JP2016162257A - Information processing device and information processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing device that is adapted so as to create teacher data less in noise when compared with preparation of the teacher data to be used in machine learning, using non-classified submission information.SOLUTION: Collection means of an information processing device is configured to collect posting information, classification means thereof is configured to classify the posting information collected by the collection means, and preparation means thereof is configured to prepare teacher data to be used in machine learning, using the posting information in an aggregate classified by the classification means.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information processing apparatus and an information processing program.

  Patent Document 1 has an object to enable learning to appropriately extract local features useful for detection and recognition of objects in different categories from a set of images. A set of data is input by an image input unit, local features having a predetermined geometric structure are detected from each input object image data by a local feature detection unit, and a plurality of detected local features are clustered. Based on the clustered result, a plurality of representative local features are selected by a feature selection unit, and an object based on the object image data is selected using a learning data set including the selected representative local features as teacher data. It is disclosed that learning for recognition or detection is performed by a learning control unit.

  In Patent Document 2, it is an object to avoid over-learning in classification of multi-class images and other feature data including very similar species. When a group is determined, the first learning unit and the second layer classifier for identifying in a group are realized. A type discriminator that performs discrimination by machine learning is generated in a state in which the discriminating state is not generated, and then a discrimination test is performed by using the type discriminator by the discriminating error counting means, and the number of times that discriminating errors are generated between categories is calculated. Aggregated, grouping processing means group categories that are prone to machine learning mistakes, and automatically sort out very similar species by counting prior identification errors. Identifying two-tier internal identification is performed, the multi-class classifier, excessive learning is disclosed to be inhibited.

  Patent Document 3 has a problem of collecting image data that is less biased in content and effective in learning, and includes a plurality of detectors that detect an image region to be detected from an image, and detection by the plurality of detectors. Integration means for combining the results of the above, and outputting a set of the image area candidate score of the learning image and the object-likeness score of the image area, a setting means for setting the adoption rate of the learning data, and the score and the setting means Disclosed is a selection unit that selects learning data from a set of an image area and a score based on a set acceptance rate, and a storage unit that stores the learning data selected by the selection unit. .

  In Patent Document 4, it is an object to provide a mobile phone use state identification device that can determine whether or not a person reflected in a camera is using a mobile phone, and a facial region is obtained from a learning image input from the camera. Is detected, a feature amount is extracted from the left and right partial regions adjacent to the face region, and an identification function for identifying whether the mobile phone is in use or not is created based on the extracted feature amount data group. Various parameter values of the function are stored in the identification function storage unit, a region with a face is detected from the identification target image input from the camera, and feature amounts are extracted from the left and right partial regions adjacent to the facial region of the identification target image, By inputting the extracted feature quantity into the discrimination function stored in the discrimination function storage unit, it is identified whether or not the mobile phone is in use (identification unit), and the discrimination result of the discrimination unit is output from the output unit Is disclosed.

  Non-Patent Document 1 has a problem that it is difficult to prepare a large amount of low-noise teacher data for machine learning, and provides clues on emoticon information included in Tweet (text called “tweet”). It is disclosed that a large amount of teacher data can be efficiently collected by collecting teacher data in a simple manner. For example, the emoticon “:-)” indicates positive, and the emoticon “:-(” indicates negative.

Japanese Patent Laid-Open No. 2005-215988 JP 2013-250809 A JP 2012-190159 A JP 2010-122838 A

Alec Go, Richa Bayani, and Lei Huang. 2009. Twiter-terentent classification using distant supervision. Technical report, Stanford University.

  The present invention provides an information processing apparatus and an information processing program that create teacher data with less noise compared to the case of creating teacher data used for machine learning using post information that is not classified. The purpose is that.

The gist of the present invention for achieving the object lies in the inventions of the following items.
The invention of claim 1 is a machine learning method using collection means for collecting post information, classification means for classifying post information collected by the collection means, and post information in the set classified by the classification means. It is an information processing apparatus characterized by comprising a creation means for creating teacher data used for.

  The invention according to claim 2 is characterized in that the classification means classifies the post information collected by the collection means using classification based on unsupervised clustering or information included in the post information. The information processing apparatus according to 1.

  The invention according to claim 3 is characterized in that the classification means classifies post information collected by the collection means by supervised clustering or machine learning using teacher data created in advance. The information processing apparatus described.

  The invention according to claim 4 is characterized in that the creation means creates one piece of teacher data used for machine learning by combining post information in the set classified by the classification means. The information processing apparatus described.

  According to a fifth aspect of the present invention, there is provided a computer using collection means for collecting post information, classification means for classifying post information collected by the collection means, and post information in a set classified by the classification means. An information processing program for functioning as a creation means for creating teacher data used for machine learning.

  According to the information processing apparatus of the first aspect, it is possible to create teacher data with less noise compared to the case of creating teacher data used for machine learning using post information that is not classified.

  According to the information processing apparatus of the second aspect, the collected posted information can be classified by using unsupervised clustering or information included in the posted information.

  According to the information processing apparatus of the third aspect, the collected post information can be classified by supervised clustering or machine learning using teacher data created in advance.

  According to the information processing apparatus of the fourth aspect, it is possible to create one piece of teacher data used for machine learning by combining post information in the classified set.

  According to the information processing program of the fifth aspect, it is possible to create teacher data with less noise compared to a case where teacher data used for machine learning is created using post information that is not classified.

It is a conceptual module block diagram about the structural example of 1st Embodiment. It is explanatory drawing which shows the system configuration example using 1st Embodiment. It is a flowchart which shows the process example by 1st Embodiment. It is explanatory drawing which shows the process example by 1st Embodiment. It is explanatory drawing which shows the process example by 1st Embodiment. It is explanatory drawing which shows the process example by 1st Embodiment. It is explanatory drawing which shows the process example by 1st Embodiment. It is explanatory drawing which shows the process example by 1st Embodiment. It is a conceptual module block diagram about the structural example of 2nd Embodiment. It is a flowchart which shows the process example by 2nd Embodiment. It is a flowchart which shows the process example by 2nd Embodiment. It is a block diagram which shows the hardware structural example of the computer which implement | achieves this Embodiment.

Hereinafter, examples of various preferred embodiments for realizing the present invention will be described with reference to the drawings.
FIG. 1 is a conceptual module configuration diagram of a configuration example according to the first embodiment.
The module generally refers to components such as software (computer program) and hardware that can be logically separated. Therefore, the module in the present embodiment indicates not only a module in a computer program but also a module in a hardware configuration. Therefore, the present embodiment is a computer program for causing these modules to function (a program for causing a computer to execute each procedure, a program for causing a computer to function as each means, and a function for each computer. This also serves as an explanation of the program and system and method for realizing the above. However, for the sake of explanation, the words “store”, “store”, and equivalents thereof are used. However, when the embodiment is a computer program, these words are stored in a storage device or stored in memory. It is the control to be stored in the device. Modules may correspond to functions one-to-one, but in mounting, one module may be configured by one program, or a plurality of modules may be configured by one program, and conversely, one module May be composed of a plurality of programs. The plurality of modules may be executed by one computer, or one module may be executed by a plurality of computers in a distributed or parallel environment. Note that one module may include other modules. Hereinafter, “connection” is used not only for physical connection but also for logical connection (data exchange, instruction, reference relationship between data, etc.). “Predetermined” means that the process is determined before the target process, and not only before the process according to this embodiment starts but also after the process according to this embodiment starts. In addition, if it is before the target processing, it is used in accordance with the situation / state at that time or with the intention to be decided according to the situation / state up to that point. When there are a plurality of “predetermined values”, they may be different values, or two or more values (of course, including all values) may be the same. In addition, the description having the meaning of “do B when it is A” is used in the meaning of “determine whether or not it is A and do B when it is judged as A”. However, the case where it is not necessary to determine whether or not A is excluded.
In addition, the system or device is configured by connecting a plurality of computers, hardware, devices, and the like by communication means such as a network (including one-to-one correspondence communication connection), etc., and one computer, hardware, device. The case where it implement | achieves by etc. is also included. “Apparatus” and “system” are used as synonymous terms. Of course, the “system” does not include a social “mechanism” (social system) that is an artificial arrangement.
In addition, when performing a plurality of processes in each module or in each module, the target information is read from the storage device for each process, and the processing result is written to the storage device after performing the processing. is there. Therefore, description of reading from the storage device before processing and writing to the storage device after processing may be omitted. Here, the storage device may include a hard disk, a RAM (Random Access Memory), an external storage medium, a storage device via a communication line, a register in a CPU (Central Processing Unit), and the like.

The information processing apparatus 100 according to the present embodiment creates teacher data used for machine learning. As shown in the example of FIG. 1, the preliminary teacher data collection module 110, the preliminary teacher data analysis module 120, the teacher The data creation module 130, the collection target data storage module 140, the preliminary teacher data storage module 150, and the teacher data storage module 160 are included.
For example, teacher data used for machine learning is created for the purpose of determining the hobby of the user who made the posting from posting information in SNS (social networking service), etc. .
In machine learning in which data units of teacher data for each case (hereinafter also referred to as teacher data units) are learning / determination targets of data consisting of text sets, the teacher data unit is satisfied when teacher data is collected by clue information. There may be cases where the number of teacher data necessary for learning cannot be collected due to the fact that the data is originally low or because part of the data is limited by security. Further, when teacher data consisting of a text set is collected and created based on clue information, a large amount of information (noise) not related to the clue information may be mixed in the data. For example, the noise in the SNS user's hobby determination includes a greeting such as “Good morning”. Anyone can create such a post, so it is not a clue for judging hobbies.
The teacher data unit is a data amount of one piece of teacher data. The divided data unit is the size of data to be extracted when data is collected in the present embodiment. This is the size of preliminary teacher data to be described later. The preliminary teacher data is data collected in divided data units based on clue information.

  When the information processing apparatus 100 collects data based on clue information, the information processing apparatus 100 collects data obtained by dividing a teacher data unit composed of a text subset (hereinafter also referred to as a divided data unit). Analyzing collected data trends and combining data with similar trends into teacher data units, and using actual data (machine learning model using teacher data generated by this embodiment, processing The teacher data is similar to the target data) and has less noise.

The collection target data storage module 140 is connected to the preliminary teacher data collection module 110. The collection target data storage module 140 stores post information that is a source of teacher data. Here, post information in the SNS (tweet, article in a blog, writing on a bulletin board, or the like) corresponds to the post information. For example, post information may be collected from a user by SNS, or post information collected by SNS may be copied and collected.
The preliminary teacher data collection module 110 is connected to the collection target data storage module 140 and the preliminary teacher data storage module 150. The preliminary teacher data collection module 110 collects post information. Then, the collected post information is stored in the preliminary teacher data storage module 150.
For example, the preliminary teacher data collection module 110 may collect post information (hereinafter also referred to as preliminary teacher data) based on the clue information. Here, the clue information includes a search keyword, a user (writer) attribute (user profile), a forum that provides a place for people with common interests to gather, a community, and the like. For example, when collecting post information of a person who has a hobby of music, post information including a certain singer's name “xxxx” as a search keyword may be searched and collected. You may make it collect the posting information of the user in which "xxxx" is included in the hobby column of the attribute of (writer).
Specifically, the preliminary teacher data collection module 110 collects preliminary teacher data in divided data units composed of text subsets based on the clue information. This makes it possible to use data that is less than the data unit and collect a large amount of data. Further, since the range extracted by the clue information is narrowed, information (noise) unrelated to the clue information is reduced.

The preliminary teacher data storage module 150 is connected to the preliminary teacher data collection module 110 and the preliminary teacher data analysis module 120. The preliminary teacher data storage module 150 stores the preliminary teacher data collected by the preliminary teacher data collection module 110 and passes the preliminary teacher data to the preliminary teacher data analysis module 120.
The preliminary teacher data analysis module 120 is connected to the teacher data creation module 130 and the preliminary teacher data storage module 150. The preliminary teacher data analysis module 120 classifies post information (preliminary teacher data stored in the preliminary teacher data storage module 150) collected by the preliminary teacher data collection module 110.
For example, the preliminary teacher data analysis module 120 may classify post information collected by the preliminary teacher data collection module 110 using classification based on unsupervised clustering or information included in the post information. That is, the preliminary teacher data analysis module 120 classifies (analyzes) the collected preliminary teacher data.

The teacher data creation module 130 is connected to the preliminary teacher data analysis module 120 and the teacher data storage module 160. The teacher data creation module 130 creates teacher data used for machine learning using the posting information in the set classified by the preliminary teacher data analysis module 120. Then, the created teacher data is stored in the teacher data storage module 160.
In addition, the teacher data creation module 130 may combine the posting information in the set classified by the preliminary teacher data analysis module 120 to create one piece of teacher data used for machine learning. That is, the teacher data creation module 130 aggregates (combines) data having the same tendency into data in units of teacher data. As a result, data having the same tendency is integrated, and teacher data similar to actual data is created. In addition, since the information is collected after the clue information is combined, the teacher data including abundant features of the tendency data is created.
The teacher data storage module 160 is connected to the teacher data creation module 130. The teacher data storage module 160 stores the teacher data created by the teacher data creation module 130.

FIG. 2 is an explanatory diagram illustrating an example of a system configuration using the first embodiment.
The information processing apparatus 100, the SNS providing apparatus 210A, the SNS providing apparatus 210B, and the user terminal 220 are connected via a communication line 290, respectively. The communication line 290 may be wireless, wired, or a combination thereof, and may be, for example, the Internet or an intranet as a communication infrastructure. The SNS providing device 210 provides an SNS service and collects post information from the user terminal 220 or the like. Then, the information processing apparatus 100 collects the posting information from the SNS providing apparatus 210A and the SNS providing apparatus 210B and generates teacher data. Further, the function of the information processing apparatus 100 may be realized as a cloud service.
Furthermore, machine learning is performed using the teacher data stored in the teacher data storage module 160 of the information processing apparatus 100. Using the model generated by this machine learning, in the above-described example, the user who has a hobby of music is specified from the posted information in the SNS providing apparatus 210A and the SNS providing apparatus 210B. Then, advertisements for products and services for individuals who have a hobby of music may be provided for the user.

FIG. 3 is a flowchart illustrating a processing example according to the first exemplary embodiment.
In step S302, the preliminary teacher data collection module 110 extracts preliminary teacher data from the collection target data storage module 140 using the clue information.
In step S304, the preliminary teacher data analysis module 120 performs a clustering process on the preliminary teacher data. Specifically, a large number of preliminary teacher data collected in step S302 is clustered and classified into several clusters composed of data having similar tendencies.
In step S306, the teacher data creation module 130 aggregates (combines) the clustered preliminary teacher data into units of teacher data. Specifically, the preliminary teacher data included in the cluster created in step S304 is used to aggregate the data amount required as the teacher data.
In step S308, the teacher data creation module 130 stores the teacher data created in step S306 in the teacher data storage module 160.

A processing example according to the present embodiment will be described by taking SNS user profile automatic determination as an example.
This technology automatically estimates profile attributes such as gender, age, residential area, occupation, and hobby of SNS users from user posts. Automatic determination of the profile is realized by creating a machine learner, but the teacher data is usually 100-200 of one SNS user (this number is an example, if more or less The annotation information is used for the posting information. That is, the teacher data unit is 100-200 post information. This is because it is difficult to estimate the user's profile attribute from a single post information. In the example shown in FIG. 4, the posted information group 420 includes “my car is cool”, “I went to a restaurant”, “I did it, I hit a lottery”, “Now, I returned to the company”, “@xxxx you, There is a sentence (posted information). This is because a plurality of post information (post information group 420) of a user is given a label 410 of “male”, which is specific to men such as “I”, “I did it”, “You” It can be seen that there are post information that appears and post information that does not appear, and it is necessary to collect a large amount of post information per person in order to make an accurate determination.

In the profile determination, the processing according to the present embodiment is applied in the teacher data creation process of “music” having a hobby polarity.
First, the preliminary teacher data collection module 110 collects preliminary teacher data using the name of the singer or the name of the instrument as clue information. The preliminary teacher data group 500 collected under the name of the singer “xxxx” as clue information is shown in the example of FIG. As the preliminary teacher data group 500, “xxxx chan is cute” posted by user a on the first line, “Go to the concert of xxxxx” posted by user b on the second line (^^), Is a "voice is good xxxxxxx" posted by user c, and the fourth line is "user's" posted "xxxxx face is small and has a good style", and the fifth line is posted by user e "I want to sing xxxxxx" In line 6, user f posted “This time I will go to the Osaka concert of xxxxxx”, in line 7, user g posted “I bought a new xxxxxx song!”, And in line 8, user h There is a post “Long eyelashes, xxxx-chan”. In this example, post information per person / person is collected, but a plurality of pieces of post information may be collected per person.

For example, the post information on the first line is a “posture” of “xxxx”, the post information on the second line is “event”, and the post information on the third line is a post focusing on “song”. Thus, it can be seen that the post information collected under the condition of including “xxxx” is data having different tendencies including different aspects (cutting points, properties) of the singer “xxxx”. If such posting information is integrated as it is, it becomes teacher data in which all the tendencies are mixed. The aspect in which the actual user is interested is considered to be different for each user, and one user is considered to focus on one or several specific aspects. For this reason, even if this data including all aspects of characteristics is directly aggregated in units of teacher data, it is expected that the determination accuracy of the machine learner will be lowered. The preliminary teacher data analysis module 120 classifies the collected post data according to those having the same aspect and tendency, and the teacher data creation module 130 aggregates the classified post data within the classification to create teacher data.
As described above, when collecting the collected preliminary teacher data with the same tendency, unsupervised clustering, information originally stored in the data to be collected, or the like is used. As an algorithm for unsupervised clustering, k-means method using words and characters as a feature, potential Dirichlet allocation method (Lent Dirichlet Allocation, LDA, “D. Blei, A. Ng, M. Jordan,“ Lent Dirichlet Allocation ”) , Journal of Machine Learning Research 3 pp. 993-1022, 2003 "), and the Girvan-Newman method. Moreover, as information that the preliminary teacher data originally has and can be used for classification and determination of tendency, for example, posting time in SNS posting, posted location, presence / absence / type of emoticon, attached image There is presence or absence and kind. The preliminary teacher data analysis module 120 classifies the preliminary teacher data based on information originally possessed by the preliminary teacher data. For example, preliminary teacher data in which the posted location is a concert venue may be collected and classified into one category.

  In the example shown in FIG. 6, post information on the singer “xxxx” is collected by “appearance” (appearance cluster 612), “song” (face cluster 622), and “event” (event cluster 632). That is, the preliminary teacher data in the preliminary teacher data group 610 classified into the appearance cluster 612 is combined to generate one piece of teacher data 614. Similarly, the preliminary teacher data in the preliminary teacher data group 620 classified into the face cluster 622 is combined to generate one piece of teacher data 624. By combining the preliminary teacher data in the preliminary teacher data group 630 classified into the event cluster 632, one piece of teacher data 634 is generated. In this way, by collecting post data having the same tendency, teacher data with comprehensive features and less noise is created. In FIG. 6, the user name is included in the preliminary teacher data group 610. However, when one piece of teacher data 614 is generated, preliminary teacher data (such as “xxxxxxxx cute”) is not included without including the user name. Join.

  In the example shown in FIG. 6, in the collection of preliminary teacher data, collection by keyword is used as clue information, but other SNS communities may be used. SNS has users who post to interested communities about matters that they are interested in, and if you collect posts and comments from such communities, you will collect post data on specific matters by different users, similar to the data collected by keywords Can be used as preliminary teacher data.

The example shown in FIG. 7 is a display example of post information posted to the community 700 of “xxxx” and its comments. Comments are also handled as post information (preliminary teacher data).
A contributor icon 715 indicating the organizer of the posting area 710 is displayed, and posting information posted by the user indicated by the contributor icon 717 is displayed in the posting area 710. A comment written by another user is displayed in the comment areas 722, 724, 726, and 728 for the posting area 710. Post information posted by the user indicated by the contributor icon 719 is displayed in the post area 730. A comment written by another user is displayed in the comment area 732 for the posting area 730.
In this case, the preliminary teacher data collection module 110 collects the posting information in the posting area 710, the comment areas 722, 724, 726, 728, the posting area 730, and the comment area 732 as posting information related to “xxxx”.

In addition, FIG. 6 illustrates an example in which posts having the same tendency are aggregated with respect to data collected for “xxxx”. However, as illustrated in the example in FIG. 8, aggregation including posts collected with other keywords is also performed. You may go. For example, a plurality of (especially, many) keywords such as singer “yyyyy”, “zzzzz” are set, a plurality (especially, many) of posting information is collected, and all the preliminary teacher data obtained are collected. Comprehensive analysis may be performed and teacher data may be created by aggregating data with the same tendency. The creation of one piece of teacher data is the same as the example in FIG. 6 except that the preliminary teacher data group (post information) to be aggregated is different.
Even when a community is used as clue information, post information of a large number of communities may be analyzed as preliminary teacher data, and teacher data may be created by aggregating the same trend data.

  In addition, Twitter, which is one of SNSs, has a hash tag as a mechanism that makes it easy to search for similar post information. This is a label that can be freely set by the user by attaching the “#” symbol to any word or sentence. Since popular hashtags are assigned to many posts, it is possible to use hashtags to collect a large amount of preliminary teacher data efficiently as one of the means to collect preliminary teacher data. .

FIG. 9 is a conceptual module configuration diagram of a configuration example according to the second embodiment.
The information processing apparatus 900 includes a teacher data analysis module 970, a teacher data storage module 980, a preliminary teacher data collection module 110, a preliminary teacher data analysis module 120, a teacher data creation module 130, a collection target data storage module 140, and a preliminary teacher data storage module. 150 and a teacher data storage module 160. In addition, the same code | symbol is attached | subjected to the site | part of the same kind as 1st Embodiment, and the overlapping description is abbreviate | omitted. In the second embodiment, a teacher data analysis module 970 and a teacher data storage module 980 are added to the first embodiment.
In the second embodiment, an ideal teacher data group created manually or the like is prepared (teacher data storage module 980), and the data is analyzed by the teacher data analysis module 970. Based on the analysis result, the preliminary teacher data analysis module 120 analyzes the preliminary teacher data. This analysis uses a supervised clustering or a machine learning discriminator. An example of a supervised clustering algorithm is a supervised latent Dirichlet allocation (PLDA). Alternatively, a method of creating a determiner that classifies and determines a tendency by machine learning using ideal teacher data and collects the same tendency may be used. Examples of this algorithm include support vector machine, naive bayes, and adaboost. In the second embodiment, teacher data closer to actual data is created from preliminary teacher data.

The teacher data storage module 980 is connected to the teacher data analysis module 970. The teacher data storage module 980 stores ideal teacher data.
The teacher data analysis module 970 is connected to the teacher data storage module 980 and the preliminary teacher data analysis module 120. The teacher data analysis module 970 analyzes ideal teacher data in the teacher data storage module 980 and reflects it as reference information in the process of combining with the data of the teacher data unit in the preliminary teacher data analysis module 120.
The preliminary teacher data analysis module 120 is connected to the teacher data analysis module 970, the teacher data creation module 130, and the preliminary teacher data storage module 150. The preliminary teacher data analysis module 120 classifies post information collected by the preliminary teacher data collection module 110 by supervised clustering or machine learning using the teacher data created by the teacher data analysis module 970.

FIG. 10 is a flowchart illustrating a processing example according to the second exemplary embodiment.
In step S1002, teacher data is extracted from the teacher data storage module 980.
In step S1004, teacher data is analyzed.
In step S1006, the analysis result is passed to the preliminary teacher data analysis module 120.

FIG. 11 is a flowchart illustrating a processing example according to the second exemplary embodiment.
In step S1102, the preliminary teacher data collection module 110 extracts preliminary teacher data from the collection target data storage module 140.
In step S1104, the preliminary teacher data analysis module 120 performs clustering using the analysis result from the teacher data analysis module 970.
In step S1106, the teacher data creation module 130 aggregates the data into teacher data units.
In step S1108, the teacher data creation module 130 stores the data as teacher data.

  Note that the hardware configuration of the computer on which the program according to the present embodiment is executed is a general computer, specifically a personal computer, a computer that can be a server, and the like, as illustrated in FIG. That is, as a specific example, the CPU 1201 is used as a processing unit (calculation unit), and the RAM 1202, the ROM 1203, and the HD 1204 are used as storage devices. For example, a hard disk or SSD (Solid State Drive) may be used as the HD 1204. A CPU 1201 that executes programs such as the preliminary teacher data collection module 110, the preliminary teacher data analysis module 120, the teacher data creation module 130, and the teacher data analysis module 970, a RAM 1202 that stores the programs and data, and the computer ROM 1203 storing the above-described programs, etc., and an auxiliary storage device (such as a flash memory) having functions such as a collection target data storage module 140, a preliminary teacher data storage module 150, a teacher data storage module 160, and a teacher data storage module 980. HD 1204, a receiving device 1206 for receiving data based on user operations on a keyboard, mouse, touch panel, microphone, etc., CRT, liquid crystal display, speaker An output device 1205 and the like, a communication line interface 1207 for connecting to a communication network such as a network interface card, and, and a bus 1208 for exchanging data by connecting them. A plurality of these computers may be connected to each other via a network.

Among the above-described embodiments, the computer program is a computer program that reads the computer program, which is software, in the hardware configuration system, and the software and hardware resources cooperate with each other. Is realized.
Note that the hardware configuration shown in FIG. 12 shows one configuration example, and the present embodiment is not limited to the configuration shown in FIG. 12, but is a configuration that can execute the modules described in the present embodiment. I just need it. For example, some modules may be configured with dedicated hardware (for example, Application Specific Integrated Circuit (ASIC), etc.), and some modules are in an external system and connected via a communication line In addition, a plurality of systems shown in FIG. 12 may be connected to each other via communication lines so as to cooperate with each other. In particular, in addition to personal computers, portable information communication devices (including mobile phones, smartphones, mobile devices, wearable computers, etc.), information appliances, robots, copiers, fax machines, scanners, printers, multifunction devices (scanners, printers, An image processing apparatus having two or more functions such as a copying machine and a fax machine) may be incorporated.

The program described above may be provided by being stored in a recording medium, or the program may be provided by communication means. In that case, for example, the above-described program may be regarded as an invention of a “computer-readable recording medium recording the program”.
The “computer-readable recording medium on which a program is recorded” refers to a computer-readable recording medium on which a program is recorded, which is used for program installation, execution, program distribution, and the like.
The recording medium is, for example, a digital versatile disc (DVD), which is a standard established by the DVD Forum, such as “DVD-R, DVD-RW, DVD-RAM,” and DVD + RW. Standard “DVD + R, DVD + RW, etc.”, compact disc (CD), read-only memory (CD-ROM), CD recordable (CD-R), CD rewritable (CD-RW), Blu-ray disc ( Blu-ray (registered trademark) Disc), magneto-optical disk (MO), flexible disk (FD), magnetic tape, hard disk, read-only memory (ROM), electrically erasable and rewritable read-only memory (EEPROM (registered trademark)) )), Flash memory, Random access memory (RAM) SD (Secure Digital) memory card and the like.
The program or a part of the program may be recorded on the recording medium for storage or distribution. Also, by communication, for example, a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a wired network used for the Internet, an intranet, an extranet, or a wireless communication It may be transmitted using a transmission medium such as a network or a combination of these, or may be carried on a carrier wave.
Furthermore, the program may be a part of another program, or may be recorded on a recording medium together with a separate program. Moreover, it may be divided and recorded on a plurality of recording media. Further, it may be recorded in any manner as long as it can be restored, such as compression or encryption.

DESCRIPTION OF SYMBOLS 100 ... Information processing apparatus 110 ... Preliminary teacher data collection module 120 ... Preliminary teacher data analysis module 130 ... Teacher data creation module 140 ... Collection target data storage module 150 ... Preliminary teacher data storage module 160 ... Teacher data storage module 210 ... SNS providing apparatus 220 ... User terminal 290 ... Communication line 900 ... Information processing device 970 ... Teacher data analysis module 980 ... Teacher data storage module

Claims (5)

A means of collecting post information;
A classifying means for classifying post information collected by the collecting means;
An information processing apparatus comprising: creation means for creating teacher data used for machine learning using post information in the set classified by the classification means. The information processing apparatus according to claim 1, wherein the classification unit classifies post information collected by the collection unit using classification based on unsupervised clustering or information included in the post information. . The information processing apparatus according to claim 1, wherein the classification unit classifies post information collected by the collection unit by supervised clustering or machine learning using teacher data created in advance. The information processing apparatus according to claim 1, wherein the creation unit creates one piece of teacher data used for machine learning by combining post information in the set classified by the classification unit. Computer
A means of collecting post information;
A classifying means for classifying post information collected by the collecting means;
An information processing program for functioning as a creation means for creating teacher data used for machine learning using post information in a set classified by the classification means.

Images