present invention is not limited to this. For example, when a plurality of question sentences are generated simultaneously, a plurality of question candidates can be registered in advance.

  The tagging processing unit 35 is means for performing tagging processing with respect to the question sentence set by the question sentence setting unit 34 with reference to the item database 21. The tag is attached to the word that is the title of the item in the item database 21.

  The appearance frequency examining unit 36 is a means for examining the appearance frequency of each word in the question sentence tagged by the tagging processing unit 35 with reference to the appearance frequency database 24. First, the tagged words are listed, and the appearance frequency (number of times) in all categories of each word is investigated. For example, using the maximum likelihood method, the probability that a problem sentence including a tagged word is generated from the population of each category based on the occurrence frequency table of the word of each category (the magnitude of the correlation with each category) ) For each category. Then, it is estimated that the category having the higher probability is closer to the problem sentence, and the category closest to the problem sentence is specified. In addition, although the case where a category is specified using the maximum likelihood method is illustrated here, the present invention is not limited to this. In short, it is only necessary to analyze the magnitude of the correlation between the question sentence and each category using the words included in the question sentence and the category and identify the optimum category. For example, you may simply select the category that contains the most frequently tagged words as question titles.

  The incorrect answer candidate investigation unit 37 refers to the item database 21 and extracts the incorrect answer candidates for the correct answer set by the correct answer setting unit 38 from the words that are the titles of the items in the item database 21. To do. In particular, in the present embodiment, the incorrect answer candidate investigation unit 37 selects an incorrect answer candidate when a word having a low appearance frequency among all the words tagged by the tagging processing unit 35 is set as an answer object (correct answer). It is a means for investigating with reference to the category hierarchy database 22. The incorrect answer candidate investigation unit 37 extracts an incorrect answer candidate from items in the same category hierarchy as the correct answer.

  For example, “Ministry of Health and Welfare” was selected as the correct answer for some question text, and the “Minister” category in FIG. 4 with a high correlation in context with the question text was identified in order to extract incorrect answers combined with this correct answer. Think about the case. The incorrect answer candidate investigation unit 37 extracts 18 items, excluding “Minister of Health and Welfare”, from the total of 19 item titles belonging to the “Minister” category as candidates for incorrect answers. Since these incorrect answer candidates belong to the same category as the “Minister of Health and Welfare” that is the correct answer, they have a parallel relationship with the “Minister of Health and Welfare” and have the same meaning as the correct answer. As a result, even if an option group in which these incorrect answer candidates and correct answers are mixed is configured, there is no sense of incongruity. Furthermore, the difficulty of the problem can be increased.

  As described above, when items that match the correct answer belong to a plurality of categories, a category closest to the context of the question sentence is selected, and an incorrect answer candidate is extracted from the category. In this case, the maximum likelihood method is used to calculate the probability that a problem sentence containing a tagged word will be generated from the population of each category based on the appearance frequency table of the word of each category. It is preferable to use these categories. That is, as shown in FIG. 7, when the word (item) “DDD” belonging to two categories “AAA” and “FFF” is set as a correct answer, the question sentence of “AAA” and “FFF” For example, if “FFF” is closer to the context of the problem sentence, “GGG”, “HHH”, “III” in the same category hierarchy with respect to the “FFF” category are selected. ”,“ JJJ ”, and“ KKK ”are extracted as incorrect answer candidates.

  Here, when specifying a category from which a candidate for an incorrect answer is extracted from a plurality of categories, determination is made based on the context of the question sentence and the proximity of words included in the category, but the present invention is not limited to this. Not. For example, if the correct item “DDD” belongs to a plurality of categories such as the “AAA” category and the “FFF” category as described above, the incorrect answer is given in the order of the small number of item lists belonging to each category. It is preferable to specify a category from which candidates are extracted. For example, since four items belong to the “AAA” category and six items belong to the “FFF” category, the “AAA” category has a smaller number of item lists. Extract wrong answer candidates from categories. This is because a large number of items belonging to a category is a broad concept of the category itself, and there is a high possibility that various items belong. Therefore, the semantic variation when extracting the wrong answer candidates increases. Further, although not illustrated here, there may be a “DDD” category having the same name as the item “DDD” which is the correct answer. In this case, it is preferable to prohibit extraction of erroneous answer candidates from the “DDD” category. This is because, in the “DDD” category, items that belong to a lower level than the item “DDD” are mainly collected, so that it is difficult to select an incorrect answer that has a parallel relationship with the item “DDD”.

  The correct answer setting unit 38 finally sets a word at a blank position as a correct answer when a predetermined number or more (at least one) of incorrect answer candidates can be acquired based on the survey result by the incorrect answer candidate survey unit 37. In addition, the correct answer setting unit 38, when the incorrect answer candidate investigating unit 37 has not acquired a predetermined number or more of the incorrect answer candidates, makes an error so as to investigate the incorrect answer candidate when another word is the correct answer. It is assumed that an instruction is given to the answer candidate survey unit 37.

  When there are a plurality of words (correct answer candidates) that can be set as correct answers in the question sentence, the appropriateness is determined with reference to the appearance frequency database 24. For example, in order to increase the difficulty level of a question, it is preferable to set a word with the lowest appearance frequency in the entire category as a correct answer. On the other hand, in order to reduce the difficulty level of the question, the word with the highest appearance frequency may be set as the correct answer. Of course, it is possible to select a word arbitrarily without being limited to these cases. The word set as the correct answer is most preferably a proper noun, but the present invention is not limited to this and may be other parts of speech. Furthermore, when comparing the appearance frequency of each word, it is preferable to correct the appearance frequency for each type of part of speech. For example, in the condition setting for setting the correct answer to the one with the lowest appearance frequency, if you want to increase the selection probability of the proper noun, divide the numerical value of the appearance frequency of the proper noun by a numerical value such as 10, for example, Let it be the appearance frequency after correction. If it does in this way, since the appearance frequency of a proper noun is adjusted formally low, it will become easy to be selected as a correct answer.

  In addition, when there are multiple correct answer candidates in the question sentence, it is determined whether at least a part of the correct answer candidate words appears repeatedly in the question sentence. It is also preferable to exclude. Furthermore, if the correct answer candidate word matches the category name in the category hierarchy database 22 or is included in a part of the category name, the difficulty level of the problem may be too low. Is preferably excluded. For example, if there are words that could be correct answers, such as “Member of Parliament” or “Parliament”, these words will be “Society> Politics> Legislation> Parliament> Parliament> Japanese Diet > The names of the “Member of the Diet” category and the “Parliament” category included in the hierarchy of “Member of the House of Representatives” match, so it is preferable to exclude these words such as “Member of the Diet” and “Parliament” from the correct answer candidates. A word that becomes a category name generally means that it is highly known. If such a word is set as a correct answer, the difficulty level of the question may be lowered. Therefore, by excluding this type of word from the correct answer candidates, it is possible to generate a question asking more specific and detailed contents.

  The error answer setting unit 39 is a means for setting at least one of the error answer candidates extracted by the error answer candidate investigation unit 37 as an error answer. Specifically, the incorrect answer setting unit 39 of the present embodiment sets a word having the same part of speech structure as the correct answer word set by the correct answer setting unit 38 as an incorrect answer. The wrong answer setting unit 39 refers to the part-of-speech structure database 23 and sets a word having the same part-of-speech structure as the correct answer from the wrong answer candidates extracted by the wrong answer candidate examining unit 37 as an incorrect answer. This is because the difficulty level of options can be increased by making the correct answer and the incorrect answer have the same or similar part-of-speech structure.

  In addition, although the case where an incorrect answer was selected from the similarity of the part of speech structure of a correct answer and an incorrect answer candidate was illustrated here, this invention is not limited to this. For example, it is preferable that the erroneous answer setting unit 39 selects an erroneous answer by giving priority to an erroneous answer candidate common to any one of Hiragana, Katakana, Kanji, and alphanumeric characters compared to the correct answer. This is because the difficulty of the problem is reduced if Hiragana, Katakana, and Kanji are mixed in the choices. Similarly, it is also preferable to select an incorrect answer with priority given to an incorrect answer candidate having a character amount close to that of the correct answer. This is because if there is a large difference in the amount of characters between the correct answer and the incorrect answer, the difficulty of the problem is likely to decrease. For example, in the example of Fig. 4, if the correct answer is "Minister of Health and Welfare", there is a difference in the amount of text. "Minister", "Minister in charge of Okinawa and Northern Measures", "Minister of Foreign Affairs (Japan)" Ministers, Ministers of Economic and Fiscal Policy, Ministers of Economy, Trade and Industry, Ministers of Health, Labor and Welfare, Ministers of Consumer and Food Safety, Deputy Prime Minister, etc. are preferably excluded from the wrong answers.

  Furthermore, it is preferable that the incorrect answer selection unit 39 selects an incorrect answer by giving priority to an incorrect answer candidate common to the start word or the end word as compared to the correct answer. In the example of FIG. 4, since the answer is “Minister of Health and Welfare”, it is preferable to select one that matches “starting” as the start word and “subject” as the end word. For example, “Minister of Health, Labor and Welfare” can be cited as a match for the starting word “thickness”. On the other hand, when it is desired to avoid that the wrong answer is too similar to the answer, it is preferable to exclude the case where “thickness” as the start word and “subject” as the end word match.

  The question generating unit 41 combines the correct answer set so far, the incorrect answer, and the question sentence in which the position of the correct answer is left blank, and stores it in the storage device 5 as one question. If the number of wrong answers set here exceeds the number (for example, three or two) desired to be listed as wrong answers, the wrong answer candidates are displayed on the output means 4 such as a display. Can be selected manually.

  In the flow control unit 43, for example, when a plurality of question sentences are registered simultaneously in the question sentence setting unit 34, a process of repeatedly applying the above problem generation procedure to the plurality of question sentences is performed. . At this time, even if one is a question sentence, one with a different selected correct answer may be another question. Therefore, by using the flow control unit 43, it is possible to generate a plurality of problems by applying the above-described problem generation procedure a plurality of times to one question sentence.

  A specific example of creating a hole-filling multiple-choice question by the automatic question creation apparatus 1 will be described with reference to FIG. In this embodiment, a case will be described in which a problem is created for the next passage of Ryunosuke Ayukawa's novel “Rashomon”. Here, it is assumed that the following passage that becomes a question sentence is input by the user operating the input device 3 and set by the question sentence setting unit 34 as a question sentence in the storage device 5.

  “It ’s the way of life one day. A genius was waiting for the rain under Rashomon. There is no one other than this man under the wide gate. However, there is a cocoon on a large cylinder that has been stripped in some places. As long as Rashomon is on Suzaku-Oji, there are likely to be a couple of other men, such as Ichi-nagasa and Hata, who are in the rain. No one else except this man. ]

  When a question sentence is set, a tagging process is executed for the question sentence with reference to the item database 21. The tag is assigned to the word that is the title of the item in the item database 21, for example, the noun word indicated by the following brackets.

  “It ’s the way of life one day. One [senior] was waiting for the rain under [Rashomon]. There is no one other than this man under the wide gate. However, there are one [蟋蟀] on a large [cylinder] stripped in some places. In addition to this man, [Rashomon] is located in [Suzaku-Oji], and in addition to this man, there are likely to be a couple of other people, such as the city girl hat and the cocoon hat. No one else except this man. ]

  Next, the appearance frequency database 24 is referred to, and the appearance frequency of the tagged word is investigated. For example, using the database constructed from Wikipedia in this embodiment, “Jinjin” is 144 times, “Rashomon” is 192 times, “Cylindrical” is 1190 times, “Midori” is 22 times, and “Suzaku Oji” is 76 times. The number of times “exceeding” is determined to be 144,415 times, and the “cap” is examined as 303 times. This investigation result is temporarily stored in the storage device 5.

  Furthermore, in order to increase the probability that a proper noun is selected as a correct answer, weighting 11 times only for proper nouns (with an appearance frequency set to 1/11), the above appearance frequency is corrected as follows. “Junior” is 144 times, “Rashomon” is 17.45 times (with correction), “Cylinder” is 1190 times, “Midori” is 22 times, “Suzaku Oji” is 6.91 times (with correction), “ The above is 124415 times, and the “hat” is 303. Next, using the corrected appearance frequency list, the word with the lowest appearance frequency (here, “Suzaku Oji”) is selected as the answer object, that is, the correct answer. Further, items that belong to the category that is in the same category hierarchy as the correct answer and that is closest to the context of the question sentence are extracted from the item database 21 as candidates for the incorrect answer of “Suzaku Oji”. According to the category hierarchy database 22, “Suzaku Oji” belongs to four categories of “road”, “city planning”, “Japanese civil engineering history”, and “civil engineering history”. In such a category state, with regard to the category of “civil engineering history”, which is the category that is identified as the closest to the context of the current problem sentence by the maximum likelihood method, Ayarakigo ruins, remains, stone walls, Ishigaki stacking, ruins, canals, etemen-an-ki, gardening, river development projects, qanat, irrigation, ring villages, megalithic monuments, trenches, construction magazines, engineers, international canals, Jie Suido, King Zezel's Pyramid, Ziggurat, Showa Pond, Suidobashi, Aqueduct, (Suzaku-Oji), Seven Wonders of the World, Eighth Wonder of the World, Bund, Jinghang Grand Canal, Reservoir, Taruyanagi Site, Flood Control, Yasukuni, Tenmaoike Park , Dujiangyan, earthen wall, permanent wall (Narashino), plate construction, wall (civil engineering), Hiraike Park, pyramid, cultural heritage (world heritage), Bemster reclaimed land, fence, moat, Pont du Gard, mustard, Manno Pond, Yamaike, Yu Coast Namiyoke stone wall, irrigation canals, landmark, watershed, lingqu canal, to extract Roman road, a word that has become a Roman aqueduct, such as a plurality of items title as a wrong answer candidate.

  Next, the part-of-speech structure of the above-mentioned candidate answer structure is investigated with reference to the part-of-speech structure database 23, and the answer candidates having the same part-of-speech structure as the correct answer part-of-speech structure are narrowed down as erroneous answers. Here, the part-of-speech structure of “Suzaku Oji” is analyzed as “noun, proper noun, region, general + noun, general”. ”,“ Rome Road ”and“ Hiraike Park ”.

  Next, the question sentence with the correct answer position blank, the correct answer, and the narrowed down answer are combined into one question and stored in the storage device 5. Here, the problem as shown in the lower part of FIG. 8 is completed. Although not particularly shown here, if the difficulty of the combination of correct and incorrect answers is low, an optimization button is selected on the question confirmation screen. When this optimization button is pressed, the wrong answer setting unit 39 changes the wrong answer selection rule. For example, even if the part-of-speech structures of the correct answer and the incorrect answer do not completely match, compared to the correct answer, it is common in Hiragana, Katakana, Kanji, and alphanumeric characters, or compared with the correct answer Precisely select candidates for wrong answers with similar category lists. As a result, “Hiraike Park”, “Tenma Oike Park”, and “Manno Pond” are selected as incorrect answers. In the present embodiment, the degree of difficulty of the problem increases when optimization is performed. When executing this optimization, the error answer setting unit 39 displays a plurality of error answer setting rules (for example, complete part-of-speech match, partial part-of-speech match, kanji numerals / katakana / hiragana match) on the screen, and However, the optimum error setting rule may be selected while viewing the error answer candidates.

  It is preferable to create a large number of problems by repeating these procedures and store a problem collection database in which the created problems are collected in the storage device 5.

  Since the automatic problem creation apparatus and creation method according to an embodiment of the present invention are configured as described above, when the user executes the program 10 of the automatic problem creation apparatus 1, the automatic problem creation apparatus 1 is defined in the program 10. It is possible to automatically create a problem according to the procedure described, and to save the labor of the problem creator for creating the problem. Moreover, since any question sentence is automatically tagged to the included word, various multiple-choice questions can be automatically generated. For example, question sentences can be set for various fields other than the above-mentioned novels, such as history, geography, current affairs, fashion, and law, and multiple-choice questions that are highly complete in a wide range of fields are automatically selected. Can be created.

  Further, the question automatic creation device 1 displays a screen prompting the user to input a question sentence on the output device 4 such as a display, so that the user can appropriately set the question sentence as necessary. In addition, the automatic question creation apparatus 1 displays wrong answer candidates and displays a screen on which an incorrect answer can be selected on the output device 4 such as a display, so that the user can freely determine the wrong answer as necessary. You can also.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to the said embodiment, A various change is possible in the range which does not deviate from the meaning of this invention.

  For example, in the above embodiment, the information in the item database 21 is acquired from Wikipedia. However, the item database 21 may be constructed by acquiring information from another medium such as a dictionary database. In addition, information from a predetermined medium is not limited to being acquired via the Internet, but may be acquired from a recording medium such as a CD in which information is stored.

  The automatic problem creation apparatus 1 may cooperate with a mobile terminal such as a mobile phone so that the user can operate the mobile terminal to execute the program 10 to automatically create a problem. At this time, it is preferable that the user inputs / selects the category of the question to be asked through the screen of the mobile terminal, or sets specific question sentences individually.

DESCRIPTION OF SYMBOLS 1 Problem automatic creation apparatus 2 Central processing unit 3 Input device 4 Output device 5 Storage device 6 Bus 10 Program 21 Item database 22 Category hierarchy database 23 Part of speech structure database 24 Appearance frequency database 31 Item database preparation part 32 Part of speech structure database preparation part 33 Appearance frequency database preparation unit 34 Question sentence setting unit 35 Tagging processing unit 36 Appearance frequency investigation unit 37 Error answer candidate investigation unit 38 Correct answer setting unit 39 Error answer setting unit 41 Problem generation unit 43 Flow control unit

Claims (7)

Completing multiple-choice questions using an item database that is composed of an electronic computer and stores the title of the item, explanation text explaining the content of the item, and the category to which the item belongs. An automatic problem creation device that automatically creates
Items that acquire information on the title, commentary, and category of the item from a predetermined medium, create the item database that stores the acquired information in a predetermined system, and store it in the storage device of the electronic computer Collection database preparation department,
A question sentence setting section for setting a question sentence to be questioned;
A tagging processing unit that tags a predetermined word in the question sentence set in the question sentence setting unit;
A correct answer setting unit for setting one of the words tagged by the tagging processing unit as a correct answer;
With reference to the item database, an incorrect answer candidate examining unit that extracts a candidate for a correct answer set in the correct answer setting unit from words that are titles of the items in the item database;
An automatic question creation apparatus, comprising: an error answer setting unit that sets at least one of error answer candidates extracted by the error answer candidate investigation unit as an error answer. Create a part-of-speech structure database that stores the result of analyzing the part-of-speech structure of each word that is the title of the item stored in the item database in a predetermined system, and prepares the part-of-speech structure database to be stored in the storage device Further comprising
2. The problem according to claim 1, wherein the incorrect answer setting unit refers to the part of speech structure database and sets, as an incorrect answer, an incorrect answer candidate in which a word set as a correct answer and a part of speech structure match. Automatic creation device. It further includes a category hierarchy database preparation unit that acquires information on the vertical relationship of the categories from the predetermined medium, creates a category hierarchy database that stores the acquired information in a predetermined system, and stores the information in the storage device,
The said incorrect answer candidate investigation part extracts the word in which the category hierarchy and the word set as a correct answer refer to the said category hierarchy database as an incorrect answer candidate, It is characterized by the above-mentioned. Problem automatic creation device. The erroneous answer candidate investigation department
4. The automatic question creation apparatus according to claim 3, wherein a category closest to the context of the question sentence is selected, and a word that is a title of an item belonging to the selected category is extracted as a wrong answer candidate. . 5. The automatic problem creation apparatus according to claim 1, wherein the tagging processing unit tags a word that is a title of an item in the item database. Appearance frequency database that stores, in a predetermined system, the results of examining the appearance frequency of each word that is the title of the item stored in the item database with respect to the commentary stored in the item database. And an appearance frequency database preparation unit to be stored in the storage device,
An appearance frequency examining unit that refers to the appearance frequency database and investigates the appearance frequency of the word tagged by the tagging processing unit;
The said correct answer setting part sets the said word as a correct answer based on the said appearance frequency, when there are two or more of the said tagged words, The any one of Claims 1-5 characterized by the above-mentioned. Problem automatic creation device. Using multiple item list databases that store the title of the item, the description that explains the content of the item, and the category to which the item belongs, the computer automatically fills the multiple-choice question This is an automatic problem creation method
The central processing unit of the electronic computer is
Obtain each item title, commentary and category information from a predetermined medium, create the item database that stores the acquired information in a predetermined system, and store it in the storage device of the electronic computer,
Set the question sentence to be questioned,
Tag certain words in the set question sentence,
Set one of the tagged words as a correct answer,
Extracting incorrect answer candidates for the set correct answer from the words that are the titles of the items in the item database,
A method for automatically creating a question, wherein a process of setting at least one of the extracted wrong answer candidates as a wrong answer is executed.

Images