JP2020187714A - Information processor, information processing system, learning device, and information processing program - Google Patents

Abstract

To provide an information processor, an information processing system, a learning device, and an information processing program for appropriately estimating an occupation of an object person.SOLUTION: An information processor 100 includes: a reception section 102 that receives an input of object person information on an object person; an extraction section 1460 that extracts a feature quantity for estimating a future occupation of the object person from the object person information; a learned estimation model 1400 that outputs a possibility of the future occupation of the object person as an estimated result on the basis of the feature quantity; and an output section 108 that outputs the estimated result in the estimation model. The estimation model 1400 is generated by learning processing using a data set for learning. The data set for learning includes learning data labeled with an occupation of the other object person for a learning parameter which is a parameter of the other object person.SELECTED DRAWING: Figure 7

Description

The present invention relates to an information processing device, an information processing system, a learning device, and an information processing program.

For example, a guardian (eg, parent) of an infant (target person) may want to predict the future occupation for which the infant is suitable. For example, future professions include, for example, athletes, singers, doctors and lawyers. Non-Patent Document 1 discloses a Web page that proposes the occupation of the subject.

[Search on May 10, 2019] Internet <URL> http: // oshigit. mydreams. jp / shindan / shindan_select1. php

However, Non-Patent Document 1 discloses a system for proposing a profession by having a subject answer a question to a subject who has reached a certain age. For example, when the subject is an infant or the like, I couldn't propose an appropriate profession.

An object of the present invention is to provide an information processing device, an information processing system, a learning device, and an information processing program that appropriately estimate the future occupation of a subject.

According to a certain aspect of the present invention, a reception unit that accepts input of target person information regarding the target person, an extraction unit that extracts a feature amount for estimating the future occupation of the target person from the target person information, and an extraction unit. Based on the feature amount, the estimation model includes a learned estimation model that outputs the possibility of the subject's future occupation as an estimation result, and an output unit that outputs the estimation result of the estimation model. Is generated by a learning process using a learning data set, and the learning data set is for learning in which the occupation of the other target person is labeled with respect to the learning parameter which is the parameter of the other target person. An information processing device containing data is proposed.

According to the present invention, the future occupation of the subject can be appropriately estimated.

It is a figure which shows the phase in the occupation estimation system according to this embodiment (the 1). It is a figure which shows the phase in the occupation estimation system according to this embodiment (the 2). It is a figure which shows the phase in the occupation estimation system according to this embodiment (the 3). It is a figure which shows an example of the occupation estimation system according to this embodiment. It is a figure which shows an example of the hardware configuration of the information processing apparatus which follows this embodiment. It is a figure which shows an example of the hardware configuration of the management apparatus according to this embodiment. It is a figure which shows an example of the functional structure of the information processing apparatus which follows this embodiment. It is a figure which shows an example of the structure of the estimation part according to this embodiment. It is a figure which shows an example of the estimation model according to this embodiment. It is a figure which shows an example of the functional structure of the management apparatus according to this embodiment. It is a flowchart of occupation estimation processing according to this embodiment. It is a flowchart of the management process according to this embodiment.

Embodiments of the present invention will be described in detail with reference to the drawings. The same or corresponding parts in the drawings are designated by the same reference numerals and the description thereof will not be repeated.

<A. Overview of occupation estimation system>
First, as a typical example of the information processing system according to the present invention, an outline of the occupation estimation system 1 according to the present embodiment will be described. In the present specification, the "subject" is a "person whose future occupation is presumed", for example, a person aged 0 to 18 years. A guardian is, for example, a person who protects the target person. The guardian is, for example, the subject's parents.

1 and 2 are diagrams for explaining a usage example of the occupation estimation system according to the present embodiment. In this embodiment, a plurality of phases are defined in using the occupation estimation system. In the information processing device of the present embodiment, the information acquired in each of the plurality of phases is input, and the future occupation of the target person is estimated based on the information. FIG. 1A is a diagram for explaining a voice acquisition phase. The voice acquisition phase is a phase in which voice is acquired in a question and answer session between the subject and the questioner. The microphone 20 acquires voice. The voice data of the voice acquired by the microphone 20 is input to the information processing device 100. The information processing device 100 stores the input voice data in a predetermined storage area. The information processing device 100 may be any device as long as it can process information. The information processing device 100 may be, for example, a PC (personal computer), a tablet, a smartphone, or the like.

For example, a question from a questioner includes a question asking the subject what profession he / she wants to have in the future. The question is, for example, a question such as "What occupation do you want to be in the future?" (Hereinafter referred to as "specific question"). Other specific questions include "what do you like", "what are you good at", and "what occupation do you want to be in the future" (hereinafter referred to as "specific questions"). May include.

For example, a person in the doctor's profession can confidently answer questions such as "what do you like" and "what are you good at" in early childhood? There is a tendency to answer "science". In addition, those who are in the profession of doctors tend to confidently answer "doctors" to questions such as "what profession do you want to be in the future?" In early childhood.

In this way, for a person in a specific occupation, the answer to a specific question includes a specific keyword and a specific way of speaking (tempo, etc.) as features. A specific keyword is a keyword indicating a specific occupation. In this way, a specific keyword in the answer to a specific question and a specific way of speaking (feature amount) are useful indexes for estimating the occupation of the subject.

FIG. 1B is a diagram for explaining a still image acquisition phase. The still image acquisition phase is a phase in which a still image including at least one of an image drawn by the target person and characters drawn by the target person is acquired. The camera 40 acquires a still image. The still image data of the still image acquired by the camera 40 is input to the information processing device 100. The information processing device 100 stores the input still image data in a predetermined storage area.

For example, a still image includes a painting (image) drawn by the subject, characters drawn by the subject, and the like. For example, people in highly educated professions (such as lawyers and doctors) tend to have beautiful childhood letters. "The characters are beautiful" means, for example, that the degree of deviation from the "model characters" is small.

Further, the painting is preferably a painting for a specific object. Specific objects are, for example, flowers, animals, and the like. For example, those who work as painters tend to have beautiful paintings in their childhood. "Paintings are beautiful" means that the characters in childhood from "model paintings" tend to be beautiful. A "model painting" is, for example, a photograph of a particular object.

In this way, the images drawn by a person in a specific occupation during childhood and the characters drawn by the subject are useful indicators for estimating the occupation of the subject.

FIG. 1C is a diagram for explaining a moving image acquisition phase. The video acquisition phase is a phase in which a video is acquired in a question and answer session between the subject and the questioner. The camera 40 acquires a moving image. The moving image data of the moving image acquired by the camera 40 is input to the information processing device 100. The information processing device 100 stores the input moving image data in a predetermined storage area.

For example, a moving image is a moving image in which a target person performs a specific action. The specific motion includes a motion of the subject walking, a motion of the subject running, a motion of the subject kicking the ball, a motion of the subject swinging a bat, and the like. For example, people in the profession of baseball players tend to have beautiful running forms and bat-waving forms in their childhood. "The form is beautiful" means that the degree of deviation from the "model form" is small. Also, for example, those who are in the profession of baseball players tend to run faster in their childhood.

In this way, a specific movement of a person in a specific occupation in childhood is a useful index for estimating the occupation of the subject.

FIG. 2D is a diagram for explaining the grade acquisition phase of the subject. The subject's grade acquisition phase is a phase in which the subject's academic achievement is acquired. The subject's academic performance includes the results of examinations at cram schools and elementary schools. Grades include, for example, test scores and test deviations. As a method of inputting the results of the subject to the information processing device 100, for example, the information processing device 100 displays a grade input screen (not shown in particular). Then, a guardian or the like (user) of the target person inputs the grade of the target person from this grade input screen. The information processing device 100 stores the input performance data of the target person in a predetermined storage area.

Those in highly educated professions (such as attorneys and doctors) tend to have high childhood academic performance. In this way, the childhood performance of a person in a specific occupation is a useful index for estimating the occupation of the subject.

FIG. 2E is a diagram for explaining a guardian information acquisition phase. The parent information acquisition phase is a phase in which parent information is acquired. Parental information includes, for example, the parent's profession, the parent's educational background, and the parent's annual income. As a method of inputting the occupation of the guardian to the information processing device 100, for example, the information processing device 100 displays a guardian information input screen (not particularly shown). Then, the guardian or the like (user) of the target person inputs the guardian information from the guardian information input screen. The information processing device 100 stores the input guardian information data in a predetermined storage area.

Many parents of people in a particular profession (such as a doctor) are in that particular profession. For example, parents of people in the profession of doctors are often doctors. In addition, for example, in order to get a profession that requires a high degree of education (for example, a lawyer, a doctor, etc.), it is necessary to go to a cram school, etc., which costs a lot of money. Therefore, it is preferable that the parents of the target person also have a large annual income. In addition, in order to get a job that requires a high degree of education (for example, a lawyer, a doctor, etc.), the subject must enter and graduate from a university with a high deviation value. In general, when the guardian of the subject is highly educated, the subject also tends to be highly educated. The reason is that if the parents are highly educated, the viewpoint of the educational environment will be adjusted, and it will be easier for the parents to devote themselves to their studies. In addition, when the blood of the guardian and the subject is connected, if the guardian is highly educated from the viewpoint of heredity, the subject is likely to be highly educated.

As described above, in the present embodiment, in the four phases shown in FIGS. 1 (A) to 1 (C) and 2 (D) and 2 (E), the information processing apparatus 100 uses the audio data and the still image data. , Moving image data, subject's performance data, and guardian information data. The information processing device 100 stores these five data in a predetermined storage area. Hereinafter, these five data are also referred to as "target person information regarding the target person" or simply "five data". Further, in the present embodiment, the information processing apparatus 100 or the like describes a mode in which all five data are used, but as a modification, at least one of the five data may be used.

FIG. 3 is a diagram for explaining the display result phase. The information processing device 100 estimates the possibility of the subject's future occupation by using the estimation model 1400 or the like based on the five data stored in the information processing device 100. The information processing device 100 outputs the estimated estimation result to the display device 200, and causes the display device 200 to display the display result. In the example of FIG. 3, the most suitable profession is displayed in order along with the characters "professions that you are suitable for". In the example of FIG. 3, the display device 200 displays the "doctor" with the most suitable (most likely) occupation as the "first place", followed by the most suitable (next possible) occupation. "Attorney" is displayed as "2nd place", and "Financial office worker" is displayed as "3rd place" for the next suitable (next possible) occupation.

In this way, the information processing device 100 outputs the possibility of the subject's future occupation to the display device 200 as an estimation result. The display device 200 displays a screen showing the estimation result. Therefore, the subject and the subject's guardian can recognize the potential of the subject's future profession. In addition, the information processing apparatus 100 may display a specific occupation (for example, a doctor) at a future rate (for example, 50%).

[Configuration example of occupation estimation system]
Next, a system configuration example of the occupation estimation system 1 according to the present embodiment will be described. First, an overall configuration example of the occupation estimation system 1 will be described, and then a hardware configuration example of the main device included in the occupation estimation system 1 will be described.

(System configuration example)
FIG. 4 is a schematic diagram showing an example of the system configuration of the occupation estimation system 1 according to the present embodiment. With reference to FIG. 4, the occupation estimation system 1 includes a management device 300, a network 2, one or more information processing devices 100, and one or more display devices connected to each of the one or more information processing devices 100. Includes 200 and.

The information processing device 100 receives the input of five data which is the target person information, and estimates the future occupation of the target person from the five data. The management device 300 manages and updates the trained model used by the information processing device 100. More specifically, the management device 300 acquires the target person information (five data) input from the information processing device 100. This target person information becomes "target person information A for learning". Then, when the target person has a profession after several years, the management device 300 associates the input "target person information A for learning" with the profession data indicating the profession. Generate a training data set.

In addition, the manager of the occupation estimation system 1 may interview a person in the occupation and acquire at least one of the five data of the person in the occupation in his or her childhood. Good. For example, "voice data of Q & A between subject A and questioner B", "still image data (paintings and characters in childhood)", and "moving image data (childhood)" in the doctor's childhood. At least one of "data on images of running period)", "results data on schoolwork", and "guardian data of parents in their childhood" is acquired. This at least one data becomes "target person information B for learning". Then, the manager inputs the acquired at least one data and occupation data indicating the occupation of the person (person who has an occupation) to the management device 300. The management device 300 generates a learning data set by associating at least one input data (“learning target person information B”) with occupational data. The management device 300 uses the generated training data set to perform training of the trained model (which may include both new learning and additional learning).

(Information processing device 100)
FIG. 5 is a schematic diagram showing an example of the hardware configuration of the information processing device 100 constituting the occupation estimation system 1 according to the present embodiment.

With reference to FIG. 5, the information processing apparatus 100 includes a processor 154, a memory 156, a network controller 158, a storage 109, a printer 120, a camera interface, and an input device interface 172 as main hardware elements. , A microphone interface and a display device interface 176.

The processor 154 is an arithmetic unit that executes processing necessary for realizing the information processing apparatus 100 by executing various programs as described later. The processor 154 includes, for example, one or a plurality of CPUs (Central Processing Units). ) And GPU (Graphics Processing Unit). A CPU or GPU having a plurality of cores may be used.

The memory 156 provides a storage area for temporarily storing a program code, a work memory, and the like when the processor 154 executes a program. As the memory 156, for example, a volatile memory device such as a DRAM (Dynamic Random Access Memory) or a SRAM (Static Random Access Memory) may be used.

The storage 160 stores an OS (Operating System) 112 executed by the processor 154, an application program 164 for realizing a functional configuration as described later, a trained model 326, and the like. As the storage 160, for example, a non-volatile memory device such as a hard disk or SSD (Solid State Drive) may be used.

As a part of the library or functional module required when executing the application program 164 on the processor 154, the library or functional module provided as standard by OS 162 may be used.

The optical drive 152 reads information such as a program stored in an optical disk 124 such as a CD-ROM (Compact Disk Read Only Memory) or a DVD (Digital Versaille Disc). The optical disk 124 is an example of a non-transient recording medium, and is distributed in a non-volatile state in which an arbitrary program is stored. The information processing device 100 according to the present embodiment can be configured by the optical drive 152 reading a program from the optical disk 124 and installing the program in the storage 160. Therefore, the subject of the present invention may be the program itself installed in the storage 160 or the like, or a recording medium such as an optical disk 124 containing a program for realizing a function or process according to the present embodiment.

FIG. 5 shows an optical recording medium such as an optical disk 124 as an example of a non-transient recording medium, but the present invention is not limited to this, and a semiconductor recording medium such as a flash memory or a magnetic recording medium such as a hard disk or a storage tape is shown. , MO (Magnet-Optical disk) and other magneto-optical recording media may be used.

Alternatively, the program for realizing the information processing device 100 is not only stored and distributed in an arbitrary recording medium as described above, but also distributed by downloading from a server device or the like via the Internet or an intranet. Good.

The camera interface 170 serves as an interface between the camera 40 and the information processing device 100. The input device interface 172 serves as an interface between the input device 50 and the information processing device 100. The input device 50 includes at least one of a keyboard and a mouse. The microphone interface 174 serves as an interface between the microphone 20 and the information processing device 100. The display device interface 176 serves as an interface with the display device 200.

(Management device 300)
FIG. 6 is a schematic diagram showing an example of the hardware configuration of the management device 300. With reference to FIG. 6, the management device 300 includes a processor 304, a memory 306, a network controller 308, and a storage 310 as the main hardware elements.

The processor 304 is an arithmetic unit that executes processing necessary for realizing the management device 300 by executing various programs. The processor 304 is composed of, for example, one or a plurality of CPUs or GPUs. A CPU or GPU having a plurality of cores may be used. In the management device 300, it is preferable to employ a GPU or the like suitable for the learning process for generating the trained model.

The memory 306 provides a storage area for temporarily storing a program code, a work memory, and the like when the processor 304 executes a program. As the memory 306, for example, a volatile memory device such as DRAM or SRAM may be used.

The network controller 308 transmits / receives data to / from an external device (information processing device 100 or the like) via the network 2. The network controller 308 may be compatible with any communication method such as Ethernet, wireless LAN, and Bluetooth.

The storage 310 includes an OS 312 executed by the processor 304, an application program 314 for realizing a functional configuration as described later, a preprocessing program 316 for generating a learning data set 324, and a learning data set 324. Stores a training program 318 and the like for generating a trained model 326 using.

The learning target person information 325 includes the above-mentioned "learning target person information A" and "learning target person information B" input to the information processing apparatus 100.

The learning data set 324 is a training data set in which vocational data 218 (correct label or tag) described later is added to the learning target person information 325. The trained model 326 is an estimation model obtained by executing a training process using the training data set 324.

As the storage 310, for example, a non-volatile memory device such as a hard disk or SSD may be used.

Some of the libraries and functional modules required to execute the application program 314, the preprocessing program 316, and the learning program 318 on the processor 304 may use the libraries or functional modules provided as standard by the OS 312. .. In this case, each of the application program 314, the preprocessing program 316, and the learning program 318 does not include all the program modules necessary to realize the corresponding functions, but under the execution environment of OS 312. By installing it in, it is possible to realize the functional configuration described later. Therefore, even a program that does not include some such libraries or functional modules may be included in the technical scope of the present invention.

The application program 314, the preprocessing program 316, and the learning program 318 include optical recording media such as optical disks, semiconductor recording media such as flash memory, magnetic recording media such as hard disks or storage tapes, and magneto-optical recording media such as MO. It may be stored in a non-transient recording medium, distributed, and installed in the storage 310. Therefore, the subject of the present invention may be the program itself installed in the storage 310 or the like, or a recording medium in which the program for realizing the functions and processes according to the present embodiment is stored.

Alternatively, the program for realizing the management device 300 may be distributed not only by being stored in an arbitrary recording medium as described above and distributed, but also by downloading from a server device or the like via the Internet or an intranet. ..

The input unit 330 accepts various input operations. As the input unit 330, for example, a keyboard, a mouse, a touch panel, a pen, or the like may be used.

FIG. 6 shows a configuration example in which the general-purpose computer (processor 304) realizes the management device 300 by executing the application program 314, the preprocessing program 316, and the learning program 318, but in order to realize the management device 300, All or part of the required functions may be realized by using a hard-wired circuit such as an integrated circuit. For example, it may be realized by using ASIC, FPGA or the like.

<C. Functions and processing of information processing device 100>
Next, the functions and processes of the information processing apparatus 100 constituting the occupation estimation system 1 according to the present embodiment will be described.

(C1: Functional configuration of information processing device 100)
FIG. 7 is a schematic diagram showing an example of the functional configuration of the information processing device 100 constituting the occupation estimation system 1 according to the present embodiment. Each function shown in FIG. 10 may be typically realized by the processor 154 of the information processing apparatus 100 executing the OS 162 and the application program 164 (both of which are referred to in FIG. 7).

FIG. 7 is a diagram showing a functional configuration example of the information processing device 100. The information processing device 100 has a reception unit 102, a storage unit 104, an estimation unit 106, and an output unit 108.

As described with reference to FIGS. 1 and 2, voice data is input to the information processing device 100 from the microphone 20. Further, moving image data and still image data are input to the information processing device 100 from the camera 40. Further, the grade data is input to the information processing device 100 from the grade input screen, and the guardian data is input to the information processing device 100 from the guardian input screen.

The reception unit 102 receives the five input data. The five data received by the reception unit 102 are temporarily stored in the storage unit 104. The five data stored in the storage unit 104 are input to the estimation unit 106 when a predetermined condition is satisfied. The predetermined condition is satisfied, for example, when a predetermined operation is performed by the user.

When the predetermined conditions are satisfied, the five data (moving image data 136, still image data 137, audio data 138, guardian data 139, and grade data 140) are input to the estimation unit 106. The estimation unit 106 outputs the possibility of the subject's future occupation based on the five data. The estimation unit 106 has an estimation model 1400 and an extraction unit 1460. The extraction unit 1460 includes the area identification module 141, the size adjustment module 142, the area identification module 143, the size adjustment module 144, the section identification module 145, the resampling module 146, the extraction module 147, and the area identification module 141 shown in FIG. It corresponds to the extraction module 148 and the like. The detailed processing of the estimation unit 106 will be described later. The output unit 108 outputs the estimation result to the outside. In the present embodiment, the "external" is the display device 200. The "external" may be other, and the "external" may be, for example, a printer or a network.

The display device 200 displays a screen (see FIG. 3) based on the estimation result output from the information processing device 100.

Further, the five data from the storage unit 104 are transmitted to the management device 300 via the output unit 108. The five data transmitted to the management device 300 are used as the target person information A for learning.

[Processing of estimation part]
Next, the details of the processing of the estimation unit 106 will be described with reference to FIG. With reference to FIG. 8, the estimation unit 106 includes an area identification module 141, a size adjustment module 142, an area identification module 143, a size adjustment module 144, an interval identification module 145, a resampling module 146, and an extraction module. It includes 147, an extraction module 148, and an estimation model 1400.

The area identification module 141 specifies the target person area in the moving image of the moving image data 136. Then, the area specifying module 141 extracts and outputs the target person moving image corresponding to the specified target person area from the moving image data 136.

Typically, the region identification module 141 identifies the subject region by extracting facial features such as eyes and nose and skeletal features such as limbs. The area identification module 141 outputs the target person moving image extracted from the moving image data 136 to the size adjusting module 142.

In the size adjustment module 142, the subject moving image is converted into a feature amount (feature amount vector) having a predetermined dimension and input to the estimation model 1400. Here, since the image size of the target person moving image extracted by the area specifying module 141 can fluctuate, the size adjustment module 142 standardizes the image size.

More specifically, the size adjustment module 142 adjusts the target person moving image from the area identification module 141 to an image having a predetermined number of pixels, and then adjusts the pixel value of each pixel constituting the adjusted image. It is input to the estimation model 1400 as the moving image feature amount 1410.

Similarly, the area identification module 143 identifies the target person area in the still image of the still image data 137. Then, the area specifying module 143 extracts and outputs the target person still image corresponding to the specified target person area from the still image data 137.

Typically, the domain identification module 143 identifies characters and paintings. The area identification module 143 outputs a character image or a painting image extracted from the still image data 137 to the size adjustment module 144.

In the size adjustment module 144, the still image of the subject is converted into a feature amount (feature amount vector) having a predetermined dimension and input to the estimation model 1400. Here, since the image size of the character image or the painting image extracted by the area specifying module 143 can fluctuate, the size adjustment module 144 standardizes the image size.

More specifically, the size adjustment module 144 adjusts the character image and the painting image from the area identification module 143 to an image having a predetermined number of pixels, and then adjusts the pixel value of each pixel constituting the adjusted image. Is input to the estimation model 1400 as a still image feature amount 1420.

The section specifying module 145 identifies the section of the voice emitted by the target person included in the voice data 138, extracts the specific section voice, and outputs the voice. Typically, the section identification module 145 analyzes the temporal change of the voice indicated by the voice data 138, and identifies the section in which the amplitude or frequency has changed with respect to the noise component around the information processing apparatus 100. By doing so, the specific section voice is extracted.

In addition, the specific section voice includes the question of the questioner B. The section specifying module 145 does not specify the question part of the questioner B, but specifies only the answer part of the target person A. In addition, the particular section audio may include information indicating the subject's current mood and tempo.

In this way, the section specifying module 145 identifies the voice of the portion of the voice collected by the microphone 20 that corresponds to the utterance of the target person.

With reference to FIG. 8 again, the specific section voice extracted by the section specifying module 145 from the voice data 138 is output to the resampling module 146. In the resampling module 146, the specific section voice is converted into a feature quantity (feature quantity vector) having a predetermined dimension and output to the estimation model 1400. Here, since the time length of the specific section voice specified by the section specifying module 145 can vary, the resampling module 146 standardizes the number of voice samples.

More specifically, the resampling module 146 samples the time waveform of the voice indicated by the specific section voice from the section specification module 145 with a predetermined number of samples, and obtains the amplitude value at each sampling point as the voice feature. It is input to the estimation model 1400 as the quantity 1430.

The extraction module 147 extracts features corresponding to the guardian's occupation, the guardian's educational background, and the guardian's annual income from the guardian data 139. The guardian features 1440 extracted by the extraction module 147 are input to the estimation model 1400. The extraction module 148 extracts the grades of the subject from the grade data 140. The performance feature amount 1450 extracted by the extraction module 147 is input to the estimation model 1400.

The estimation model 1400 is built on the trained model 326 which defines the network structure and the corresponding parameters. By inputting the moving image feature amount 1410, the still image feature amount 1420, the audio feature amount 1430, the guardian feature amount 1440, and the performance feature amount 1450 into the estimation model 1400, the arithmetic processing defined by the estimation model 1400 is executed. Therefore, the score for each occupation is calculated as the estimation result 1449. Here, the score for each occupation is a value indicating the possibility of each occupation suitable for the target person.

The estimation model 1400 is generated by a learning process using the learning data set 324 as described later.

As described above, the estimated model 1400, which is the trained estimation model, receives the inputs of the moving image feature amount 1410, the still image feature amount 1420, the audio feature amount 1430, the guardian feature amount 1440, and the performance feature amount 1450. The possibility of the subject's future occupation is output as an estimation result.

FIG. 9 is a schematic diagram showing a network configuration example of the estimation model 1400 shown in FIG. With reference to FIG. 9, the estimation model 1400 is a network classified as a DNN (Deep Neural Network). The estimation model 1400 includes a preprocessing network 1460, 1470, 1480, 1490, 1500 classified as a CNN (Convolutional Neural Network), an intermediate layer 1510, an activation function 1492 corresponding to an output layer, and a Softmax function 1494. Including.

The preprocessing networks 1460, 1470, 1480, 1490, and 1500 are based on moving image features 1410, still image features 1420, audio features 1430, guardian features 1440, and performance features 1450, which have relatively large orders. It functions as a kind of filter for extracting features that are effective for calculating the estimation result 1449. Each of the pretreatment networks 1460, 1470, 1480, 1490, and 1500 has a configuration in which convolutional layers (CONV) and pooling layers (Pooling) are alternately arranged. The number of convolutional layers and pooling layers does not have to be the same, and an activation function such as ReLU (rectified linear function) is arranged on the output side of the convolutional layer. Further, in FIG. 9, the details of the preprocessing networks 1470, 1480, 1490, and 1500 are omitted, but for example, the preprocessing networks 1470, 1480, 1490, and 1500 have the same configuration as that of the preprocessing network 1460. Will be done.

More specifically, the preprocessing network 1460 receives the input of the moving image feature amount 1410 (x11, x12, ..., X1a) and outputs the internal feature amount indicating the degree of beauty of characters and paintings in the moving image. Constructed to output. The preprocessing network 1470 receives the input of the still image feature amount 1420 (x21, x22, ..., X2b) and outputs the internal feature amount indicating the degree of cleanliness of characters and paintings in the still image. Will be built. The preprocessing network 1480 receives the input of the voice feature amount 1430 (x31, x32, ..., X3c), and receives the information for identifying the occupation in the speech and the current feeling (mood) of the subject. It is constructed to output internal features that indicate the information to be shown. The preprocessing network 1490 is constructed so as to receive the input of the guardian feature amount 1440 (x41, x42, ..., X4d) and output the guardian's occupation, the guardian's educational background, the guardian's annual income, etc. Will be done. The preprocessing network 1500 is constructed so as to receive the input of the grade feature amount 1450 (x51, x52, ..., X5e) and output the grade of the subject.

The intermediate layer 1510 consists of a fully coupled network with a predetermined number of layers and outputs from each of the preprocessing networks 1460, 1470, 1480, 1490, 1500 with the weights and biases determined for each node. Combine sequentially every time.

In the example of FIG. 9, it is assumed that the preprocessing networks 1460, 1470, 1480, 1490, and 1500 all have the same configuration.

An activation function 1492 such as ReLU is arranged on the output side of the intermediate layer 1510, and finally, after being normalized to a probability distribution by the Softmax function 1494, the estimation result 1449 (y1, y2, ... , YN) is output.

In the learning phase as described below, the parameters of each element that builds the network of the estimation model 1400 are optimized.

[Learning phase]
Next, the learning phase will be described. FIG. 10 is a diagram for explaining the processing contents in the learning function of the management device 300. With reference to FIG. 10, the management device 300 has the area identification module 141, the size adjustment module 142, the area identification module 143, the size adjustment module 144, the section identification module 145, and the resampling module 146 as learning functions. And an extraction module 147 and an extraction module 148. In these modules, the information processing apparatus 100 is substantially the same as each module described with reference to FIG. Therefore, the detailed description of these modules will not be repeated.

Further, the management device 300 includes an optimization unit 362 as a learning function. The optimization unit 362 generates the trained model 326 by optimizing the model parameter 364 for defining the estimation model 1400.

The optimization unit 362 uses each set (learning data) of moving image data 136, still image data 137, audio data 138, guardian data 139, and grade data 140 included in the learning data set 324 to model. Optimize parameter 364.

More specifically, the management device 300 is based on the moving image data 136, the still image data 137, the audio data 138, the guardian data 139, and the performance data 140 of each set included in the learning data set 324. A quantity 1410, a still image feature quantity 1420, a voice feature quantity 1430, a guardian feature quantity 1440, and a performance feature quantity 1450 are generated. The moving image feature amount 1410, the still image feature amount 1420, the audio feature amount 1430, the guardian feature amount 1440, and the performance feature amount 1450 are input to the estimation model 1400 to calculate the estimation result 1449. Then, the optimization unit 362 calculates an error by comparing the estimation result 1449 output from the estimation model 1400 with the corresponding occupation data 218 (correct answer label), and according to the calculated error (calculated error). Optimize (adjust) the value of model parameter 364 (to make it smaller).

That is, the optimization unit 362 corresponds to the learning unit, and the training data set 324 (moving image data 136, still image data 137, audio data 138, guardian data 139, and grade data 140 are labeled with occupational data 218. Estimated output by inputting moving image feature amount 1410, still image feature amount 1420, audio feature amount 1430, guardian feature amount 1440, and performance feature amount 1450 extracted from the data) into the estimation model 1400. The estimation model 1400 is optimized so that the result 1449 approaches the occupational data 218 labeled in the training data. In other words, the optimization unit 362 extracts feature quantities from the moving image data 136, the still image data 137, the audio data 138, the guardian data 139, and the performance data 140 included in the learning data set 324, and estimates the model 1400. The model parameter 364 is adjusted so that the estimation result 1449 calculated when inputting to is consistent with the corresponding occupation data 218.

In the same procedure, the estimation model 1400 is based on the training data (moving image data 136, still image data 137, audio data 138, guardian data 139, and performance data 140) included in the training data set 324. The trained model 326 is generated by iteratively optimizing the model parameter 364.

When the optimization unit 362 optimizes the value of the model parameter 364, an arbitrary optimization algorithm can be used. More specifically, as the optimization algorithm, for example, SGD (Stochastic Gradient Descent), Momentum SGD (Inertia Term Addition SGD), AdaGrad, RMSprop, AdaDelta, Adam (Adaptation), etc. Gradient method can be used.

The estimation model 1400 whose model parameter 364 is optimized by the optimization unit 362 corresponds to the trained model 326 and is transmitted to the information processing apparatus 100.

[Flow of occupation estimation processing]
Next, a flowchart of the occupation estimation process by the information processing apparatus 100 will be described with reference to FIG. In step S2, the information processing apparatus 100 extracts the moving image feature amount 1410 (see FIG. 8) from the moving image data 136. Next, in step S4, the information processing apparatus 100 extracts the still image feature amount 1420 (see FIG. 8) from the still image data 137. Next, in step S6, the information processing apparatus 100 extracts the voice feature amount 1430 (see FIG. 8) from the voice data 138. Next, in step S8, the information processing apparatus 100 extracts the guardian feature amount 1440 (see FIG. 8) from the guardian data 139. Next, in step S10, the information processing apparatus 100 extracts the grade feature amount 1450 (see FIG. 8) from the grade data 140.

Next, in step S12, the information processing apparatus 100 is based on five feature amounts (moving image feature amount 1410, still image feature amount 1420, audio feature amount 1430, guardian feature amount 1440, and performance feature amount 1450). , Estimate the occupation of the subject (output the estimation result 1449 in FIG. 8).

[Learning process flow]
Next, a flowchart of the learning process by the management device 300 will be described with reference to FIG. First, in step S101, the management device 300 acquires a learning data set from each information processing device 100. Here, each information processing device 100 is all the information processing devices 100 connected to the management device 300 via the network 2 (see FIG. 4).

Further, the management device 300 transmits a request signal to all the information processing devices 100 at a predetermined opportunity. The predetermined trigger includes, for example, a first trigger in which the administrator performs an operation for transmitting a request signal to the management device 300, and a second trigger in which the predetermined period elapses.

Next, in step S102, the management device 300 selects one set (learning data set) of the learning data sets acquired in step S101.

Next, in step S104, the management device 300 extracts the feature amount of another target person from the selected learning data set. The processing of step S104 includes, for example, the area identification module 141, the size adjustment module 142, the area identification module 143, the size adjustment module 144, the section identification module 145, the resampling module 146, the extraction module 147, and the extraction module 148 in FIG. Do.

Next, in step S106, the management device 300 inputs the extracted five feature quantities into the estimation model 1400 and generates an estimation result 1449. Next, in step S108, the management device 300 optimizes the parameters of the trained model based on the error between the occupation (label) of the selected learning data set and the occupation related to the estimation result. Specifically, the management device 300 updates the parameters of the trained model so that the error between the occupation (correct label) of the selected learning data set and the occupation related to the estimation result becomes small.

Next, in step S110, the management device 300 determines whether or not all the learning data sets have been processed. When the management device 300 determines that all the learning data sets have not been processed (NO in step S110), the management device 300 returns to step S102. On the other hand, in step S110, when the management device 300 determines that all the learning data sets have not been processed (YES in step S110), the management device 300 proceeds to step S112.

In step S112, the management device 300 transmits the trained model to all the information processing devices 100. All the information processing devices 100 update the stored learned model with the transmitted trained model.

[Brief Summary]
According to the information processing apparatus 100 of the above-described embodiment, for example, as shown in FIG. 7, the reception unit 102 accepts the input of the target person information regarding the target person. The extraction unit 1460 extracts the feature amount for estimating the future occupation of the subject from the subject information. Further, the estimation model 1400 is a trained estimation model, and the estimation model 1400 outputs the possibility of the subject's future occupation as an estimation result based on the feature amount. The output unit 108 outputs the estimation result of the estimation model 1400. Further, as shown in FIG. 10 and the like, the estimation model 1400 is generated by the learning process using the learning data set, and the learning data set is the same as the learning parameters which are the parameters of other subjects. Includes learning data labeled with the profession of another subject (occupational data 218). Thereby, the information processing apparatus 100 can estimate the future occupation of the target person based on the target person information of the target person by using the trained estimation model 1400. Therefore, the information processing device 100 can appropriately estimate the future occupation of the target person.

The features include the subject's answer to the question asked by the questioner (voice feature amount 1430), the subject's academic performance (performance feature amount 1450), the image drawn by the subject (still image feature amount 1420), and the subject. Includes at least one of the characters drawn by (still image feature amount 1420) and the moving image of the subject (moving image feature amount 1421). Therefore, the information processing apparatus 100 can appropriately estimate the occupation of the target person based on the feature amount appropriate for estimating the future occupation of the target person.

In addition, the question asked by the questioner includes a question asking the profession that the subject wants to get in the future. Such a question and the answer to the question are useful indicators for estimating the occupation of the subject. Therefore, the information processing device 100 can appropriately estimate the occupation of the target person.

In addition, the feature amount includes a guardian parameter which is a parameter for the guardian of the target person. The guardian parameters of the subject's guardian are useful indicators in estimating the subject's occupation. Therefore, the information processing device 100 can appropriately estimate the occupation of the target person.

The parental parameter also includes at least one of the parent's occupation, parent's educational background, and parent's annual income. The profession of the guardian, the educational background of the guardian, and the annual income of the guardian are useful indicators for estimating the profession of the target person. Therefore, the information processing device 100 can appropriately estimate the occupation of the target person.

Further, in the present embodiment, as shown in FIG. 3, the estimation result is information based on the probability of becoming a specific occupation. In the example of FIG. 3, the specific occupation is a doctor, an attorney, a financial office worker, and the like. Therefore, the information processing device 100 can make the target person, the guardian of the target person, and the like recognize a profession based on the probability of becoming a specific profession.

Further, the management device 300 as a learning device receives an input of a feature amount for estimating the future occupation of the target person, and outputs an estimation model 1400 that outputs the possibility of the future occupation of the target person as an estimation result. Generate. The management device 300 acquires a learning data set. In addition, the learning data set 324 includes a plurality of learning data in which the occupations of the other subjects are labeled with respect to the feature quantities of the other subjects. The management device 300 obtains the feature amounts of other subjects (moving image feature amount 1410, still image feature amount 1420, audio feature amount 1430, guardian feature amount 1440, and grade feature amount 1450) from the learning data set 324. Extract. In addition, the optimization unit 362 inputs the feature amount of the other target person received by the reception unit into the estimation model and outputs the estimation result, which is the occupation of the other target person whose training data is labeled. The estimation model 1400 is optimized so as to approach (occupation data 218). Therefore, the management device 300 can update the estimation model 1400 so that an appropriate estimation result can be output.

100 information processing device, 102 reception unit, 104 storage unit, 106 estimation unit, 108 output unit.

Claims (9)

The reception department that accepts input of target person information about the target person,
An extraction unit that extracts a feature amount for estimating the future occupation of the target person from the target person information,
Based on the feature quantity, a learned estimation model that outputs the possibility of the subject's future occupation as an estimation result, and
It is equipped with an output unit that outputs the estimation result of the estimation model.
The estimation model is generated by a learning process using a learning data set, and the learning data set labels the occupation of the other target person with respect to the learning parameter which is a parameter of the other target person. An information processing device that contains the learning data. The feature amount is at least one of the subject's answer to the question asked by the questioner, the subject's academic performance, the image drawn by the subject, the characters drawn by the subject, and the video of the subject. The information processing apparatus according to claim 1, which includes. The information processing device according to claim 2, wherein the question by the questioner includes a question asking the profession that the target person wants to take in the future. The information processing device according to any one of claims 1 to 3, wherein the feature amount includes a guardian parameter which is a parameter for the guardian of the target person. The information processing apparatus according to claim 4, wherein the guardian parameter includes at least one of the guardian's occupation, the guardian's educational background, and the guardian's annual income. The estimation result according to any one of claims 1 to 5, wherein the estimation result includes at least one of information indicating whether or not the person can become a specific occupation and information based on the probability of becoming the specific occupation. Information processing device. An information processing device that inputs features for estimating the future occupation of the target person into the trained estimation model and outputs the future occupation of the target person.
A learning device for generating the estimation model is provided.
The information processing device
It is provided with an extraction unit for extracting the feature amount.
The estimation model is trained to receive the input of the feature amount and output the possibility of the subject's future occupation as an estimation result.
The information processing device includes an output unit that outputs information about the future occupation of the target person based on the estimation result.
The learning device includes an acquisition unit for acquiring a learning data set.
The learning data set includes a plurality of learning data in which the occupations of the other subjects are labeled with respect to the features of the other subjects.
The learning device further
The reception section that accepts the input of the feature amount of the other target person,
The estimation result output by inputting the feature amount of the other target person received by the reception unit into the estimation model approaches the occupation of the other target person labeled in the learning data. , An information processing system including a learning unit that optimizes the estimation model. It is a learning device for generating an estimation model that receives input of a feature amount for estimating the future occupation of the target person and outputs the possibility of the future occupation of the target person as an estimation result.
Equipped with an acquisition unit to acquire the training data set
The learning data set includes a plurality of learning data in which the occupations of the other subjects are labeled with respect to the features of the other subjects.
The learning device is
An extraction unit that extracts the features of the other subject from the learning data set, and
The estimation model is optimized so that the estimation result output by inputting the feature amount of the other target person into the estimation model approaches the occupation of the other target person labeled in the learning data. A learning device that includes a learning unit that is optimized. On the computer
The steps to extract the features for estimating the future occupation of the target person from the target person information, and
Based on the feature quantity, the possibility of the future occupation of the subject is output as an estimation result, and a step of outputting an estimation result by a trained estimation model is executed.
The estimation model is generated by a learning process using a learning data set, and the learning data set labels the occupation of the other target person with respect to the learning parameter which is a parameter of the other target person. An information processing program that includes the learning data.

Images