JP6054039B2 - Vocabulary learning speed estimation apparatus, method, and program - Google Patents

Description

  The present invention relates to a vocabulary learning speed estimation device, method, and program, and more particularly, to a vocabulary learning speed estimation device, method, and program for estimating vocabulary learning speed of an infant.

  Although human language development can provide important scientific knowledge and suggestions for thinking about `` what is human beings '', there are many unsolved problems at present, so progress in measurement technology related to language development There is almost no commercial service development. In particular, there has been little progress in science and technology related to vocabulary acquisition, among basic abilities such as speech recognition, vocabulary acquisition, and grammar manipulation. However, technological development in this field is considered to be important in view of the need for education that moderately supports healthy development and the early detection and support of developmental disabilities including language development delays.

  One of the phenomena that is particularly characteristic and important in grasping individuality in the language development of young children is vocabulary learning speed. However, it has been thought that this vocabulary learning speed cannot be obtained without the phenomenon called vocabulary explosion (or vocabulary spurt). Vocabulary explosion is a phenomenon that development psychologists have been paying attention to since the middle of the 20th century, and refers to a sudden change in vocabulary learning speed that occurs in the second half of the year. Basically, toddlers begin to utter their first words around their first birthday, but for a while they learn words at a very moderate rate. However, after the age of one and a half years, the words suddenly start to utter, so the dramatic change has been called “explosion” and “spurt”. The vocabulary explosion is so dramatic that many parents consciously notice it, so it is well known not only in the field of psychology but also in the childcare industry. For this reason, it is thought that the vocabulary explosion must be modeled in order to quantify the state of child vocabulary development.

  Conventionally, in the field of developmental psychology, the phenomenon of vocabulary explosion has been confirmed in multiple languages using large-scale group data using a vocabulary checklist (questionnaire survey based on parents' answers). It has been found that when the average value of the group data is plotted for each age, it becomes a quadratic curve showing a gradual increase, and its inflection point appears around 18-20 months. From these collective data, we have regarded vocabulary explosion as a common phenomenon seen in many children.

  In order to estimate the vocabulary learning speed, it is important for vocabulary explosions to determine when vocabulary explosions occur for each individual and how to detect and estimate the vocabulary explosion time (when vocabulary explosion starts) However, the following four methods for estimating the vocabulary explosion period and estimating the speed of vocabulary learning before and after that have been proposed.

  The first is a visual method in which a graph is drawn without any particular calculation and is visually determined. The second is the 50-word achievement standard method that defines the time when 50 words are learned as the vocabulary explosion time. The third is a specific period achievement standard method in which a period when an achievement standard (for example, 30 words or more) is satisfied in a specific period (for example, three weeks) is set as a vocabulary explosion period. The fourth is a logistic regression approximation method in which the velocity component of the vocabulary acquisition data is approximated to a logistic regression equation, and the inflection point is the vocabulary explosion time (see Non-Patent Document 1).

  Up to now, these methods have been used to determine vocabulary explosion and to determine the vocabulary learning speed before and after that.

Ganger, J., & Brent, M. R. (2004). Reexamining the vocabulary spurt. Developmental Psychology, Vol. 40, No. 4, 621-632.

  However, the first method can confirm the presence or absence of a phenomenon to some extent, but has a problem that the vocabulary speed cannot be automatically determined. The second method is based on empirical data, but based on a very small sample of English-speaking middle classes, there is no guarantee that it will apply to various children in many cultural zones. There is a problem. Another problem is that no individual differences in vocabulary explosion are assumed.

  The third method can detect changes in vocabulary learning speed in a specific time range, but the unique and arbitrary achievement criteria setting takes into account the vocabulary learning speed between individuals. There is a problem that some individuals may misunderstand the vocabulary explosion time completely.

  The fourth method approximates the data for each individual, so it can handle even if the vocabulary learning speed between individuals is different, but it does not accurately capture the characteristics of infant vocabulary development. However, there is a problem that the approximation accuracy of logistic regression is low and the existence of the vocabulary explosion itself cannot often be confirmed. In addition, this method is based on the premise that there is abundant data after the vocabulary explosion, and there is also a problem that it cannot be applied unless the data up to 2 years old are available.

  In order to point out the problems of the fourth method, the vocabulary learning curves for infants are shown in FIGS. This shows the infant's age on the horizontal axis and the cumulative vocabulary acquired on the vertical axis. The fourth method is a complex in which a large discontinuity as shown in FIG. 14B and an S-shaped curve as shown in FIGS. 14C, 15A, and 16A appear. Vocabulary learning curves cannot be modeled. Also, a complicated curve as shown in FIG. 17C could not be modeled.

  As described above, the method that assumes the phenomenon of vocabulary explosion is extremely difficult to model, and it has been impossible to stably obtain the vocabulary learning speed peculiar to an infant using any method.

  In addition, since these methods assume a vocabulary explosion, modeling of vocabulary learning speeds that are completely different before and after the vocabulary explosion is performed, and feature quantities that represent infant characteristics over all periods are extracted. It is not.

  The present invention has been made in view of the above problems, and can calculate a vocabulary learning curve for an infant without assuming a vocabulary explosion time, and can stably estimate the vocabulary learning speed with high accuracy. It is an object to provide a learning speed estimation device, method, and program.

In order to achieve the above object, the device for calculating a function relating to vocabulary learning according to the present invention is the age at which an infant begins to speak a new word, and the infant speaks before the age. With respect to the time series of data indicating the relationship with the cumulative number of words, a portion where the age interval until the next new word is spoken is larger than a predetermined value within the range of 4 to 15, The replacement means for replacing each age of the data so that the interval of the ages is within a range of 1 to 3 smaller than the predetermined value, and the data replaced by the replacement means are used. It is configured to include a function approximation means for calculating a function that approximates the change in the time series of the data had as a function related vocabulary learning, the.

Further, a method of calculating a function relating vocabulary learning of the present invention, substituted means, and to a method of calculating a function relating to vocabulary learning in an apparatus for calculating a function relating to learning vocabulary containing function approximation hand stage, the vocabulary learning The function for calculating the function relating to the relationship between the age at which the infant began to speak a new word and the cumulative number of words that the infant has spoken up to the age by the replacement means. With respect to the time series of the data shown, for the portion where the age interval until the next new word is spoken is greater than a predetermined value within the range of 4 to 15, the age interval is the predetermined Replacing the age of each of the data so as to become a reference value within a range of 1 to 3 smaller than the value, and replaced by the replacement means by the function approximation means Calculating a function that approximates the change in the time series of the data using the serial data as a function relating vocabulary learning, to perform include.

  According to the present invention, the data indicating the relationship between the age at which the infant started to speak a new word and the cumulative number of words that the infant has spoken up to the age by the replacement means. With respect to the time series, for the part where the age interval until the next new word is spoken is larger than a predetermined value, the age interval becomes a reference value smaller than the predetermined value. Replace each age of data.

Then, the function approximation means, for calculating a function of time approximation the change in the sequence of the data using the data that have been replaced by the replacing means as a function related to learning vocabulary.

In this way, for each part where the age interval until the next new word starts to be spoken is larger than the predetermined value, the age of each data is replaced so that the age interval becomes the reference value. by approximate to Rukoto function, without assuming the vocabulary explosion timing can be calculated function approximating the transition of the time series of the data as a function relating to learning vocabulary.

Moreover, the program of this invention is a program for functioning a computer as each means which comprises the apparatus which calculates the function regarding said vocabulary learning.

As described above, according to the apparatus, method, and program for calculating a function relating to vocabulary learning of the present invention, a portion where the interval of days until the next new word starts to be spoken is a predetermined value or more. Assuming the vocabulary explosion period by replacing each age of the data so that the interval of days of age becomes the reference value, approximating with the function, and estimating the vocabulary learning speed based on the approximated function Therefore, there is obtained an effect that the true vocabulary learning curve of the infant can be calculated and the vocabulary learning speed can be stably estimated with high accuracy.

It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a block diagram which shows the functional structure of the vocabulary learning speed estimation apparatus of this Embodiment. It is a figure which shows an example of an input screen. It is a figure which shows an example of an input data set. It is a flowchart which shows the content of the vocabulary learning speed estimation processing routine in the vocabulary learning speed estimation apparatus of this Embodiment. It is a flowchart which shows the flow of the process which removes the plateau in the vocabulary learning speed estimation apparatus of this Embodiment. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve. It is a graph which shows an infant's vocabulary learning curve.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Outline of the invention>
First, the natural law regarding vocabulary learning discovered by the inventors will be explained. Inventors found in FIG. 14 to FIG. 17 that vocabulary learning curves frequently occur during a certain period (for example, 6 days or more) during which no word is learned. The part of the learning curve that does not remember one word for more than a certain period of time will be called a plateau (a technical term in learning psychology that means “plain”).

  Our model of vocabulary learning curve assumes that there is a true learning curve in vocabulary learning for infants, and this plateau is frequently inserted in the early stages of vocabulary learning, so the phenomenon of vocabulary explosion is observed. It is said that.

  In order to demonstrate this, the plateau portion was replaced by a certain period (for example, 1 day) (plateau removal) and the vocabulary learning curve was redrawn. As a result, it was found that the vocabulary learning curves were aligned. The inventors assume that this curve is a true vocabulary learning curve. The state is shown in FIGS. -Is the vocabulary learning curve shown in FIGS. 14 to 17, and + indicates the start of a plateau. ○ indicates the learning curve after plateau removal. The solid line represents a straight line that approximates the learning curve. When the plateau is removed in this way, the learning curve is almost a straight line, and its slope and intercept show characteristics specific to infants. The natural phenomena discovered here are summarized as follows.

(1) The plateau lasts from the beginning of vocabulary learning to the end of the period previously called vocabulary explosion, and then decreases in number and length.

(2) Excluding the plateau, the complicated vocabulary learning curve that seems to be due to the vocabulary explosion can be modeled by a simple curve close to a straight line.

  In the present invention, using this newly discovered natural phenomenon, a vocabulary learning curve unique to an infant that does not depend on vocabulary explosion is obtained, and the vocabulary learning speed is obtained from the curve.

<System configuration>
As shown in FIG. 9, the vocabulary learning speed estimation apparatus 10 according to the present embodiment includes an input unit 12 that receives input of various data, a calculation unit 14 that estimates a vocabulary learning speed, and an output that outputs an estimation result. Part 16.

  The input unit 12 is realized by an input device such as a known keyboard, mouse, or storage device, and receives input data.

  Here, in order to estimate an infant's vocabulary learning speed, what kind of data is referred to is a problem first. If it is possible to record all of the infant's utterances with electronic media such as a digital video recorder, it can be said that the most accurate method is to analyze it, but the cost of data acquisition is enormous and the infant's ambiguous utterance data Since there is no engineering technology that automatically recognizes and analyzes at the word level, it is very difficult to realize. On the other hand, there is also a method in which a questionnaire is answered every predetermined period (for example, once every three months) to grasp a change in the number of words newly spoken by an infant. In this case, if the predetermined period is long, it is difficult to grasp the exact timing of the vocabulary explosion. In addition, if the predetermined period is short, the burden on the respondent of the questionnaire (the parent of the infant) increases. Therefore, in practice, it is desirable to have a method capable of acquiring data at fine time points while reducing the burden on the parent who records the data.

  Therefore, in the present embodiment, data acquisition using the web diary method is applied. In this method, when an infant learns a new word (speaks), a specific site on the web is accessed via a network from a mobile phone or a personal computer, and the word that the infant remembers is recorded along with the diary of the day. (Non-Patent Document 2 “Tetsuo Kobayashi, Masaaki Nagata (2009),“ Infant Language Development Research Using the Web: Trial of Large-Scale Longitudinal Data Collection ”, Proc. 15th Annual Conference of the Language Processing Society, pp. 534-537., Non-Patent Document 3, “Tetsuo Kobayashi, Masaaki Nagata (March 2010),“ Reliability Verification of Infant Vocabulary Development Data Collected on the Web ”, 16th Annual Conference of the Language Processing Society of Japan See the collection of papers, pages 403-406.). The effectiveness of this method is very good in that it has been scientifically verified.

  In addition, the advantage of data acquisition by this method is that it is relatively easy for parents to record, but the date of recording (the date when the infant learned a new word) and the date of birth of the infant From the difference, it is possible to calculate the age at which the infant learned a new word. By using the data acquired in this way, the vocabulary learning speed estimation apparatus 10 of the present embodiment can estimate how many words are learned per day.

  For example, an input screen 50 as shown in FIG. 10 is provided in the input unit 12 as an input interface to input data. In the input screen 50 of FIG. 10, a date input area 52, a word input area 54, a date of birth display area 56, and a registration correction button 58 are provided.

  In the date input area 52, the date (word acquisition date) when the infant spoke a new word can be input by direct input or selection from a pull-down menu. Further, when the input screen 50 is opened, the date of the day may be input as an initial value. In the word input area 54, the utterance and meaning of a word newly learned by the infant can be input by direct input. The date of birth display area 56 displays the date of birth of the infant registered in advance. When the date of birth is not registered, or when the registered date of birth is to be corrected, by pressing the registration correction button 58, the date of birth input screen is displayed and the input of the date of birth is accepted.

  By accepting the input data in this way, as shown in Fig. 11, when (eg: September 12, 2009) what word (eg: doggie) and what meaning (eg: dog) was spoken A data set composed of the date of birth, the date of word acquisition, the utterance, and the meaning is obtained.

  The calculation unit 14 is a computer that includes a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory) that stores a program for executing a vocabulary learning speed estimation processing routine described later. It is configured. Functionally, the calculation unit 14 is functionally a word acquisition age calculation unit 20, a data set storage unit 22, a plateau removal unit 24, a plateau data storage unit 26, a function approximation unit 28, and a vocabulary learning speed calculation unit 30. And can be represented by a configuration including The plateau removal unit 24 is an example of a replacement unit, and the vocabulary learning speed calculation unit 30 is an example of an estimation unit.

The word acquisition age calculation unit 20 indicates the date of acquisition of each word from the difference between the word acquisition date and the date of birth in the data set input from the input unit 12. Calculate age. For example, if the word acquisition date is “September 12, 2009” and the date of birth is “September 12, 2008”, the age of acquisition = September 12, 2009-September 12, 2008 = 365 days old. The calculated acquisition ages of the words are arranged in ascending order, and an integer series of 1, 2, 3,... Is assigned in ascending order to calculate the cumulative number of words (the most remembered word). Thereby, a time series of a data set including a set of the acquired age x _i and the cumulative word number y _i is generated and stored in the data set storage unit 22.

  In addition, it is good also as a format which acquires in the input part 12 the data set which consists of a set of acquisition age and the cumulative number of words directly. In this case, in the calculation part 14, the structure of the word acquisition age calculation part 20 can be abbreviate | omitted. Moreover, it is good also as a format which acquires the data set previously accumulate | stored in the memory | storage device.

The data set storage unit 22 stores a time series of a data set including a set of the acquisition date age x _i and the cumulative word number y _i generated by the word acquisition age calculation unit 20.

The plateau removing unit 24 is a plateau in which the interval between days until the next new word is spoken from the time series of the acquired age and cumulative vocabulary data sets acquired from the data set storage unit 22 is greater than p. (day intervals old) it is, so that all the reference value p ', replacing the acquisition date age x _i of each data set. As a result, a time series of a data set of new acquired age x _i ′ and cumulative vocabulary number y _i is obtained. The plateau length threshold to be removed is set as p, and a reference plateau for replacing the plateau is set as p ′ in advance and stored in the plateau data storage unit 26.

The function approximating unit 28 replaces the plateau removing unit 24 with the time series of the data set composed of the combination of the acquired age x _i ′ and the accumulated word number y _i , the accumulated word number as the y axis, and the acquired age as the acquired date. Plot in the coordinate system with x-axis and approximate each plotted data point with a function.

Specifically, a function y = f (x ′) that approximates the relationship between two variables of the acquired age x _i ′ and the cumulative word number y _i is assumed. In the present embodiment, a straight line such as a linear function y = ax ′ + b is assumed. These a and b can be obtained by the least square method as shown in the following equation (1).

  Here, the square distance between the data and the straight line is used as the norm, but another norm may be used.

  The vocabulary learning speed calculation unit 30 differentiates the function y = f (x ′) approximated by the function approximation unit 28 by x ′ to calculate the vocabulary learning speed. In the example using the above formula (1), the vocabulary learning speed has a slope a and is constant regardless of time.

  The output unit 16 is mounted with a display, a printer, a magnetic disk, or the like, and the calculation result of the calculation unit 14 is output. For example, the vocabulary learning speed calculated by the vocabulary learning speed calculation unit 30 is output.

<Operation of Vocabulary Learning Speed Estimator>
Next, a vocabulary learning speed estimation processing routine executed by the vocabulary learning speed estimation apparatus 10 of the present embodiment will be described with reference to FIG.

  In step S100, a data set composed of a date of birth, a word acquisition date, an utterance, and a meaning is acquired.

Next, in step S102, the acquisition date of each word is calculated from the difference between the date of acquisition and the date of birth of the data set acquired in step S100. Then, the calculated acquisition ages of the words are arranged in ascending order, and an integer series of 1, 2, 3,... Is assigned from the smallest to calculate the cumulative number of words. Thereby, a time series of a data set composed of a set of the acquired age x _i and the cumulative word number y _i is generated and stored in the data set storage unit 22.

  In the next step S104, each data set is replaced so as to remove the plateau from the time series of the data set generated in step S102.

  Here, step S104 is realized by the processing routine shown in FIG.

First, the length L of the time series sequence and the data set when the data set acquisition date age x _i and the cumulative vocabulary y _i is obtained. Also, let p be the threshold for the length of the plateau to be removed, and p ′ be the reference plateau for replacing that plateau.

In step S120, the variable i indicating the time-series order of the data set is set to 1, and S for calculating the cumulative value of the plateau is set to 0. In step S122, it is subtracted from x _i entered the set accumulated plateau S, and records replaced with new Nichiyowai x _i '.

In step S124, it is determined whether i is less than length L. If i is less than length L, it is determined that there is a next data set. In step S126, the next data is set. The difference between the set age x _{i + 1} and the current age x _i (the age interval) is compared with the threshold value p, and if the difference in age is greater than p, in step S128, the cumulative Let the plateau be S = S + x _{i + 1} −x _i −p ′. This determines how much value should be subtracted from the age of each subsequent data set to replace a plateau greater than p (the age interval) with p ′.

  In step S130, i is incremented by 1, and the process returns to step S122. In this way, the operations in steps S122 to S128 are repeated for all L data sets.

  If it is determined in step S124 that i is greater than or equal to the length L, it is determined that there is no next data set, and in step S132, all xj 'and yj recorded in step S122 are output. The processing routine ends.

According to the above processing routine, an updated value of the data set of acquired age x _i ′ and cumulative vocabulary number y _i is obtained. Here, p is assumed to be about 4 to 15, and p ′ is assumed to be about 1 to 3.

Then, in step S106 of FIG. 12, the time series of the data set composed of the combination of the acquired age x _i ′ and the cumulative vocabulary number y _i replaced so as to remove the plateau in step 104 is represented by the cumulative word count. Is plotted on a coordinate system with the y-axis and the acquired age as the x-axis, and a linear function (y = ax + b) is determined so as to minimize the norm with each plotted data point according to the above equation (1).

  Next, in step S108, the value of the slope a of the linear function obtained in step S106 is calculated as the vocabulary learning speed. In step S110, the vocabulary learning speed calculated in step S108 is output, and the process ends.

  Table 1 shows the vocabulary learning speed obtained by estimating the vocabulary learning speed using the actual data of the 15 speakers shown in FIGS. However, p is 4 to 7, and p 'is 1 and 2. Although the individual values are different, the relative vocabulary learning speed relationship does not change much with any value.

  As described above, according to the vocabulary learning speed estimation device of the present embodiment, the age interval is a reference for a plateau in which the age interval until the next new word is spoken is equal to or greater than a threshold value. By replacing each age of the data set to be a value, approximating with a linear function, and estimating the slope of the approximated linear function as the vocabulary learning speed, without assuming the vocabulary explosion time, The vocabulary learning speed can be estimated accurately and stably.

  In addition, the plateau is removed from the data indicating the relationship between the age at which the infant began speaking a new word and the cumulative number of words, the vocabulary learning curve is approximated by a function, and the vocabulary learning speed is estimated from the slope. . Thereby, a true vocabulary learning speed that does not depend on the vocabulary explosion can be obtained. One of the effects of estimating this vocabulary learning speed is that not only can you estimate the true vocabulary learning speed from a small number of data at the beginning of vocabulary learning (for example, the first 20 words you learned), but you can also use this period of data It can be used to estimate the vocabulary learning speed. By using such vocabulary learning speed unique to infants, it is possible to make tailored education tailored to each individual's characteristics and development more effectively, and it can be a very valuable index in commerce.

  In the above embodiment, the case where the function approximating unit approximates the vocabulary learning curve to a linear function has been described, but it may be approximated to another function.

  The present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the gist of the present invention.

  In addition, the vocabulary learning speed estimation apparatus described above has a computer system inside, but the “computer system” includes a homepage provision environment (or display environment) if a WWW system is used. Shall be.

  In the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium.

DESCRIPTION OF SYMBOLS 10 Vocabulary learning speed estimation apparatus 12 Input part 14 Operation part 16 Output part 20 Word acquisition age calculation part 22 Data set storage part 24 Plateau removal part 26 Plateau data storage part 28 Function approximation part 30 Vocabulary learning speed calculation part

Claims (4)

Next to the time series of data indicating the relationship between the age at which the infant began speaking a new word and the cumulative number of words spoken by the infant until the age, the next new word The reference value in which the interval between the ages before the utterance is greater than a predetermined value within the range of 4 to 15 is within the range of 1 to 3 smaller than the predetermined value. A replacement means for replacing the age of each of the data,
A function approximation means for calculating a function, which approximates the change in the time series of the data using the data that have been replaced by the replacing means as a function related vocabulary learning,
A device that calculates functions related to vocabulary learning.
The function approximating unit is configured to minimize the square sum of the difference between the linear function approximated by the function and each data with respect to the time series transition of the data using the data replaced by the replacing unit. calculating the linear function that approximates the change in the time series of the data, as a function relating to the vocabulary learning,
The apparatus which calculates the function regarding the vocabulary learning of Claim 1.
A method of calculating a function relating to vocabulary learning in an apparatus for calculating a function relating to learning vocabulary comprising a substitution unit and function approximation hand stage,
An apparatus for calculating a function related to vocabulary learning is as follows:
With respect to the time series of data indicating the relationship between the age at which the infant began to speak a new word by the replacement means and the cumulative number of words that the infant began to speak before the age In a portion where the age interval until the next new word is spoken is larger than a predetermined value within the range of 4 to 15, the age interval is within a range of 1 to 3 smaller than the predetermined value. Substituting the age of each of the data to be a reference value within,
By the function approximation means, calculating a function, which approximates the change in the time series of the data using the data that have been replaced by the replacing means as a function relating to the vocabulary learning,
To calculate a function related to vocabulary learning to be performed.
The program for functioning a computer as each means which comprises the apparatus which calculates the function regarding the vocabulary learning of Claim 1 or 2.