JP4018673B2 - Multiple choice language test question automatic creation program - Google Patents

Description

  The present invention relates to an apparatus for automatically creating a test question, and more particularly to an apparatus for automatically creating a multiple-choice language test question.

  Language education tests include so-called descriptive questions and multiple-choice questions. The language education test is indispensable for measuring the learner's learning proficiency level. For this reason, language teachers spend a great deal of time creating test questions. Therefore, there is a need for measures that facilitate problem creation.

  Comparing the descriptive test problem with the multiple-choice test problem, there are the following differences. First, for descriptive test questions, it is relatively easy to create a question. However, in the case of a descriptive test, there is a problem that it is difficult to input and score an answer to the question. In particular, there is a problem that scoring varies considerably depending on the subjectivity of the grader. In the case of a descriptive test, there is also a problem that it is very difficult to mechanize the scoring.

  Compared to this, the multiple-choice language test question has the advantage that it is easy to input an answer and easy to score. In addition, it has the advantage of being objective in the sense that there are no fluctuations in scoring by the evaluator. Therefore, when there are a large number of examinees or in a computer-based test (CBT), a multiple-choice test is widely used. In addition, so-called e-learning advocating the goal of “learning anytime, anywhere” is due to problems specific to mobile devices such as mobile phones or PDAs (Portable Digital Assistants), such as restrictions such as small screens and poor input devices. In view of this, the multiple-choice language test questions may play an important role.

  However, in the case of a multiple-choice type test, there is a problem that the time and effort required to create a problem is considerably large compared to a descriptive test. In particular, there is a problem that it is difficult to create a problem that matches the domain to be learned (travel conversation, business conversation, news, etc.) and that matches the target of the subject. For educational institutions that do not have many years of experience or accumulation in the domain to be learned, it is very difficult to create a multiple-choice language test problem for a desired domain in accordance with a predetermined level of achievement. In order to solve such a problem, it is necessary to mobilize a large number of personnel, which is a factor in increasing the cost of preparing the test questions. It would be desirable to be able to automatically create a multiple choice language test question in some way and that would be solved.

Therefore, an object of the present invention is to provide a program that allows a computer to function to automatically create a multiple choice language test question.

Another object of the present invention is to provide a program that allows a computer to function to automatically create a problem of a desired domain that is a multiple choice language test problem.

Still another object of the invention is a multiple-choice language exam is to provide a program capable of making a computer function to automatically create the desired levels reached problems desired domain .

A program for automatically creating a multiple-choice language test question according to the first aspect of the present invention relates to a corpus storage means for storing a computer-readable corpus , a word dictionary, a thesaurus, and a word for creating options . Language information storage means and corpus storage means for storing linguistic information in a computer-readable format including an error list in which errors are collected, a word list listed together with the frequency and importance of words, or any combination thereof From the corpus stored in the sentence selection means for selecting the sentence that is the basis of the multiple-choice language test problem according to a predetermined selection method, and among the components that constitute the sentence selected by the sentence selection means, a component determining means for determining the components of the target portion of the multiple-choice by a predetermined condition, stored in the language information storage unit Of the language information, based on the language information related to the determined component replacement for creating replacement candidate satisfying the Shi pairs components determined by the configuration element determination means given conditions In the selected sentence, the validity of the sentence obtained by replacing the determined constituent element with the replacement candidate in the selected sentence is determined in accordance with a predetermined element of the constituent string including the replacement candidate included in the obtained sentence. Validity determination means for determining in accordance with statistical criteria based on the appearance frequency in the corpus of the corpus, replacement candidates determined by the validity determination means that the sentence is not valid, and the original sentence in a predetermined form a problem preparing means for preparing multiple-choice language test questions causing a computer to function and by combining.

A sentence is selected from the corpus storage means, and a replacement candidate for a component satisfying a predetermined condition is created. The validity of the sentence after replacing the component with a replacement candidate is determined by the validity determination means according to a statistical criterion based on the frequency of occurrence of the component string containing the replacement candidate included in the sentence in a predetermined corpus. Then, the multiple choice language test question is automatically created by combining the replacement candidate determined to be invalid and the original sentence selected first, in a predetermined form. If you change the corpus, you can automatically create multiple-choice language questions with the desired domain or level. Furthermore, by leaving only the sentence that is not valid from the replacement candidates, it is possible to create an appropriate multiple-choice language test question in combination with the sentence selected first. A large number of multiple-choice language test questions can be created in a short period of time without the need for humans to think about question sentences and option sets or to verify the correctness of problems.

Preferably, the replacement candidate creating unit creates a plurality of replacement candidates for the determined component based on language information related to the determined component among the language information stored in the language information storage unit. includes means for, problems creating means, among the selected sentence, means for creating a substituted problem statement the determined component in space, among the plurality of replacement candidates, reasonable An option creation means for creating an option set consisting of a predetermined number of replacement candidates determined by the sex determination means and a predetermined number of replacement candidates determined, and a problem sentence and an option set as a predetermined Shaping means for shaping into a problem form.

More preferably, problems creating means, among the selected sentence, the determined components for making substitution to issue statements appropriateness determination means the substitution candidate is determined that statement is not valid Means and a shaping means for shaping the question sentence so as to display a candidate for replacement and a predetermined number of words or word groups other than the candidate for replacement as option candidates.

Preferably, the multiple-choice language test question automatic creation program further causes the computer to function as a means for receiving input from the user regarding a given condition and giving it to the conditional replacement candidate creation means.

The automatic program for creating multiple-choice language test questions further reads a corpus designated by the user from the computer-readable corpus group storage means for storing a corpus group consisting of a plurality of corpora and stores it in the corpus storage means The computer is made to function as a corpus reading means for doing this .

  Preferably, the corpus is a bilingual corpus including a plurality of parallel translations of sentences in the first language and sentences in the second language, and the sentence selecting means is selected from the bilingual corpora stored in the corpus storage means according to the selection method. A bilingual selection means for selecting a bilingual source of the selected language test question is included.

More preferably, the constituent element determining means includes means for determining a constituent element satisfying a predetermined condition among the constituent elements of the selected bilingual sentence in the second language, and the replacement candidate creating means includes the language information. based on the stored language information in the storage means includes means for creating a replacement candidate in the plurality of second language for the determined components, problems creating means, the first language of the selected translation And a means for creating a question sentence including the sentence in which the determined component is replaced with a blank space among the sentences in the second language of the selected parallel translation, and among a plurality of replacement candidates, Option creation means for creating a set of choices consisting of a predetermined number of second language replacement candidates determined in advance by the validity judgment means and the determined constituent elements, and question sentences and options A set with a given problem form And a shaping means for the form.

Problem creating means is determined and sentence in the first language of the selected translation, of the sentence in the second language of the selected translation, the determined elements, the statement is not valid due appropriateness determination means Means for creating a question sentence including a sentence obtained by replacement with a replacement candidate, and a replacement candidate and a predetermined number of words or word groups other than the replacement candidate among sentences in the second language of the question sentence And a shaping means for shaping the question sentence so as to be displayed as option candidates.

Preferably, the constituent element determining means includes means for determining an entire sentence in the second language of the selected parallel translation as a constituent element satisfying a predetermined condition, and the replacement candidate creating means is stored in the language information storage means. based on the language information includes a plurality of machine translation means for mechanically translating the sentence of the first language of the selected translation, issue creation means, second language sentence of the selected translation, and more A means for creating an option set composed of sentences determined not to be valid by the validity determination means, a sentence in the first language of the selected parallel translation, and an option set. Means for creating a multiple choice language test question by combining in a predetermined form.

More preferably, multiple-choice language exam automatic creation program statement selected means, component determining means, substituted candidate creating unit, adequacy judging means, and controls the and problems creating means, a predetermined number of multi A computer functions as a control means for creating a limb selection language test question.

[Overview of multiple choice language test questions]
The multiple-choice language test problem includes a space filling problem shown in FIG. 1, an error detection problem shown in FIG. 3, a correct answer sentence or an error sentence selection problem (not shown), and the like. An apparatus for creating a multiple-choice language test problem according to an embodiment described below can create a plurality of types of multi-choice language test problems including at least both a space filling problem and an error detection problem.

  First, the empty space replenishment problem shown in FIG. 1 will be described. Referring to FIG. 1, in the empty space replenishment question, the examinee is required to replenish the empty space 24 in the question sentence 22 by selecting from the options 26 so that the sentence becomes a correct sentence.

  In the device according to the present embodiment, in such empty space replenishment problem, it is utilized that the correct answer is a sentence and the incorrect answer is a sentence. In addition, a corpus that collects exemplary sentences is prepared so that correct answers can be created without error, and a multiple-choice language test question is created using the sentences included in the corpus as seeds.

  More specifically, questions and options (candidates for correct and incorrect answers) are generated based on a corpus that collects exemplary sentences. The correct answer can be extracted from the corpus. It is determined whether or not each combination of a question and an incorrect answer option candidate is valid as a sentence. Here, since incorrect option candidates are generated, those that are determined to be appropriate as sentences in the above determination are excluded from the options. In the following embodiment, the validity of a sentence is calculated by a predetermined method, and only those whose validity falls below a predetermined threshold are left as options.

  FIG. 2 shows, as an example, a sentence group 30 generated from a problem and a solution candidate. Referring to FIG. 2, this sentence group includes a total of five sentences 40, 42, 44, 46 and 48. The underlined portion is a solution candidate, and the one shown in bold is the correct answer. In the example shown in FIG. 2, both the first sentence 40 and the fifth sentence 48 are correct. In general, in the case of multiple choice, the number of options and the number of correct answers are constant. In the present specification, the former is 4 and the latter is 1 unless otherwise specified. Therefore, in this case, the first option 40 or the fifth option 48 is deleted.

  The error detection problem can be realized in almost the same way as the empty space filling problem. FIG. 3 shows an example 50 of the error detection problem. Referring to FIG. 3, the problem is that an error part is selected from the four options a) to d) of the sentence 52. In this case, a sentence is first determined, a word is determined as an option, word substitution is performed on one of the sentences, and it is confirmed that the sentence is validated as a non-sentence. Note that the same is true if the alternative word is determined after the position of the replacement word is determined first.

[First Embodiment]
<Configuration>
Hereinafter, the configuration and operation of an apparatus according to an embodiment for realizing such a procedure will be described. The following apparatus is merely one embodiment of the present invention, and the present invention is not limited to the following embodiment.

  FIG. 4 shows a multiple-choice language test question creation system 60 according to the present embodiment in the form of a block diagram. Referring to FIG. 4, this multiple-choice language test question creation system 60 is a condition selection unit 62 used by the user to set conditions for question creation, and serves as a seed for questions according to various domains. A corpus group 66 composed of a plurality of corpuses each including a correct sentence and a corpus suitable for a designated domain are selected from the corpus group 66 according to the conditions set by the condition selection unit 62, and the above-described empty space replenishment problem A problem creation device 64 for automatically creating a multiple-choice language test question 74 consisting of any of the error detection problems and various language information data to be referred to when the problem creation device 64 creates a problem Various data storage units 68. The question creation device 64 also has a function of checking the validity of a sentence during creation of a multiple choice language test question using the so-called Internet 70.

  It is assumed that the multiple-choice language test question creation system 60 is realized by a so-called computer system. The condition selection unit 62 includes a keyboard (not shown), a pointing device such as a mouse, software for performing input / output using a condition setting screen, a display device, and the like. In the present embodiment, the condition selection unit 62 specifies the following conditions.

(1) Domain to be created (2) Type of multiple choice language test question to be created (vacancy refilling problem or error detection problem)
(3) Number of questions to be created (4) Specified sentence selection conditions at the time of question creation (5) Conditions for determining the position of the vacant space for the vacancy replenishment problem (6) Condition (7) Candidate Validity Verification Method The problem creating device 64 is realized by computer hardware and a computer program executed on the computer, as will be described later. In this embodiment, the corpus group 66 and the various data storage units 68 are realized by a storage device provided in the same computer as that for realizing the problem creation device 64.

  The problem creating device 64 uses the condition selection unit 62 to select a corpus of a domain designated by the user from the corpus group 66, and stores the corpus selected by the corpus selection unit 80. A corpus storage unit 82 and a seed sentence selection for selecting a seed sentence that is a source of a problem from among corpora stored in the corpus storage part 82 according to a condition specified by a user using the condition selection unit 62 Part 84.

  Further, the problem creating device 64 further creates a space filling problem according to the type of problem designated by the user using the condition selecting unit 62 among the seed sentences selected by the seed sentence selecting unit 84. Select the word at the position according to the specified condition as a blank, and when creating an error detection problem, select the word position that will be the choice as many as the number of choices, along with information indicating the type of the problem and the seed sentence Various data stored in the various data storage unit 68 are output in accordance with the information indicating the problem type and the problem position for the problem position selection unit 86 for output and the seed sentence given from the problem position selection unit 86. An option candidate generation processing unit 88 for generating option candidates for a multiple-choice language test by creating an option while referring to a sentence related to the option candidate generated by the option candidate generation processing unit 88 In order to determine the validity, the validity determination data search unit 92 for searching the Internet 70 and the sentence related to the option candidate received from the option candidate generation processing unit 88 are valid. In order to search the validity determination data using the sex determination data search unit 92 and determine the validity as a sentence, and also determine the appropriateness of the problem itself according to the problem type given from the condition selection unit 62 The validity determination unit 90 is included.

  When the validity determination unit 90 determines that the problem is not appropriate as a result of determining the validity of the option candidates, the validity determination unit 90 discards the problem and creates a next problem for the seed sentence selection unit 84. Has a function to direct.

  The problem creation device 64 further includes a shaping unit 94 for shaping the problem determined to be appropriate by the validity determination unit 90 into an appropriate format according to the type of problem, and the problem shaped by the shaping unit 94. A related information addition processing unit 96 for assigning information (example sentence, definition, meaning, etc.) related to the problem from the data on the Internet 70 and the data stored in the various data storage unit 68; And an output unit 98 for performing a process of writing, as a multiple-choice language test question 74, a problem to which related information is given and shaped, which is output from the assignment processing unit 96, into a storage device.

  In the case of multiple choice language test questions, it is not desirable that the same word appears in different question sentences. Therefore, the question creating device 64 includes the already-presented option storage unit 100 for storing the options in the multiple-choice language test question output by the output unit 98, and the option candidate generation processing unit 88 is the already-chosen options at the time of problem creation. With reference to the memory | storage part 100, it has the function to select the word at the time of question preparation so that the already-existing word may not be included in a subsequent question.

  FIG. 5 shows a configuration example of the corpus 110 stored in the corpus storage unit 82. Referring to FIG. 5, corpus 110 is a monolingual corpus in the present embodiment and includes a plurality of example sentences 116. Each example sentence 116 includes an identification number (ID) 112 and a text 114. The corpuses included in the corpus group 66 do not have to be all in the same format, but in order to simplify the configuration of the problem creating apparatus 64, it is desirable that the corpora be in a unified format.

  FIG. 6 shows the configuration of various data storage units 68 shown in FIG. Referring to FIG. 6, various data storage unit 68 uses error list 120 including a plurality of lists in which errors related to a certain word are collected according to the type of error, and a dictionary related to words, which are used when creating options. 122, a thesaurus 124 that records synonyms, antonyms, and the like, and similar relationships between words, and a word list 126 that lists predetermined words along with their frequencies and importance levels.

  The error list 120 includes a list 130 in which a set of words with similar spellings is collected, a list 132 in which a set of a certain word and a word misspelled with respect to the word is collected, and so-called Japanese-English. And a list 134 collected together with words (word strings).

  As an example of a set of words included in the list 130, there is “moss, miss, message”. As an example of a set of words included in the list 132, there are “flight, flight, flit, flite” and the like. As an example of a set of words included in the list 134, there is “night (night), night game”.

  As will be described later, the multiple-choice language test question creation system 60 according to the present embodiment can be realized by computer hardware and software executed thereon. FIG. 7 shows the flow of the main process of the program constituting the software for realizing the multiple choice language test question creation system 60 in the form of a flowchart. Hereinafter, the control structure of the program will be described with reference to FIG.

  Referring to FIG. 7, when this program is activated, settings are read in step 140. This setting is set by the condition selection unit 62 shown in FIG. 4 and written in the storage device or the like. Subsequently, in step 142, the type of test is determined according to the read setting. If the test type is a vacancy replenishment problem, the process proceeds to step 144. If the test type is an error detection problem, the process proceeds to step 146. Similarly, appropriate processes in steps 144 to 148 are executed in accordance with the selected test type.

  FIG. 8 is a flowchart showing a control structure of a program for performing the empty space replenishment problem creation process executed in step 144. Referring to FIG. 8, first, in step 150, 0 is substituted into variable i representing the problem number being created. In step 152, 1 is added to the variable i. In step 154, it is determined whether or not the value of the variable i exceeds the value MAX set as the number of problems to be created. If the value of the variable i exceeds the value MAX, the process is terminated. If the value of variable i is less than or equal to value MAX, control proceeds to step 156.

  In step 156, a seed sentence is selected by a predetermined method from the sentences stored in the corpus storage unit 82 shown in FIG. As will be described later with reference to FIG. 9, the length of the seed sentence selected in this step is limited.

  Subsequently, in step 158, a position to be vacant in the seed sentence is determined. The position of the vacant space can also be selected randomly from the whole, or can be selected on the basis of the part of speech and the position of the word. For example, “leftmost verb (eg“ lost ”in“ I lost the return ticket of my flight ”)”, “frequent preposition (eg“ of ”above)”, or “rightmost determiner (eg above) "My") "etc. can be selected.

In step 160, an option set made up of option candidates is created. In this step, after saving (storing) the word existing in the empty position selected in step 158 as “correct answer”, a plurality of error option candidates are created, and option candidates for the i-th problem are created. Let it be set A _i . In the present embodiment, the maximum number of option candidates created is M, and each option candidate is represented as A _{i, m} (1 ≦ m ≦ M). Details of the processing in step 160 will be described later with reference to FIG. As will be described later, in the process of step 160, it is considered inappropriate to select the same word as the word included in the already-explained problem as an option candidate. Therefore, such a word is not regarded as an option candidate.

Thereafter, in step 162, 0 is substituted into a repetition control variable m for controlling the following repetition. 1 to the summing variable m in step 164. Subsequently, at step 166, it is determined whether or not the value of the variable m exceeds the value M. If the value of variable m exceeds value M, control proceeds to step 172. Otherwise, control proceeds to step 168. The processing after step 172 will be described later.

In step 168, it is determined whether or not the sentence including the option candidates A _{i, m} is a valid sentence. In the present embodiment, the validity of the sentence is determined using the Internet 70. Specifically, in this embodiment, a word string formed of a word sequence and choice of up before and after the content words of the cavity searches in U E Bed Internet 70, including choice candidates A _i, a _m if Judge that the sentence is valid. The purpose here is, since it is possible to find alternatives candidate as an error, selection candidates Ai, if the sentence containing the m is determined to be a valid statement, choice candidates Ai from the set A _i choice candidate in step 170, Delete m and return to step 164. Otherwise, the sentence including the option candidate A _{i, m} is a non-sentence, and the option candidate A _{i, m} is suitable as an option, so that it remains in the option candidate set A _i and returns to step 164 without doing anything.

If the value of variable m exceeds value M at step 166, control proceeds to step 172. In step 172, it is determined whether or not the number of elements | A _i | included in the option candidate set A _i is greater than 2 (that is, 3 or more). If it is greater than 2, the process proceeds to step 174. In other cases, the problem cannot be constructed because the number of options is insufficient, so that the problem creation is stopped for this option set A _i , and the process returns to step 156 to the next. Redo from the selection of seeds.

In step 174, correct choices are added to the choice set A _i , arranged in a random order, and the contents are shaped to give a problem as shown in FIG. In step 176, related information such as example sentences searched on the web for each option, meanings read from the dictionary, and the like are given to the problem. Further, in step 178, this problem is stored as an existing option in a storage device such as the existing option storage unit 100 shown in FIG. This information is used in the option candidate creation process in step 160.

  In step 180, the problem created in this way is output to a predetermined storage device. Thereafter, control returns to step 152 to begin creating the next question.

  With reference to FIG. 9, the seed sentence selection process performed in step 156 of FIG. 8 will be described. First, in step 181, a seed sentence candidate is selected from the corpus. At this time, in the present embodiment, the user can designate a seed sentence selection method from a plurality of methods. For example, selecting a seed sentence at random, conditions regarding sentence attributes (length, frequency, readability, etc.), grammatical items included in the sentence or scenes where the sentence is used, such as sentences containing verbs, exclamation sentences, It is possible to select at random from narrowed down conditions such as a sentence including a specific preposition, etc., or to select in a predetermined order according to the above attribute values.

  In step 182, it is determined whether or not the morpheme length of the seed sentence candidate selected in step 181 is greater than a predetermined threshold value TH1. Such a determination is made because there is not enough information in a too short sentence, it becomes difficult for the examinee to select an option, and the test may not be established. If the morpheme length is larger than the threshold value TH1, the process proceeds to step 184. Otherwise, the process returns to step 181 because it is inappropriate as a seed sentence, and the next seed sentence candidate is selected. On the other hand, in step 184, it is determined whether or not the morpheme length is smaller than another threshold value TH2 (TH2> TH1). Such a determination is made because if the problem sentence is long, it may be inconvenient to take a place and display the problem on a narrow screen such as a mobile phone. If the morpheme length is smaller than the threshold value TH2, the seed sentence selection process is terminated using this seed sentence candidate as a seed sentence. Otherwise, return to Step 181 and repeat from the selection of the next seed sentence candidate.

FIG. 10 is a flowchart showing details of the option candidate creation process executed in step 160 of FIG. Referring to FIG. 10, first, in step 190, set A _i consisting of option candidates is set to empty set φ, and 0 is assigned to repetition control variable j. In step 191, 1 is added to the variable j. In step 192, it is determined whether the value of the variable j exceeds the maximum number M of option candidates. If j exceeds M, the option candidate creation process is terminated. If j is M or less, the process proceeds to step 193.

  In step 193, the candidate word is selected with reference to the various data storage units 68 shown in FIG. In this case, the following selection method is possible.

  (1) Maintaining the part of speech and utilization (past tense, singular present, etc.) of an alternative word and replacing it with another word so as not to deviate from the grammar even if the word in question is replaced In this case, words can be randomly selected from the dictionary 122, synonyms or antonyms can be selected from the thesaurus 124, and words can be extracted from the word list 126 and replaced in the order of word frequency and importance.

  (2) The part of speech and the utilization (past tense, singular present, etc.) are replaced while maintaining the word base of the word at the position so as to deviate from the grammar by replacing the word. For example, referring to the dictionary 122, the utilization form of the word is changed, or a certain word is replaced with a word with a different part of speech corresponding to the word.

  (3) A set of error words including the word at the position is taken out from the error list 120 shown in FIG. 6 and replaced with the words in the error list 120 so as to reproduce errors with high frequency. For example, as shown in FIG. 6, the original word position is replaced with a similar spelling word, a misspelled word, Japanese-made English, or the like.

  After selecting an option candidate, in step 194, the frequency of the word is checked with reference to various data storage units 68, and it is determined whether or not the frequency is smaller than a predetermined threshold value TH3. This is because, in the case of words that are too infrequent, the examinee may think that the word of the choice is incorrect just because he / she is not familiar with it, and is considered inappropriate as an option. Therefore, if it is determined in step 194 that the frequency is smaller than the threshold value TH3, the process returns to step 193 to select the next option candidate. Otherwise, go to step 195.

In step 195, it is determined whether or not the option candidate is the same as the word that has already appeared in the created problem with reference to the content stored in the already-existing option storage unit 100 shown in FIG. Such determination is made because the already-existing word is not valid as an option candidate. Therefore, if the option candidate is an existing word, the process returns to step 193, and the above-described processing is repeated for the next option candidate. If not, the process proceeds to step 196, and this option candidate is added to the option candidate set A _i . Thereafter, the process returns to step 191.

  The above is a flowchart of the program executed when, for example, the problem creating device 64 creates the empty space refilling problem.

  On the other hand, the error detection problem creation processing executed in step 146 shown in FIG. 7 can also be realized by a similar program. FIG. 11 is a flowchart showing the control structure of the program adopted in the problem creating apparatus 64 of this embodiment. In FIG. 11, the same processes as those in FIG. Therefore, detailed description thereof will not be repeated.

  In FIG. 11, after selecting a seed sentence in step 156, control proceeds to step 200. In step 200, a predetermined number of words (4 by default, but can be specified by the user) are selected as options, and one of them is selected as an error word position.

Then in step 202, by replacing the selected word in another word, to convert the Tanebun B _i to another sentence B _i ^'. In the present embodiment, the replacement word here is determined using each list of the error list 120 shown in FIG. It should be noted that the word determined as an option in step 200 and the replacement word determined in step 202 are both words that have already been mentioned in the previous problem (regardless of whether space filling or error detection is used). Make it not exist. This replacement word is limited to the content word in the experiment described later, and the function word is excluded from the candidates.

In step 204, it is determined whether or not the replaced sentence B _i ^′ is a non-sentence. This determination can be realized by a process similar to the process performed in step 168 of FIG. If the sentence B _i ^′ is a non-sentence, this sentence B _i ^′ can be regarded as an error detection problem.

Therefore, the sentence B _i ^′ is shaped in step 206 to give a problem as shown in FIG. In step 208, the example sentence acquired from the Internet 70 or the like in the determination of step 204, the meaning information acquired from the dictionary, etc. are assigned to this problem as related information. In step 210, this problem is stored as an existing option in the existing option storage unit 100 of FIG. In step 212, this problem is output. Thereafter, the process returns to step 152 to create the next problem.

  In this way, an error detection problem can be automatically created. There are other forms of multiple-choice language test questions, but they can be realized in the same way as described above. Therefore, the control structure of each program will not be described here.

<Operation>
The multiple choice language test question creation system 60 described above operates as follows. Referring to FIG. 4, it is assumed that corpus groups 66 corresponding to various domains and levels are prepared in advance. It is also assumed that an error list 120, a dictionary 122, a thesaurus 124, and a word list 126 for creating a problem are prepared in advance.

  The user performs various settings using the condition selection unit 62. For example, the domain to be questioned, the type of multiple choice language test question, the number of questions to be created, the level of the question to be created, etc. are designated.

  The corpus selection unit 80 selects an appropriate level corpus in the appropriate domain from the corpus group 66 in response to the designation by the user, and stores it in the corpus storage unit 82.

  The seed sentence selection unit 84 selects a seed sentence from the corpus stored in the corpus storage unit 82 in accordance with the condition specified by the condition selection unit 62, and provides the selected problem sentence to the problem position selection unit 86 together with the specified condition. For example, if the problem to be created is a vacancy replenishment problem, the problem position selection unit 86 determines which word is vacant. If the problem to be created is an error detection problem, the problem position selection unit 86 determines a word position as an option in the seed sentence, and further determines which word is replaced with the error word. This order may be reversed.

  The option candidate generation processing unit 88 receives the seed sentence in which the problem position is determined by the problem position selection unit 86, refers to the various data storage units 68, and generates option candidates according to the type of the problem. For example, if the problem is an empty space replenishment problem, for example, a thesaurus 124 shown in FIG.

  The validity determination unit 90 determines whether the option candidate thus generated is valid. If it is a space filling problem, it is determined whether a sentence in which words included in the option set are filled in the space becomes a non-sentence except for one. If it is an error detection problem, it is determined whether or not a sentence obtained as a result of inserting an error word at the problem position becomes a non-sentence. At this time, the validity determination unit 90 searches the web on the Internet 70 using the validity determination data search unit 92 and determines the validity of the sentence.

  In the case of an empty space supplementation problem, if all sentences except for one (correct answer) are determined to be non-sentences except for one (correct answer), the validity determination unit 90 selects an option candidate. Is given to the shaping unit 94. In the case of an error detection problem, if the sentence obtained by inserting the error word at the problem position in the seed sentence is a non-sentence, the validity determination unit 90 gives this problem to the shaping unit 94. In other cases, the validity determination unit 90 discards the problem given from the option candidate generation processing unit 88. Whether or not the problem is adopted, the validity determination unit 90 gives a signal that instructs the seed sentence selection unit 84 to start creating the problem again.

  When a problem is given from the validity determination unit 90, the shaping unit 94 shapes it into an appropriate shape according to the type of the problem and gives it to the related information addition processing unit 96. The related information addition processing unit 96 assigns example information obtained from the Internet 70 and related information such as word meanings obtained from the various data storage units 68 to the options, and provides them to the output unit 98. The output unit 98 writes this problem in the existing choice storage unit 100 and then outputs the problem as a multiple-choice language test question 74.

  When the signal for instructing the creation of the next question is given from the validity determination unit 90, the seed sentence selection unit 84 determines whether or not the number of questions already created has reached the number of questions specified by the condition selection unit 62. If it is determined and reached, the process is terminated. If not reached, the seed sentence selection unit 84 selects the next seed sentence from the corpus storage unit 82 and gives it to the problem position selection unit 86.

  Hereinafter, the seed sentence selection unit 84, the problem position selection unit 86, the option candidate generation processing unit 88, the validity determination unit 90, the validity determination data search unit 92, the shaping unit 94, the related information addition processing unit 96, and the output unit 98. The already-described option storage unit 100 repeats the above-described operation until a specified number of questions are created.

The validity determination described above is based on whether or not a word string consisting of a word string up to the content word before and after the empty space and an option exists on the web. However, the present invention is not limited to such an embodiment. In addition to this, a method of searching for a sentence in which a blank portion of the original sentence is replaced with the option candidate A _{i, m} on the web and determining that it is a valid sentence if there is a sentence including the option candidate A _{i, m} A method of determining that all N-grams (for example, trigrams) constituting a sentence are valid if found on the web, and a sentence likelihood including option candidates A _{i, m} using the N-gram likelihood obtained from the web There is a method of calculating and determining an appropriate sentence if it exceeds a predetermined threshold. In the last method, the frequency of the web N-gram is replaced with the number of documents including the word string that can be searched by the search system, that is, the document frequency df, instead of the normal word string frequency. For example, bigram _w 1 the document frequency of _{w 2 df (w 1 w 2} ), when the document frequency of the unigram _{w 1} and df _(w 1), bigram document likelihood P _(w 1 _{w 2)} is df (w _{1 w 2) / df (w} 1) and that Do not.

<Experiment>
The effectiveness of the multiple-choice language test question creation system 60 according to the above-described embodiment was verified by experiments. In the experiment, a corpus of two domains of travel conversation and business conversation was prepared as a corpus group 66, and the results obtained by letting multiple examinees solve the problem obtained by the multiple choice language test question creation system 60, and each test The effectiveness of the multiple-choice language test question creation system 60 was verified by examining whether or not the TOEIC (registered trademark) score of the person was correlated.

  The following options were used in the experiment. Select one of the above corpora, and select the seed sentence in order. The problem is a void filling problem. The vacant position is the leftmost verb in the seed sentence. As an option candidate, a word having the same meaning code as the word at the problem position is selected from the thesaurus 124. In the validity determination, the document frequency (the number of documents including the word string that can be searched by the validity determination data search unit 92) is obtained on the web using the content words before and after the option candidate as a key. As an option.

  As a seed sentence, sentences having a morpheme number of less than 6 or 21 or more were excluded. Also, words with a frequency of less than 9 were excluded from the choice candidates. The word string adopted as an option in the previous problem was excluded from the option candidates in another problem. As the validity determination data search unit 92, GOOGLE (registered trademark) was used.

Candidates who have a TOEIC score should solve the problem created in this way, and examinee by item response theory (refer to textbooks such as Kenji Otomo, “Introduction to Item Response Theory”, Daishukan Shoten, 1996). English ability (θ) was estimated. The horizontal axis TOEIC score candidates, to that shown by the estimated English language candidates θ to the vertical axis is the graph of Figure 12. As shown in FIG. 12, the correlation between the two was as high as 82%. From this, it is verified that the multiple choice question / vacancy replacement problem automatically created by the multiple choice language test question creation system 60 according to the above-described embodiment is effective in measuring the English ability of the examinee. did it.

  In this experiment, a 20,000-question multiple choice language test question was created in one personal computer in about an hour. Traditionally, institutions that administer English proficiency tests have taken several years to create and maintain questions. Moreover, the number of problems prepared by such institutions is around 4,000. Therefore, it can be seen that the multiple-choice language test question creating system 60 according to the present embodiment is very effective in creating a question.

[Second Embodiment]
In the first embodiment, a monolingual corpus is used with a seed sentence as a source. However, the present invention is not limited to the one using a monolingual corpus. For example, as in the second embodiment, a multiple-choice language test question can be created using a bilingual corpus in which parallel translations are collected. If a bilingual corpus is used as a seed source, a multiple-choice language test problem in which the original text acts as a constraint can be realized.

<Configuration>
FIG. 13 is a block diagram showing the configuration of the multiple-choice language test question creation system 220 according to the present embodiment. Referring to FIG. 13, a multiple-choice language test question creation system 220 includes a condition selection unit 222 similar to the condition selection unit 62 used in the first embodiment, and a bilingual corpus group including a plurality of bilingual corpora. 234, a plurality of machine translation devices 226, various data storage units 228 for storing various data as shown in FIG. 6, as well as various data storage units 68 used in the first embodiment, and condition selection A problem for automatically creating a multiple-choice language test question 232 connected to the unit 222, the plurality of machine translation devices 226, various data storage units 228, and the Internet 230 according to the conditions specified by the condition selection unit 222 Creating device 224.

  The various data storage unit 228 stores the same data as the various data storage unit 68 shown in FIG. Therefore, the description thereof will not be repeated here.

  In the present embodiment, the plurality of machine translation devices 226 are made up of different types of machine translation devices. The machine translation device 226 is used to create a sentence that becomes an error option, as will be described later.

  The problem creation device 224 uses a corpus selection unit 240 for selecting one bilingual corpus from the bilingual corpus group 234 according to an instruction given by the user using the condition selection unit 222, and the bilingual corpus selected by the corpus selection unit 240. A bilingual corpus storage unit 242 is stored. As this bilingual corpus, generally, a bilingual corpus is selected that includes a large number of parallel translations of a sentence in the native language of the examinee and a sentence in the language to be examined. For example, when an English test is imposed on a Japanese person, a bilingual corpus including many Japanese and English translations is used. In this embodiment, a bilingual corpus of Japanese and English is used. In this case, a plurality of machine translation devices 226 are used that translate from Japanese to English.

  The question creation device 224 is further selected from a bilingual corpus stored in the bilingual corpus storage unit 242 by a seed sentence selection unit 244 for selecting a parallel translation including a seed sentence for creating a question, and a seed sentence selection unit 244. The English sentence of the translated text includes a problem position selection unit 246 for selecting a space in the space replacement problem, an option position in the error detection problem, and the like. As will be described later, in the present embodiment, a problem form in which the original Japanese is presented and the correct one is selected from a plurality of translated sentences is also possible. When creating such a problem, the problem position selection unit 246 collectively sets the entire English sentence as the problem position.

  The question creation device 224 further receives a parallel translation including a seed sentence, and creates a choice candidate generation unit 248 for creating a choice candidate according to a question format specified by the user using the condition selection unit 222, and a choice candidate generation process. A validity determination unit 250 for determining validity based on a result of searching the Internet 230 for the option candidates generated by the unit 248, leaving only appropriate ones as problems and deleting inappropriate ones; A validity determination data search unit 252 for searching the Internet 230 in response to a request from the sex determination unit 250 and returning the result to the validity determination unit 250 is included.

  The option candidate generation processing performed by the option candidate generation processing unit 248 differs depending on the format of the multiple choice language test question to be created. FIG. 14 shows an example of a multiple choice language test question that can be created by the question creation device 224 according to the present embodiment. Referring to FIG. 14, problems that can be created by the problem creating device 224 are selected from the options in which a Japanese original sentence is given and an appropriate word is used in the corresponding English sentence space as in question 270. The question to choose, the problem of selecting the appropriate English translation corresponding to the Japanese sentence as shown in question 272, the problem of selecting the appropriate English translation from the four options, the problem of selecting the corresponding Japanese translation as given in Japanese There are a problem of selecting an appropriate word to be filled in the empty space, a problem of giving a Japanese compound word as in question 276, and selecting a correct translation from the choices, and the like. The creation method for the empty space replenishment problem is basically the same as that of the first embodiment. In the case of the problem 272, an English sentence generated by giving the original sentence to the plurality of machine translation devices 226 described above can be prepared as an incorrect answer. In the case of the problem 276, each element word of the Japanese compound word can be translated using a dictionary in the various data storage unit 228, and an erroneous answer can be generated by combining these translated words.

  In any of these cases, since the Japanese sentence is included in the seed sentence selected by the seed sentence selecting unit 244, it is possible to attach the Japanese sentence to the question sentence as shown in FIG. In this case, for example, in the empty space replenishment problem, the examinee considers not only English but also Japanese as a constraint condition and answers it. Therefore, it is easy to answer even if the seed sentence is short, and there is an effect that even a short sentence can be adopted as a seed sentence.

  Referring again to FIG. 13, the problem creating device 224 is further shaped by the shaping unit 254 for shaping what is determined to be appropriate as a problem by the validity judgment unit 250 according to the problem format, and the shaping unit 254. The related information addition processing unit 256 for adding information such as examples retrieved from the Internet 230, meanings stored in various data storage units 228, and usage information, and the related information addition processing unit 256 An output unit 258 for writing the question marked with a multiple choice language test question 232 to an external storage device, and an existing option storage unit 260 for storing the options in the problem output by the output unit 258; including.

<Operation>
The multiple choice language test question creation system 220 operates as follows. For the empty space supplementation problem, the operation of the multiple-choice language test question creation system 220 performs the same operation as that of the multiple-choice language test question creation system 60 according to the first embodiment. The difference is that the selected seed sentence consists of a parallel translation of a Japanese sentence and an English sentence, that even a short sentence can be used as a seed sentence, a Japanese sentence is added to the problem sentence as a constraint condition, etc. is there. Therefore, hereinafter, the operation of the multiple-choice language test question creation system 220 when creating the question 272 shown in FIG. 14 will be described.

  Assume that the condition selection unit 222 is designated to create a multiple-choice language test question such as the question 272. It is assumed that the corpus selected by the corpus selection unit 240 from the bilingual corpus group 234 according to a user instruction is stored in the bilingual corpus storage unit 242 in advance. In the following description, this corpus is assumed to be a bilingual corpus of Japanese and English.

  The seed sentence selection unit 244 selects a pair of parallel translations as a seed sentence from the corpus stored in the bilingual corpus storage unit 242. At the time of this selection, both the Japanese sentence and the English sentence are considered as the length of the seed sentence. However, the selected seed sentence may be shorter than in the case of the first embodiment.

  In the case of a problem of selecting an appropriate sentence from a plurality of sentences like the problem 272, the problem position selection unit 246 selects the entire seed sentence as the problem position and gives it to the option candidate generation processing unit 248.

  The option candidate generation processing unit 248 gives a Japanese sentence to the machine translation device 226 among the parallel translations of the given seed sentence, and causes the machine translation apparatus 226 to perform machine translation. For example, consider a case where the seed sentence is a translation of the sentence “My favorite dessert is ice cream.” And “My favorite dessert is ice cream.” The option candidate generation processing unit 248 gives this Japanese sentence “favorite dessert is ice cream” to each of the plurality of machine translation devices 226 and receives the result. The option candidate generation processing unit 248 gives the received machine translation result and the parallel translation of the seed sentence to the validity determination unit 250.

  The validity determination unit 250 uses each of the machine translation results as option candidates, performs an Internet search using the validity determination data search unit 252, and performs a non-sentence in the same manner as in the first embodiment. It is determined whether or not. Discard non-sentence candidates and leave only non-sentence candidates. If the number of non-sentence candidates is 2 or less, the number of options is insufficient. Therefore, the option candidate generation processing unit 248 stops the problem creation related to the seed sentence being processed, and instructs the seed sentence selection unit 244 to select the next seed sentence. When the number of candidates for non-sentences is 3 or more, the option candidate generation processing unit 248 selects only three sentences and gives them to the shaping unit 254 together with the parallel translation of the seed sentences. At this time, the validity determination unit 250 instructs the seed sentence selection unit 244 to select the next seed sentence. In response to this, the seed sentence selection unit 244 selects the next seed sentence, and creation of the next question is started.

  The shaping unit 254 that has received the option candidate and the parallel translation of the seed sentence from the validity determination unit 250 shapes the problem into a format like the problem 272 shown in FIG. At this time, the shaping unit 254 merges the English sentence of the seed sentence and the option candidate, and places them in any one of a), b), c), and d) according to a random number.

  The related information addition processing unit 256 gives information such as examples retrieved from the Internet 230, meanings, synonyms, antonyms, and related words stored in various data storage units 228 to the problem thus shaped. This is given to the output unit 258.

  The output unit 258 outputs this problem to the external storage device as a multiple-choice language test question 232. The output unit 258 also stores the words in the question sentence to be output in the already-existing option storage unit 260.

  The above process is repeated, and when the number of problems specified by the user using the condition selection unit 222 has been created, the process ends.

  As described above, according to the multiple-choice language test question creation system 220 of this embodiment, since a bilingual corpus is used, a problem using the original text in the corpus as a constraint can be created. Therefore, the examinee can select an answer even if the question sentence is short. As a result, it is possible to create many types of multiple choice language test questions using the corpus more effectively.

  Also, in both the first embodiment and the second embodiment, by preparing a corpus having various attributes and selecting a seed sentence from the corpus, a domain to be tested (travel conversation, business) Conversation, news, etc.) and the level of the target examinee (junior high school, high school, university, etc., or the rank of various examinations). For example, by preparing a corpus that includes sentences (and their translations) that appear in a specific unit of a specific textbook, the learner assumed by that textbook (for example, junior high school 1st year, high school 3rd year, business English students, etc.) The multiple choice language test questions suitable for the learning level can be created easily and in a short time.

  Furthermore, by preparing a corpus, a dictionary, a thesaurus, and the like for each language, a multiple-choice language test question can be easily created for a plurality of languages. In addition, by changing the combination of languages in the bilingual dictionary, various combinations of the examinee's mother tongue and the language to be tested can be easily realized.

  Also, by searching the Internet, dictionary, thesaurus, etc., it is possible to add appropriate examples, meanings, synonyms, synonyms, antonyms, etc. as related information to options that become incorrect answers, so you can simply solve the problem You can create a problem that can be expected to have an educational effect.

  In the above-described embodiment, for example, a case where a multiple-choice language test question is provided using a narrow screen such as a mobile phone is assumed, and therefore, the problem sentence is prevented from becoming long. However, this restriction is not essential. For example, when the space for presenting the question sentence is not particularly restricted, it is not necessary to restrict the length of the question sentence.

  In the above-described embodiment, when checking the validity of the option candidates, a web page on the Internet is searched and the result is used. Since a web page can be considered as an extremely large corpus, the appearance frequency of a linguistic expression composed of a certain word string can approximate the validity of the linguistic expression. In other words, if the frequency is high, it is an indication that the linguistic expression is very often used. If the frequency is very low or zero, it means that the linguistic expression is ungrammatical or unnatural. You can expect to be there. Therefore, it can be expected that the validity check is performed with high accuracy by using the Internet as described above.

  Of course, though it is grammatical, the new expression may not be on any web page, and there is no guarantee that all the correct expressions are on any web page. Therefore, it should be understood that the validity check described above is only approximate. Thus, a large corpus of correct sentences can be used instead of the web.

[Realization by computer]
Each of the problem creation device 64 according to the first embodiment and the problem creation device 224 according to the second embodiment includes a computer system, a computer program executed by the computer system, This can be realized by data stored in a storage device in the computer system or an external storage device. FIG. 15 shows the external appearance of the computer system 330, and FIG. 16 shows the internal configuration of the computer system 330.

  Referring to FIG. 15, this computer system 330 includes a computer 340 having an FD (flexible disk) drive 352 and a CD-ROM (compact disk read only memory) drive 350, and a keyboard 346 that is connected to the computer 340. A mouse 348 and a monitor 342.

  Referring to FIG. 16, in addition to FD drive 352 and CD-ROM drive 350, computer 340 includes CPU (central processing unit) 356 and bus 366 connected to CPU 356, FD drive 352 and CD-ROM drive 350. And a read only memory (ROM) 358 for storing a boot-up program and the like, and a random access memory (RAM) 360 connected to the bus 366 for storing a program command, a system program, work data, and the like. Computer system 330 further includes a printer 344.

  Although not shown here, the computer 340 may further include a network adapter board that provides a connection to a local area network (LAN).

  A computer program for causing the computer system 330 to operate as the problem creation device 64 or the problem creation device 224 is stored in a CD-ROM 362 or FD 364 inserted in the CD-ROM drive 350 or FD drive 352, and Transferred to the hard disk 354. Alternatively, the program may be transmitted to the computer 340 through a network (not shown) and stored in the hard disk 354. The program is loaded into the RAM 360 when executed. The program may be loaded directly into the RAM 360 from the CD-ROM 362, from the FD 364, or via a network.

  This program includes a plurality of instructions that cause the computer 340 to operate as the problem creation device 64 or the problem creation device 224 described above. Some of the basic functions required for this are provided by operating system (OS) or third party programs running on the computer 340 or various toolkit modules installed on the computer 340. Therefore, this program does not necessarily include all functions necessary for realizing the system and method of this embodiment. This program calls out an appropriate function or “tool” in a controlled manner so as to obtain a desired result, thereby realizing the operation as the problem creating device 64 or 224 described above. Only need to be included. The operation of computer system 330 is well known and will not be repeated here.

  In the above description of the embodiment, for example, as shown in FIG. 7, the main part of the program is completely different depending on the type of option problem to be created. However, the present invention is not limited to such an embodiment, and the same program code may be executed in portions that execute common processing. In the above description, a flowchart is used for convenience of description of the control structure. However, an actual program does not necessarily have the control structure represented by such a flowchart. For example, like an object-oriented program, each object may operate separately, and the functions of the apparatus of the above-described embodiment as a whole may be realized.

  Furthermore, in the above-described embodiment, the corpus, the condition selection unit, each unit of the problem creation device, and the storage device that stores the multiple-choice language test questions are all assumed to be in the same place. However, the present invention is not limited to such an embodiment. For example, the problem creation device may be realized by a computer connected to the Internet, and the condition selection unit may be realized by an information terminal device such as a mobile phone that can be connected to the computer through a wireless telephone network and the Internet. Also, do not accumulate multiple-choice language test questions, create a multiple-choice language test question using the question creation device each time a question creation request is made via the condition selection unit, and send it back to the request sender A simple configuration may be adopted.

  The embodiment disclosed herein is merely an example, and the present invention is not limited to the above-described embodiment. The scope of the present invention is indicated by each of the claims after taking into account the description of the detailed description of the invention, and all modifications within the meaning and scope equivalent to the wording described therein are intended. Including.

It is a figure which shows the example of an empty space replenishment problem. It is a figure which shows typically the sentence group produced | generated from the problem obtained in the middle of question preparation, and the solution candidate. It is a figure which shows the example of an error detection problem. 1 is a block diagram of a multiple-choice language test question creation system 60 according to a first embodiment. FIG. 3 is a diagram illustrating an example of a configuration of a corpus 110 stored in a corpus storage unit 82. FIG. It is a figure which shows an example of a structure of the various data memorize | stored in the various data storage part. It is a flowchart which shows the control structure of the computer program which implement | achieves the multiple choice language test question creation system 60 which concerns on 1st Embodiment. It is a flowchart which shows the control structure of the computer program which produces an empty space replenishment problem. It is a flowchart of a seed sentence selection process. It is a flowchart of an option candidate creation process. It is a flowchart which shows the control structure of the computer program which produces an error detection problem. It is a graph which shows the correlation with the candidate's English ability ((theta)) obtained by making a candidate solve the problem obtained by experiment, and a candidate's TOEIC score. It is a block diagram of the multiple choice language test question creation system 220 concerning the 2nd Embodiment of this invention. It is a figure which shows typically the various forms of the multiple choice language test question which can be created in 2nd Embodiment. 1 is an external view of a computer system. FIG. 16 is a block diagram of the computer shown in FIG. 15.

Explanation of symbols

  60,220 Multiple Choice Language Test Question Creation System, 62,222 Condition Selection Unit, 64,224 Question Creation Device, 66 Corpus Group, 68,228 Various Data Storage Units, 70,230 Internet, 80 Corpus Selection Unit, 82 Corpus Storage unit, 84, 244 Seed sentence selection unit, 86, 246 Problem position selection unit, 88, 248 Option candidate generation processing unit, 90, 250 Validity determination unit, 92, 252 Validity determination data search unit, 94, 254 Shaping unit, 96, 256 related information adding processing unit, 98, 258 output unit, 100, 260 existing option storage unit, 242 bilingual corpus storage unit

Claims (10)

Corpus storage means for storing a computer-readable corpus;
Computer-readable linguistic information including word dictionary, thesaurus, list of errors related to the word, or a list of words listed with the frequency and importance of the word, or any combination thereof, for making choices Language information storage means for storing in a form;
Sentence selection means for selecting a sentence that is a source of a multiple choice language test question from the corpus stored in the corpus storage means according to a predetermined selection method;
Among the constituent elements constituting the sentence selected by the sentence selecting means, the constituent element determining means for determining a constituent element that is a target location for multiple selection according to a predetermined condition;
Of the language information stored in the language information storage unit, a given condition is satisfied for the component determined by the component determination unit based on language information related to the determined component A replacement candidate creation means for creating a replacement candidate;
In the selected sentence, the determined components included in the sentence which is obtained by replacing the replacement candidate, the components string including the replacement candidate, for Oite search in a given corpus Search means;
In response to a search result by the searching means, wherein for each of replacement candidate, to calculate a statistical measure of the appearance frequency definitive in said predetermined corpus of the component string based on the search result by the searching means Means for calculating
Among the replacement candidates, multiple-choice selection is performed by combining a replacement candidate whose statistical measure calculated by the calculating means is smaller than a predetermined threshold and the original sentence in a predetermined form. An automatic multiple-choice language test question creation program that causes a computer to function as a question creation means for creating a language test question. The replacement candidate creating means includes:
Means for creating a plurality of replacement candidates for the determined component based on language information related to the determined component of the language information stored in the language information storage unit ,
The problem creating means is:
Means for creating a question sentence in which the determined component of the selected sentence is replaced with a space;
Among the plurality of replacement candidates, an option set including a predetermined number of replacement candidates in which the statistical measure calculated by the calculation unit is smaller than the threshold and the determined component is created. Choice creation means to
The multiple-choice language test question automatic creation program according to claim 1, further comprising: shaping means for shaping the question sentence and the option set into a predetermined question form. The problem creating means is:
Among the selected sentence, the determined component, the statistical measure that is calculated by the calculation means to create a replacement to issue statements in small intention determination has been substituted candidates than the threshold value Means for
2. The shaping unit for shaping the question sentence so as to display the replacement candidate and a predetermined number of words or word groups other than the replacement candidate as option candidates. Multiple-choice language test question automatic creation program described. The multiple-choice selection according to claim 1, further comprising a computer functioning as a means for receiving a user's input designating the given condition and providing the designated given condition to the replacement candidate creating means. Language test question automatic creation program. Further, the computer functions as a corpus reading means for reading a corpus designated by a user from a computer-readable corpus group storing means for storing a corpus group consisting of a plurality of corpora and storing it in the corpus storing means. The program for automatically creating multiple-choice language test questions according to claim 1. The corpus is a bilingual corpus including a plurality of parallel translations of sentences in a first language and sentences in a second language,
The sentence selection means includes a parallel translation selection means for selecting a parallel translation that is a source of a multiple-choice language test question according to the selection method from the bilingual corpus stored in the corpus storage means. Multiple-choice language test question automatic creation program described. The component determining means includes means for determining a component that satisfies the predetermined condition among the components of the sentence in the second language of the selected parallel translation,
The replacement candidate creating means includes:
Means for creating a plurality of replacement candidates for the second language for the determined component based on the language information stored in the language information storage means;
The problem creating means is:
Creating a question sentence including the sentence in the first language of the selected parallel translation and a sentence in which the determined component is replaced with a blank space among the sentences in the second language of the selected parallel translation Means for
Among the plurality of replacement candidates, the statistical measure that is calculated and replacement candidate of the second language a predetermined number of predetermined intended is determined smaller than the threshold value, which is the determined by the calculating means Choice creation means for creating a choice set consisting of components;
The multiple-choice language test question automatic creation program according to claim 6, comprising shaping means for shaping the question sentence and the option set into a predetermined question format. The problem creating means is:
Of the sentence in the first language of the selected parallel translation and the sentence in the second language of the selected parallel translation, the statistical measure calculated by the calculating means is the statistical measure calculated by the calculating means. means for creating a problem statement including a statement obtained by substituting at small the Most the determined replacement candidate than the threshold value,
Shaping means for shaping the question sentence so as to display the replacement candidate and a predetermined number of words or word groups other than the replacement candidate among the sentences in the second language of the question sentence as option candidates; The multiple-choice language test question automatic creation program according to claim 6, comprising: The component determining means includes means for determining the entire sentence in the second language of the selected parallel translation as a component satisfying the predetermined condition,
The replacement candidate creating means includes a plurality of machine translation means for machine translating the sentence in the first language of the selected parallel translation based on language information stored in the language information storage means,
The problem creating means is:
The second language sentence of the selected translation, and of the translation result by the plurality of machine translation unit, wherein the statistical measure that is calculated by the calculating means is the Most determined smaller than the threshold value statements Means for creating a set of choices comprising:
The multi-selection language test problem according to claim 6, comprising means for creating a multiple choice language test question by combining the sentence of the first language of the selected parallel translation and the set of choices in a predetermined form. Automatic creation program for limb choice language test questions. Control for creating a predetermined number of multiple-choice language test questions by controlling the sentence selection means, the constituent element determination means, the replacement candidate creation means, the search means , the calculation means , and the question creation means The multiple-choice language test question automatic creation program according to any one of claims 1 to 9, further causing a computer to function as a means.

Images