JP2008032953A - Language acquisition supporting device and program for the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a language acquisition supporting device and a program for the same enabling a user to feel as if the user is playing against a human by using a making-word-chains game where a computer sometimes loses in a comparatively natural manner. <P>SOLUTION: The language acquisition supporting device A1 and the program for the same are provided in which: candidate words having, as the character at the beginning of the word, the character at the end of a presented word whose input is received are extracted from a prescribed word dictionary D; probability of remembering is imparted to these candidate words; any one of the candidate words is output as an answer word at the probability of remembering or the probability imparted by a function using the probability of remembering. When none of the candidate words can be output or when a candidate extracting section cannot extract the candidate words, the information indicating the defeat is output. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a language acquisition support apparatus and a program using a serial game.

  Ciritri play, which everyone has become accustomed to since childhood, exists in many countries and has the aspect of simply enjoying the fun of a game. Has the educational aspect of helping other languages learn, and the academic aspect of being a subject of research on the theme of human-machine communication. In recent years, for example, serial game software has been developed for the purpose of assisting language acquisition, which is an educational aspect.

The inventor of the present application has also already performed a language acquisition support apparatus using a ritual game and an invention related to the program (see Patent Document 1). In the invention according to this document, based on the ending character of the word input by the user, the ending character is used as the initial character, and the candidate word excluding the already accepted input or the already output word is automatically extracted from the predetermined word dictionary. , Calculating the preference degree indicating the degree of advantage in the citation that depends on the character for each candidate word based on the predetermined quantitative statistical data for winning the citation for each character included in the word dictionary, and this calculation A method is adopted in which an answer word for determining a candidate word that can be an answer word is determined based on the preference level and then output.
JP 2004-077995 A

  By the way, in the above-mentioned patent document and other serial game software, when a human user and a machine computer (language acquisition support device) face each other, the computer usually wins. This is because the number of vocabularies registered in a vocabulary database or dictionary that can be used by a computer is generally larger than the number of vocabularies that a user can remember and remember. However, in the game where the user always loses, the enjoyment is reduced, and eventually the user loses interest, and the educational purpose of learning the user's language cannot be achieved as well as the fun of the game. . Therefore, it may be possible to adjust the program of the game software so that the computer loses to the user, but if the computer loses in an unnatural way, the user will wake up instead. It can be.

  Therefore, according to the present invention, in the same program as the language acquisition support apparatus using the ritual game, the computer can be defeated relatively naturally so that the user can feel as if he is playing against humans. The main purpose is to do.

  That is, the language acquisition support apparatus according to the present invention is a language acquisition support apparatus A1 based on a rule relating to a predetermined citrine that outputs an answer word having the initial letter of the input presentation word as an initial character. As shown in the functional block diagram, an input receiving unit 1 that receives an input of a presentation word, and a candidate word that uses the ending character as a prefix based on the ending character of the presentation word received by the input receiving unit 1 are predetermined. A candidate word extraction unit 2 that is extracted from the word dictionary D, a recall probability assigning unit 3 that provides a recall probability for the candidate word extracted by the candidate word extraction unit 2, and a recall probability that is provided by the recall probability provision unit 3 Or if any one of the candidate words is output as an answer word with a probability given by a function using the recall probability or if all candidate words cannot be output It is an essential feature by comprising, an output unit 4 for outputting information indicating defeated if it can not extract the candidate words in the candidate extraction unit.

  The present invention is based on the assumption that, as a premise, the ending character of the input presentation word is the initial character and, unlike the presentation word, an already accepted input word or an answer word excluding the already output word is output, A siritri game is performed with a computer, and through this, it is intended to be useful for user language learning and natural language education.

  In the present invention, first, the input receiving unit 1 receives a word input from a user. Word input by the user is performed by an appropriate input device such as a keyboard, a tablet, or a voice input device connected to the computer constituting the language acquisition support apparatus A1, or an input device of another computer connected to the computer through a network. The input receiving unit 1 stores the word input through an appropriate input interface as a presentation word. Depending on the rules of the ritual game, a word that has already been input or a word that has already been output may be excluded from a presentation word that receives input, or if such a word is input, the user loses (or language acquisition support device) (A1 win). Next, the candidate word extraction unit 2 extracts an initial character and a candidate word from a predetermined word dictionary based on the final character of the presented word that has been accepted. Again, depending on the rules of the ritual game, words that have already been input or words that have already been output may be excluded from candidate words extracted from the word dictionary D. Here, the word dictionary D may be provided in the language acquisition support device of the present invention, or may be used by the language acquisition support device A1 accessing and using the external word dictionary D.

  Subsequently, according to the present invention, the recall probability assigning unit 3 gives a probability of recalling the candidate word extracted by the candidate word extracting unit 2. Here, the “probability of remembering” is an expression that the language acquisition support device A1 that is a computer is regarded as a human who competes with a user in a siritri game. The language acquisition support device A1 provides “ In other words, the probability of answering. For example, when “probability to remember” is P and the sum of “probability to remember” and “probability that cannot be remembered” is 1, “probability that cannot be remembered” is obtained by the expression 1-P.

  In the present invention, the output unit 4 outputs any one of the candidate words as an answer word with the probability that the recall probability assigning unit 3 gives and recalls, or cannot output all candidate words with the recall probability Alternatively, if the candidate extraction unit cannot extract a candidate word, information indicating a loss is output. When the output unit 4 outputs the answer word, the user inputs a word having the initial letter of the answer word as the initial word, and if the user cannot input the next word, the user (Ie, the language acquisition support device A1 wins). On the other hand, if the output unit outputs a method indicating the loss, the language acquisition support device A1 loses (that is, the user wins).

  By the way, in the present invention described above, the recall probability given to each candidate word by the recall probability assigning unit 3 can be considered in various modes. For example, a predetermined constant value is fixedly set for all candidate words. Can be granted. In this case, all candidate words may be output as answer words with equal probability.

  Further, a value based on the appearance probability of each candidate word in a predetermined corpus C (see FIG. 1) in which a natural language sentence that can be used by the language acquisition support device A1 of the present invention is given to each candidate word as a probability May be. In this case, it may be a probability of recalling the appearance probability itself, or a probability of remembering a value obtained by multiplying the appearance probability by a predetermined constant. In particular, when using the corpus C as described above, there may be a plurality of the same words in the corpus C, but the same words may be “candidate words” as a single word or there are a plurality of the same words. The same word may be set as “candidate words” as different words. This is the same for the embodiment using the corpus C.

  In this case, for the appearance probability, a value based on the appearance frequency obtained from the ratio of the number of appearances of the candidate word to the total number of words in the corpus C is adopted, or the number of appearances of the presented word in the corpus C is determined. It is also possible to employ a value based on the association degree obtained from the ratio of the number of the presented word and the candidate word appearing simultaneously (that is, the number of the presented word and the candidate word co-occurd in the same sentence). In the former case, for example, the appearance frequency itself can be used as the appearance probability, or a value obtained by multiplying the appearance frequency by a predetermined constant can be used as the appearance probability. In the latter case, the association degree itself can be used as the appearance probability, A value obtained by multiplying by a predetermined constant can be used as the appearance probability.

  In addition, a value determined according to the number of appearances of the candidate word in a predetermined corpus C can be applied to the probability of remembering. That is, adopt a mode in which the appearance frequency itself or the value obtained by multiplying the appearance frequency by a predetermined constant is used as the probability of recalling, or the appearance frequency is provided with several stages and a predetermined recall probability is assigned to each stage. You can also.

  Thus, even if the probability given to the candidate word by various methods varies depending on the number of times the user inputs the word or the number of times the language acquisition support device outputs the answer word, the language acquisition support device is human. Process can be performed. That is, as an internal function element in the recall probability assigning unit 3, the candidate word is temporarily set according to the number of times the input word is input by the input receiving unit 1 or the number of times the answer word is output by the output unit 4. A configuration may be considered in which the recall probability re-assignment unit 31 (see FIG. 2, language acquisition support device A1 ′) is provided to adjust the given recall probability and re-assign the adjusted remember probability to the candidate word.

  In this case, the recall probability re-assignment unit 31 is once assigned to the candidate word as the number of times the input of the presented word is received by the input receiving unit 1 or the number of times the answer word is output by the output unit 4 increases. An example is one in which adjustment is performed so that the probability of recalling is reduced, and processing for re-assigning the recalled probability reduced by the adjustment to the candidate word is performed. This is because in the case of a siritri game played between humans, the next word as the siritri game progresses due to the effect that the human brain's processing power is used little by little in the task of remembering the word that was already used first Since the probability of recalling is considered to decrease, the process of the recall probability re-assignment unit 31 corresponds to such a phenomenon. When “recalling probability” is used in the processing of each unit described below, the “recalling probability” handled in the processing means that the recalling probability assigning unit 3 of the language acquisition support apparatus includes the recalling probability re-assigning unit 31. If it is, it means the “recall probability” reassigned by the recall probability re-assignment unit 31. If the recall probability re-assignment unit 3 does not include the recall probability re-assignment unit 31, “ It means “probability to remember”.

  Furthermore, it is useful in adjusting the winning or losing of the ritual game in which order a plurality of candidate words are provided to the output unit 4 and output with the recalling probability. For this purpose, the language acquisition further includes a candidate word arrangement order determination unit 5 that determines the arrangement order of the candidate words extracted by the candidate word extraction unit 3 to the configuration of the language acquisition support apparatus A1 of the present invention as described above. Whether or not the support device A2 is configured (see FIG. 3) and the output unit 4 can output the candidate words as the answer words with the probability of recalling the candidate words sequentially in the arrangement order determined by the candidate word arrangement order determination unit 5 It is preferable to judge.

  Here, various arrangement orders of candidate words by the candidate word arrangement order determination unit 5 can be considered. For example, the candidate word extracted by the candidate word extraction unit 2 may be randomly determined. Can be adopted.

  Further, for example, the candidate word arrangement order determination unit 5 performs predetermined calculation using the recall probability in the order of the recall probability given by the recall probability assigning unit 3 for the candidate words extracted by the candidate word extraction unit 4 The arrangement order may be determined in descending order of the result values.

  In addition, for the candidate words extracted by the candidate word extraction unit 4, the candidate word arrangement order determination unit 5 simultaneously displays the presented word and the candidate word for the number of appearances of the presented word in the predetermined corpus C. The order may be determined in descending order of the association degree obtained from the ratio of the numbers.

  In addition to this, the candidate word arrangement order determination unit 5 indicates the degree of advantage in the citrine obtained from the candidate words extracted by the candidate word extraction unit 4 based on a plurality of quantitative statistical data for winning the citri The order of arrangement may be determined in descending order of preference.

  Alternatively, for the candidate words extracted by the candidate word extraction unit 4, the candidate word arrangement order determination unit 5 counts the number of appearances of the presented word and the candidate word for the number of appearances of the presented word in a predetermined corpus C at the same time. The associative degree obtained from the ratio of the above and the preference degree indicating the degree of advantage in the citrine obtained based on a plurality of quantitative statistical data for winning the citrine are arranged in descending order of the value of the predetermined calculation result. It may be determined.

  Further, the candidate word extraction unit 2 excludes candidate words extracted from the word dictionary D from candidate words whose appearance frequency is less than a predetermined number in a predetermined corpus C in which natural language sentences are stored, or the corpus By providing a function of excluding candidate words whose association degree obtained from the ratio of the number of appearances of the presented word and the candidate word to the number of appearances of the presented word in C at the same time is less than a predetermined value, By narrowing down the number, it is possible to realize a state where the computer loses to the user in a natural way.

  In addition, in the output unit 4, instead of the function of the output unit 4 as described above, the priority of the output of the answer word and the output of the information indicating the loss is changed, and the probabilities of remembering given to each candidate word first. Is used to calculate the probability that the language acquisition support devices A1 and A2 lose, output information indicating the loss with the loss probability that is the result of the calculation, and output any information indicating that the loss is not output. Even if it has a function of outputting candidate words as answer words with the probability of remembering each assigned, it is possible to adjust the winning or losing.

  In addition to the above-described configuration of the present invention, the output according to one or both of the remaining number of candidate words or the losing probability calculated using the recalling probability given to each candidate word The language acquisition support device A3 (see FIG. 4) may further be provided, which further includes an output time adjusting unit 6 that adjusts the time for the unit 4 to output information indicating the answer word or losing. In this way, for example, the smaller the remaining number of candidate words or the higher the probability that the language acquisition support device A3 loses, the output unit 4 outputs either the answer word or the information indicating the loss. It is possible to create a state close to that the user is fighting against a human being in a clitoris game, such as by delaying the timing to perform.

  In addition to the above-described configuration of the present invention, after the output unit 4 outputs the answer word, the presentation word input time measurement unit 7 that measures the time until the input reception unit 1 receives the input of the presentation word next time. When the time measured by the presented word input time measuring unit 7 exceeds a predetermined time limit, the output unit 4 provides the language acquisition support device A4 (see FIG. 5). By using the device A4 to output information indicating winning, the use of the present invention with a sense of tension can be realized by limiting the input time of the words presented to the user.

  Further, the language acquisition support program according to the present invention causes a computer to perform information processing based on a rule relating to a predetermined citrine that outputs an answer word having a prefix of an input presentation word as an initial character. 1) an input receiving step for receiving an input of a presentation word, and 2) a prefix of the presentation word based on the suffix of the presentation word received in the input receiving step. A candidate word extraction step for extracting candidate words to be extracted from a predetermined word dictionary, and 3) a probability of recalling the candidate word extracted in the candidate word extraction step (= probability that the device can answer, constant value / individual value possible) And 4) a probability given by the recall probability given in the recall probability giving step or by a function using the recall probability. An output step of outputting any one of the candidate words as an answer word, or outputting information indicating losing when all candidate words cannot be output or when the candidate extraction unit cannot extract candidate words; Is a basic feature.

  According to the present invention, by introducing an algorithm that adjusts the winning or losing of the ritual game “recall probability” to a candidate word that is a candidate for the answer word for the presented word, the answer word cannot be output with the recall probability. Since the acquisition support device is “losing” in the siritri game, the computer will never continue to win, and the user will have a natural feeling as if they are fighting against humans. The device can be used.

  Embodiments of the present invention will be described below with reference to the drawings. Here, the language acquisition support apparatus A2 shown in FIG. 2 will be mainly described as an example of the embodiment of the present invention, but a language acquisition support apparatus having another configuration as appropriate is also referred to as needed.

  As a computer that realizes the language acquisition support apparatus A2 of the present embodiment, an ordinary personal computer is sufficient as an example. Therefore, in this embodiment, a general-purpose personal computer is adopted as the computer. Since the hardware configuration of such a personal computer is very general, it will not be described in detail, but usually, a processor such as a central processing unit (CPU), a main memory (RAM), an auxiliary storage such as a hard disk drive (HDD), etc. Devices and the like are provided in a mode of being connected by communication lines (bus lines or the like), and these operate under the control of a system controller, an I / O controller, or the like. Furthermore, the personal computer is a pointing device such as a keyboard and mouse as an input device, a display control device such as a display and graphic chip that displays information as an image or video as one form of an output device, and data exchange with an external device. It is assumed that a communication device is provided. Usually, the language acquisition support program of the present invention is installed in the HDD or the like in addition to the operation software (OS), and when the personal computer functions as the language acquisition support apparatus A2, the book read out to the RAM appropriately. Each unit is operated by processing of the CPU based on the program. Data generated in this processing step and data input or acquired from the outside are temporarily stored in a RAM or the like and used for subsequent processing. As an example, a general Japanese electronic word dictionary can be adopted as the word dictionary D, and a corpus C uses a database storing Japanese natural language sentences such as newspaper articles as an example. can do. Language information such as part of speech or syntactic structure may be given to words stored in the word dictionary D or the corpus C.

  As described above, the personal computer in which the language acquisition support program is installed functions as the language acquisition support apparatus A2 of the present embodiment, so that the input reception unit 1, the candidate word extraction unit 2, the recall probability assignment unit 3, The function of each part of the output part 4 and the candidate word arrangement | sequence order determination part 5 is show | played. In addition to the functions of each unit described above, the language acquisition support program includes rules for citrines that use Japanese words in this embodiment. For example, “n” is used as a suffix. When a word is input, or when a word that has already been input as a suggested word or output as a reply word is input as a new suggested word between the start and end of a single ritual game, When a presentation word not entered (for example, not registered in the word dictionary D) is input, the output unit 4 outputs information indicating either the user's loss or the language acquisition support device A2 wins. It is assumed that it is set to.

  The input receiving unit 1 is means for receiving and temporarily storing an input of a presentation word input by a user using an input device. In the input receiving unit 1, whether or not the ending character of the presentation word is “n”, whether or not the inputted presentation word has already been input as a presentation word or output as a reply word, the presentation word has a meaning. Checks whether a presentation word that conforms to the rules of the ritual game, such as whether or not it is a word that is not to be made, is accepted, and accepts the input of the presentation word only if it is a presentation word that complies with the rule. In this case, the output unit 4 can be set to output information indicating either the user's defeat or the language acquisition support device A2.

  The candidate word extracting unit 2 is a means for extracting a word having the ending character of the presented word received by the input receiving unit 1 from the word dictionary D and temporarily storing it. In addition, in the candidate word extraction part 2, it can set so that the word already input as a presentation word or the output as an answer word may be excluded, for example in one tribute game. Or, if input / output is allowed again even if it is the same as the presented word that has already been input or the answer word that has already been output, or if the same word has a different meaning or usage, it will be input / output again In such a case, such a word may be extracted as a candidate word. Or when extracting a candidate word from the word dictionary D, you may set so that only the word which appears in the said corpus C may be extracted with reference to the corpus C. In this case, the words appearing in the corpus C may be treated as the same word even if the same word appears multiple times, or all the words appearing with the same word appearing multiple times as different words. It may be handled by "B". Furthermore, among candidate words extracted from the word dictionary D, candidate words whose appearance frequency is less than a predetermined number in the corpus C are excluded, or the presented words and the candidates for the number of appearances of the presented words in the corpus The candidate word extraction unit 2 may be provided with a function of excluding candidate words whose association degree obtained from the ratio of the number of words appearing simultaneously does not reach a predetermined value. A method for calculating the appearance frequency and the association degree will be described later.

For each candidate word extracted by the candidate word extraction unit 2, the recall probability assigning unit 3
It is a means for giving a probability of remembering, in other words, a “probability of answering”. Here, when the sum of the “probability to remember” and the “probability that cannot be remembered” is 1 as the probability P to be remembered, the “probability that cannot be remembered” can be obtained by the operational formula <1-P>. Normally, if P = 0.01 or so, the balance between the user and the computer (language acquisition support device A2) is moderate, but if P = 0.02 or more, the computer is more likely to win. Tend to be. Various methods for determining the recall probability P can be considered. For example, any one of the following methods can be employed.

  <A> A uniform recall probability P is determined for all candidate words.

  <B> Let P be the appearance probability of each candidate word in the corpus C, or P be a value obtained by multiplying the appearance probability by a predetermined constant (P = appearance probability X constant). As for the appearance probability, the ratio of the number of appearances of each candidate word to the total number of words in the corpus C {(number of appearances of candidate words) / (total number of words)}, It can be determined by the ratio of the number of co-occurrence with each candidate word (appearing in the same sentence) {(number of co-occurrence of presented word and candidate word) / (number of appearance of presented word) = association degree}.

  <C> The number of appearances of each candidate word is divided in stages, and a probability P for remembering a different value is set for each stage. For example, P = 0.01 when the number of appearances is 100 or more, P = 0.099 when the number of appearances is 10 or more, P = 0.098 when the number of appearances is less than 10, and so on. Probability P is set stepwise. Note that the relationship between the number of appearances and the probability P to remember is merely an example, and the number of appearances can be replaced with the appearance frequency.

  The candidate word arrangement order determination unit 5 determines the arrangement order based on a predetermined algorithm for a plurality of candidate words extracted by the candidate word extraction unit 2 and given a probability P to be remembered by the recall probability assignment unit 3. The means for sequentially supplying candidate words to the output unit 4.

  Various methods for determining the arrangement order of the candidate words can be considered. For example, any of the following methods can be adopted.

  <a> Arrange candidate words in a random (random) order.

  <B> Candidate words are arranged in descending order of association degree Q (see <B> above).

  <C> Candidate words are arranged in descending order of preference R indicating the degree of advantage in the citrus obtained based on a plurality of quantitative statistical data for winning the citrine. Here, the “quantitative statistical data” includes, for example, 44 types excluding Japanese “n” from a word dictionary based on the “categorized vocabulary table (National Institute for Japanese Language, Shuei Publishing, 1964)”. A collection of quantitative statistical data on the characters (Hiragana) can be used. Specifically, data obtained by statistically counting a plurality of elements for winning a Japanese ritual game can be used as statistical data. For example, the elements include 1) the suffix of a candidate word The ratio of the number of words that have the end of the candidate word to the end of the number of words that have as the initial and that do not have the letter that means losing in the rules of the siritri as the end, 2) 3) When comparing the total number of words with the initial letter of the candidate word as the initial character, the number of words having the initial character of the candidate word as the initial character and the character “n” as the initial character 4) the number of words that have the ending characters of the candidate words as the initial and ending characters, 5) the number of types of ending characters of the words in the set of words having the ending characters of the candidate words as the ending characters, 6 ) Candidate The word endings character number of types of prefix characters thereof word in the set of words consisting of only two letters to ending character, it can be exemplified. The preference R is obtained by summing up values obtained by multiplying the values of the respective elements by a constant using one or more of the values of the elements 1) to 6), or a function of such a simple linear sum. Instead, it is obtained by a function that adds a second-order or higher-order term, a logarithmic term, a reciprocal term, or a combination thereof in a more complicated manner.

  <D> An arithmetic expression using the association degree Q and the preference degree R, for example, a value obtained by multiplying each of the association degree Q and the preference degree R by a constant, or a second order or more using the association degree Q and the preference degree R Candidate words are arranged in descending order of a value obtained by a function to which a term of logarithm, a logarithmic term, a reciprocal term, or a combination thereof is added.

  <E> Candidate words are arranged in descending order of the probability P of remembering. In this case, only those whose recall probability P is greater than or equal to a predetermined threshold are selected, only a predetermined number is selected among those whose recall probability P is greater than or equal to a predetermined threshold, or those whose recall probability P is greater than or equal to a predetermined threshold. Variations may be made in the selection of candidate words, such as obtaining a value obtained by multiplying a maximum value by a predetermined value and selecting only a value that is greater than or equal to the predetermined value. These threshold values and predetermined values may be determined in advance, or may be set and changed as appropriate by the user. Note that, in addition to the recall probability P, the threshold value or the like can be limited even when candidate words are arranged in descending order using the association degree Q and the preference degree R described above.

  <F> Candidate words are arranged in descending order of the value given by the function using the recall probability P. For example, the ratio of the recall probability P of each candidate word to the sum Pt of recall probabilities P of all candidate words remaining at that time is calculated (P / Pt), and the candidate words are arranged in descending order of the value.

  The output unit 4 determines whether or not the candidate words are sequentially recalled in the order determined by the candidate word arrangement order determination unit 5 or the probability obtained by a function using the recall probability P, or not. If the candidate word can be output, it is a means for outputting it as an answer word. The output unit 4 also indicates that the computer (language acquisition support device A2) has “lost” when all candidate words cannot be output or when the candidate word extraction unit 2 cannot extract any candidate words. The information shown is output. When the output unit 4 determines whether or not to output a candidate word, it is determined by the value of the recall probability P itself or by the probability given by the function using the recall probability P. Either of them can be adopted.

<I> Each candidate word is output based on the recall probability P. In this case, it outputs with the probability P in an order from the upper word of the candidate word for which the arrangement order was determined. For example, if there are n candidate words, the probability that each candidate word is remembered is represented by Pn, the probability that the first candidate word is output is P ₁ , and the probability that it is not output is (1−P ₁ ). Similarly, the probability that the nth candidate word is not output is given by the expression {1- (1-P ₁ ) (1-P ₂ )... (1-Pn)}. For example, if the probability P is remembered as a uniform value for all candidate words, the probability that the computer (language acquisition support device A2) will output any candidate word is the expression {1- (1-P) ⁿ }, and the probability of outputting information indicating a loss is given by the expression {(1-P) ⁿ }.

<II> Each candidate word is output based on a function using the recall probability P. For example, when a uniform recall probability P is given to all candidate words, it is output as an answer word with the probability P of remembering the first candidate word, and when the first candidate word is not output, it is the second place. Is output as an answer word by the expression {(1-P) P}. Thereafter, the probability of recalling the nth candidate word is output as an answer word by the expression {(1-P) ^n- 1P}. When the lowest candidate word is not output as an answer word, the output unit 4 outputs information indicating loss.

  Regardless of the output method (I) or (II) described above, if the output unit 4 outputs an answer word, the language acquisition support device A2 determines the next word presented by the user, that is, the answer word. The input of a word whose initial character is the initial character is waited, and the above-described processing is repeated until information indicating losing is output for either the user or the language acquisition support device A2.

  The flow of information processing by the language acquisition support apparatus A2 so far is summarized and shown in a simple flowchart as shown in FIG. Here, as a function of the output unit 4, a case of an aspect (above (I)) in which the candidate word is output with a probability P remembered is shown. That is, first, the input unit 1 accepts input of a presentation word (step S1), the candidate word extraction unit 2 extracts candidate words (n) from the word dictionary (step S2), and the recall probability assignment unit 3 extracts each candidate word. Is given a probability Pn to remember by any one of the methods <A> to <C> (step S3). Then, the candidate word arrangement order determining unit 5 determines the arrangement order of these candidate words by any one of the methods <a> to <f> (step S4), and the output unit 4 selects the higher order candidates. It is determined whether the nth candidate word can be output with probability Pn in order from the word (step S5). In this determination, if the candidate word can be output (S5; Yes), the candidate word is output as an answer word (step S6), and input of the next presentation word is waited (step S7). If there is an input of the next presentation word, the process returns to the step of accepting the input of the presentation word (step S1). On the other hand, when the candidate word cannot be output in the step of determining whether or not to output a candidate word (S5; No), it is further determined whether or not the number of remaining candidate words is 0 (step S8). If there are remaining candidate words (S8; No), the process returns to the step of determining whether or not the nth candidate word can be output with the probability Pn for the next candidate word (step S5). If the number of remaining candidate words is 0 (S8; Yes), information indicating that the computer (language acquisition support device A2) is losing is output (step S9), and the ritual game ends. If the input of the presentation word is not accepted in step S1, the user cannot input the next presentation word. Therefore, the output unit 4 loses the user or the computer (language acquisition support device A4) wins. Will be output.

  Further, when the output time adjustment unit 6 is provided as in the language acquisition support device A3 described above, for example, the following method is used when the number of remaining candidate words is small or the computer (language acquisition support device A3) When the probability of losing is high, the output time of the answer word by the output unit 4 can be delayed.

  B) The output of answer words is delayed as the number of remaining candidate words decreases. For example, a step of counting the number of remaining candidate words is provided between step S5 and step S8, and when the number of candidate words is less than a predetermined threshold, the output time of the answer word by the output unit 4 is a predetermined time (for example, (Several seconds) is added to delay the output time in step S6. Alternatively, for each predetermined number of the remaining number of candidate words, a predetermined table is provided so that the output time of the answer word is set longer as the number of candidate words decreases, and the output of the answer word in step S6 is performed according to the table. Determine the time.

B) The probability that the computer (language acquisition support device A3) loses is calculated by, for example, the expression {(1-P) ⁿ } or the expression {(1-P ₁ ) (1-P ₂ ) (1-P ₃ ) (1- P _n )} is provided, and the larger the value, the longer the output time of the answer word in step S6. Alternatively, for each predetermined value of a certain losing probability, a predetermined table is provided such that the longer the value is, the longer the response word output time is set, and the response word output time in step S6 is determined according to the table. To do.

  C) A predetermined function f (x) for determining the output time of the answer word is defined with the number of remaining candidate words and the probability of losing as a variable x, and the output time of the answer word in step S6 is defined by this f (x). decide.

  Furthermore, when the presentation word input time measurement unit 7 is provided as in the language acquisition support device A4 described above, the next time after the output unit 4 outputs the answer word between step S7 and step S1, A step of measuring a time until receiving an input of a presentation word and a step of determining whether or not the measurement time exceeds a certain time limit are provided. When the time limit is exceeded, the output unit 4 loses the user or Information indicating the winning of the computer (language acquisition support device A4) may be output.

In addition, in the output part 4 mentioned above, after trying whether it can output as an answer word about all the candidate words as mentioned above, when the answer word was not able to be output, the information which shows loss was output. On the contrary, when the probability of outputting information indicating the loss is obtained in advance, and the information indicating the loss cannot be output, the probability P of recalling any one candidate word or the probability P of recalling is used. It can be set to output with the probability given by the function. For example, when there are n candidate words and the probability of recalling each candidate word is Pn, the probability that the computer (language acquisition support device A2) outputs information indicating the loss is represented by the expression {(1-P ₁ ) ( 1-P ₂ ) (1-P ₃ ) (1 (1-Pn)) or its approximate expression {1-P ₁ -P ₂ -P ₃ -... -Pn}. If no information indicating that the loss is obtained with the probability obtained by these arithmetic expressions is output, any one candidate word is output as an answer word with the recall probability Pn. Alternatively, when information indicating loss is not output, any one candidate word is output at a ratio (Pn / Pt) of the recall probability Pn of the candidate word to the total probability Pt of recall of all candidate words. You may do it.

  In addition, by providing the recall probability re-assignment unit 31 as an internal function element of the recall probability assigning unit 3 as in the language acquisition support device A1 ′ described above, the value changes each time the probability P recalled to the candidate word is given. You may let them. Specifically, for example, the processing in the recall probability re-assignment unit 31 is performed once on the candidate word as the number of times the input word is input by the input receiving unit 1 or the number of times the answer word is output by the output unit 4 increases. It is possible to adjust so that the given remembering probability P is lowered, and to re-assign to each candidate word the remembering probability P ′ lowered by the adjustment. More specifically, as an adjustment for reducing the recall probability P, based on the number of times the input acceptance unit 1 has accepted the input of the presented word or the number of times the output unit 4 has output the answer word in the previous ritual game, An example is shown in which it is determined that it is the rth time that the probability P is given to the candidate word at the time, and the value obtained by multiplying the recall probability P by the value “1 / r” is used as the recall probability P ′. can do. In this way, it is possible to cause the computer to realize a human process that makes it difficult to remember the next word as the ritual game progresses. In addition to the above-described mode, the value “r” is the number of times the input word is received by the input receiving unit 1, the number of times the output word is output by the output unit 4, or a value obtained by adding 1 to the number of times. Various meanings can be given, such as the total number of times the input word is received by the input receiving unit 1 and the number of times the answer word is output by the output unit 4. In addition, the value “r” is limited to only words having the same initial letter, the number of times of giving the recall probability P for the word having the initial letter, and the number of times of receiving the presented word in the input receiving unit 1 itself. , The number of times the answer word is output by the output unit 4, the value obtained by adding 1 to the number of times, various values such as the number of times the input word is received by the input receiving unit 1 and the number of times the answer word is output by the output unit 4. Can also be given meaning.

  In addition, this invention is not limited to the above-mentioned embodiment, About the specific structure of each above-mentioned part and each process, it is not restricted to the said embodiment, A various deformation | transformation is possible in the range which does not deviate from the meaning of this invention.

The block diagram which shows an example of a function structure of the language acquisition assistance apparatus which concerns on this invention. The block diagram which shows an example of a function structure of the language acquisition assistance apparatus which concerns on this invention. The block diagram which shows an example of a function structure of the language acquisition assistance apparatus which concerns on this invention. The block diagram which shows an example of a function structure of the language acquisition assistance apparatus which concerns on this invention. The block diagram which shows an example of a function structure of the language acquisition assistance apparatus which concerns on this invention. The flowchart which shows an example of the flow of the information processing by the language acquisition assistance apparatus which concerns on one Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Input reception part 2 ... Candidate word extraction part 3 ... Recall probability provision part 31 ... Recall probability re-assignment part 4 ... Output part 5 ... Candidate word arrangement | sequence order determination part 6 ... Output time adjustment part 7 ... Presentation word input time measurement part A1, A1 ', A2, A3, A4 ... Language acquisition support device C ... Corpus D ... Word dictionary

Claims (19)

A language acquisition support device based on a rule relating to a predetermined siritri that outputs an answer word having a prefix of an input presentation word as an initial character,
An input receiving unit for receiving an input of the presented word;
A candidate word extraction unit for extracting from the predetermined word dictionary a candidate word whose initial character is the initial character based on the final character of the presented word received by the input reception unit;
A recalling probability giving unit for giving a probability of individually recalling candidate words extracted by the candidate word extracting unit;
When any one of the candidate words is output as an answer word with a probability given by the recall probability assigning unit or a probability given by a function using the recall probability, or not all candidate words can be outputted Alternatively, when the candidate extraction unit cannot extract a candidate word, an output unit that outputs information indicating a loss;
A language acquisition support apparatus comprising:   The language acquisition support apparatus according to claim 1, wherein the recalling probability is a predetermined constant value for all candidate words.   The language acquisition support apparatus according to claim 1, wherein the recall probability is a value based on an appearance probability of each candidate word in a predetermined corpus in which a natural language sentence is stored.   The language acquisition support apparatus according to claim 3, wherein the appearance probability is a value based on an appearance frequency obtained from a ratio of the number of appearances of the candidate words to the number of all words in the corpus.   4. The language acquisition support according to claim 3, wherein the appearance probability is a value based on an association degree obtained from a ratio of the number of appearances of the presented word and the candidate word to the number of appearances of the presented word in the corpus. apparatus.   The language acquisition support apparatus according to claim 1, wherein the recall probability is a value determined according to the number of appearances of the candidate word in a predetermined corpus in which a natural language sentence is stored.   The recall probability assigning unit adjusts the remembering probability once given to the candidate word according to the number of times the input of the presented word is accepted by the input accepting unit or the number of times the answer word is output by the output unit. The language acquisition support device according to claim 1, further comprising a recall probability re-assignment unit that re-assigns the adjusted recall probability to the candidate word.   The recall probability re-granting unit increases the number of times the input word is accepted by the input accepting unit or the number of times the output word is output by the output unit. The language acquisition support apparatus according to claim 7, wherein the processing is performed such that adjustment is performed so as to decrease, and the recall probability reduced by the adjustment is reassigned to the candidate word. A candidate word arrangement order determining unit that determines the arrangement order of the candidate words extracted by the candidate word extraction unit;
The language according to any one of claims 1 to 8, wherein the output unit determines whether or not candidate words can be output as the answer word with the probability of recalling them sequentially in the arrangement order determined by the candidate word arrangement order determination unit. Acquisition support device.   The language acquisition support device according to claim 9, wherein the candidate word arrangement order determination unit is configured to randomly determine the arrangement order of the candidate words extracted by the candidate word extraction unit.   The candidate word arrangement order determination unit, for the candidate words extracted by the candidate word extraction unit, in the order of the recall probability given by the recall probability giving unit, or a value of a predetermined calculation result using the recall probability The language acquisition support apparatus according to claim 9, wherein the order is determined in descending order.   The candidate word arrangement order determination unit is configured to simultaneously display the candidate word extracted from the candidate word extraction unit with respect to the number of appearances of the presented word in a predetermined corpus in which a natural language sentence is stored and the candidate word. The language acquisition support apparatus according to claim 9, wherein the order of arrangement is determined in descending order of the association degree obtained from the ratio of the number of appearances.   The candidate word arrangement order determining unit is arranged in descending order of preference indicating the degree of advantage in the citrine obtained based on a plurality of quantitative statistical data for winning the citri for the candidate words extracted by the candidate word extracting unit. The language acquisition support apparatus according to claim 9, wherein the arrangement order is determined.   The candidate word arrangement order determination unit is configured to simultaneously display the candidate word extracted from the candidate word extraction unit with respect to the number of appearances of the presented word in a predetermined corpus in which a natural language sentence is stored and the candidate word. The value of the predetermined calculation result using the association degree obtained from the ratio of the number of appearances and the preference degree indicating the degree of advantage in the citri to be obtained based on a plurality of quantitative statistical data for winning the citri The language acquisition support apparatus according to claim 9, wherein the arrangement order is determined.   The output unit outputs any one of the candidate words as an answer word with a probability given by the recall probability giving unit or a probability given by a function using the recall probability, or all candidates If the word cannot be output or if the candidate extraction unit cannot extract the candidate word, the language acquisition support is performed using the recall probability given to each candidate word instead of the function of outputting information indicating losing If the device calculates the probability of losing, outputs information indicating losing with the losing probability that is the result of this operation, and does not output information indicating the losing, the probability of remembering each candidate word given to each or 15. The language according to claim 1, which has a function of outputting as a reply word with a probability given by a function using the probability of remembering. Obtained support device.   The candidate word extraction unit excludes candidate words whose appearance frequency is less than a predetermined number in a predetermined corpus in which a natural language sentence is stored among candidate words extracted from the word dictionary, or the word in the corpus 16. The method according to any one of claims 1 to 15, which has a function of excluding candidate words whose association degree obtained from a ratio of the number of appearances of the presented word and the candidate word to the number of appearances of the presented word is less than a predetermined value. The language acquisition support apparatus according to the above.   In response to either or both of the remaining number of candidate words or the probability of losing the language acquisition support device calculated using the recall probability given to each candidate word, the output unit answers the word The language acquisition support apparatus according to claim 1, further comprising an output time adjustment unit that adjusts a time for outputting information indicating losing. After the output unit outputs the answer word, it further comprises a presentation word input time measurement unit that measures the time until the input reception unit receives the input of the presentation word next,
18. The output unit according to claim 1, wherein when the time measured by the presented word input time measurement unit exceeds a predetermined time limit, the output unit outputs information indicating that the language acquisition support device is winning. Language acquisition support device. A program for causing a computer to perform information processing based on a rule relating to a predetermined siritri that outputs an answer word having a prefix of an input presentation word as an initial character,
An input receiving process for receiving an input of the presented word;
A candidate word extracting step of extracting from the predetermined word dictionary a candidate word having the ending character as an initial character based on the ending character of the presented word received in the input receiving step;
A recall probability assigning step for giving a probability of individually recalling candidate words extracted in the candidate word extraction step;
When any one of the candidate words is output as an answer word with a probability given by the recall probability giving step or a probability given by a function using the recall probability, or not all candidate words can be outputted Or an output step for outputting information indicating a loss when a candidate word cannot be extracted in the candidate extraction step;
Language acquisition support program characterized by having

Images