JP6601888B1 - Information processing apparatus, information processing method, and information processing program - Google Patents

Abstract

An information processing apparatus, an information processing method, and an information processing program capable of estimating the relationship between individual analysis results of different services are provided. An information processing apparatus according to the present application includes an acquisition unit, a conversion unit, and an estimation unit. The acquisition unit acquires feature information including at least one of a model for estimating a user feature using a user attribute and an estimation result regarding the user feature estimated using the attribute from each of different services. The conversion unit converts the feature information acquired by the acquisition unit into a vector value. The estimation unit estimates the relationship between the feature information based on the vector value converted by the conversion unit. [Selection] Figure 3

Description

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

  2. Description of the Related Art Conventionally, for example, there are information processing apparatuses that provide various services to a user and estimate the user's characteristics by analyzing user attributes obtained by using the service. In this type of information processing apparatus, a service provided by the information processing apparatus may be optimized based on the estimated user characteristics.

Japanese Unexamined Patent Publication No. 2018-032252

  However, the technology described above analyzes the user's characteristics independently for each service and applies the analysis result to the optimization of its own service. Consideration is given to using the analysis results of other services. Was not. For this reason, it is difficult to estimate the relationship between individual analysis results of different services.

  The present application has been made in view of the above, and an object thereof is to provide an information processing apparatus, an information processing method, and an information processing program capable of estimating the relationship between individual analysis results of different services.

  An information processing apparatus according to the present application includes an acquisition unit, a conversion unit, and an estimation unit. The acquisition unit acquires feature information including at least one of a model for estimating the user's characteristics using the user's attributes and an estimation result regarding the user's characteristics estimated using the attributes from different services. To do. The conversion unit converts the feature information acquired by the acquisition unit into a vector value. The estimation unit estimates a relationship between the feature information based on the vector value converted by the conversion unit.

  According to one aspect of the embodiment, there is an effect that the relationship between individual analysis results of different services can be estimated.

FIG. 1 is a diagram illustrating an example of information processing according to the embodiment. FIG. 2 is a diagram illustrating a configuration example of the information processing system according to the embodiment. FIG. 3 is a block diagram illustrating a configuration example of the information processing apparatus according to the embodiment. FIG. 4 is a diagram illustrating an example of shared information according to the embodiment. FIG. 5 is a diagram illustrating an example of service information according to the embodiment. FIG. 6 is a diagram illustrating an example of model program information according to the embodiment. FIG. 7A is a process explanatory diagram (part 1) of the information processing according to the embodiment. FIG. 7B is a process explanatory diagram (part 2) of the information processing according to the embodiment. FIG. 7C is a process explanatory diagram (part 3) of the information processing according to the embodiment. FIG. 7D is a process explanatory diagram (part 4) of the information processing according to the embodiment. FIG. 7E is a process explanatory diagram (part 5) of the information processing according to the embodiment. FIG. 7F is a process explanatory diagram (No. 6) of the information processing according to the embodiment. FIG. 7G is an explanatory diagram (part 7) of the information processing according to the embodiment. FIG. 8 is a flowchart (part 1) illustrating a processing procedure executed by the information processing apparatus according to the embodiment. FIG. 9 is a flowchart (part 2) illustrating a processing procedure executed by the information processing apparatus according to the embodiment. FIG. 10 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing apparatus according to the embodiment.

  Hereinafter, a mode for carrying out an information processing apparatus, an information processing method, and an information processing program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the information processing apparatus, the information processing method, and the information processing program according to the present application are not limited by this embodiment. In addition, the embodiments can be appropriately combined within a range that does not contradict processing contents. In the following embodiments, the same portions are denoted by the same reference numerals, and redundant description is omitted.

[1. Example of information processing)
First, an example of information processing according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of information processing according to the embodiment. In FIG. 1, a model in which the information processing apparatus 1 according to the present application included in the information processing system S estimates the characteristics of the user U using the attributes of the user U, and the characteristics of the user U estimated using the attributes described above. Obtains feature information including at least one of the estimation results for each from different services, converts the obtained feature information into vector values, and executes a process of estimating the relationship between the feature information based on the converted vector values An example is shown.

  As illustrated in FIG. 1, the information processing system S according to the embodiment includes an information processing apparatus 1, a service providing server 10 (10-1, 10-2, 10-3), and a user terminal 100.

  The user terminal 100 is a terminal device owned by the user U. The user U uses the service Y (Y1, Y2, Y3) provided by each service providing server 10 by operating the user terminal 100 to access the service providing server 10 (steps S1-1, S1-2). , S1-3).

  The service providing server 10 is a server device that provides various services Y to the user U. The service Y provided by the service providing server 10 is a service that distributes information through various applications installed in the user terminal 100, for example. Such services include, for example, news providing services, auction services, weather forecast services, shopping services, financial transactions (stock transactions, etc.) services, route search services, map providing services, travel services, restaurant introduction services, blogs There are services. Note that the service Y is not limited to a service that distributes information via an application, and may be a service that provides information to the user U via a browser, for example.

  Each service providing server 10 estimates the characteristics of the user U by analyzing the attributes of the user U using the service Y. Specifically, each service providing server 10 uses at least one of a model for estimating the characteristics of the user U using the attributes of the user U and an estimation result regarding the characteristics of the user U estimated using the attributes of the user U. Is generated (steps S2-1, S2-2, and S2-3).

  The attribute referred to here is an attribute determined by the user U or a feature of the user U. For example, action information regarding the action of the user U when using the service Y such as a search log or purchase history, It is a concept including demographic attributes such as sex, age, address, etc., and psychographic attributes such as life cycle, values, personality, purchasing motivation, and the like.

  The characteristics of the user U are properties of the user U estimated by including various attributes of the user U, or estimated attributes. For example, the user U's interests, hobbies, and behaviors Includes patterns (search behavior, purchasing behavior, etc.).

  Moreover, the above-described model in the feature information can input the attribute of the user U and output the feature of the user U as a score, for example. Further, the above estimation result in the feature information is, for example, a score output by the model. The estimation result may be information that does not depend on the model, for example, information obtained by simply quantifying the attribute of the user U (for example, a score that increases or decreases according to the number of searches). In the example illustrated in FIG. 1, the service Y <b> 1 as a shopping service generates feature information including a model “fishing rod” of a user U who purchased a fishing rod through online shopping and a score “0.5” as an estimation result.

  In addition, the service Y2 as a search service generates feature information including the model “fishing spot” of the user U who searched for the fishing spot and the score “0.4” as an estimation result. In addition, the service Y3 as a car auction service generates feature information including a model “Luxury Car Ownership” of the user U who owns a luxury car and a score “0.9” as an estimation result. In FIG. 1, the higher the score, the stronger the characteristics of the user U with respect to the thing indicated by the model (for example, strong interest, preference, and relevance).

  Moreover, although the above-mentioned score showed the case where it was a continuous value, a discrete value may be sufficient, for example. Specifically, whether or not a fishing spot has been searched may be represented by a discrete value (ie, Boolean value) score of either “1” (present) or “0” (not present). Alternatively, the estimation result is not limited to the score shown as an example, and may be text information.

  Here, conventionally, the feature information individually generated by the services Y1, Y2, and Y3 is only used for the optimization of its own service, and the use of the feature information of other services is not considered.

  For this reason, the feature information of other services cannot be used with each other, and the relationship between the individual analysis results of different services cannot be estimated.

  Therefore, the information processing apparatus 1 according to the embodiment is configured such that the feature information generated by each service Y can be shared by other services Y. Specifically, the information processing apparatus 1 according to the embodiment acquires the above-described feature information from each different service Y (steps S3-1, S3-2, and S3-3), and associates the feature information for each user. Generate shared information. The information processing apparatus 1 according to the embodiment is not limited to acquiring feature information including the above-described model and a score that is an estimation result, and includes feature information including only the model or feature information including only the score. May be obtained.

  Subsequently, the information processing apparatus 1 converts the acquired feature information into a vector value (step S4). For example, the information processing apparatus 1 uses a user ID and a score of each feature information as a data set, and converts each feature information into a multi-dimensional real value vector using a machine learning method or the like using a neural network. (Embedding). Here, as can be seen from the fact that the user ID and score of each feature information are used as a data set, in such embedding, from the viewpoint of “what user U has what score of what model”, the user ID and score The combination becomes the main feature element (so-called feature in machine learning). Of course, for example, information representing the characteristics of the model such as a model name or a model description may be added as a feature.

  And the information processing apparatus 1 estimates the relationship between feature information based on this vector value (step S5). For example, the information processing apparatus 1 determines the similarity between feature information based on the vector value. This makes it possible to determine the proximity of feature information. For example, the information processing apparatus 1 calculates the cosine similarity of these vector values for two pieces of feature information, and if the similarity is equal to or greater than a predetermined value, for example, the user U not included in one feature information Can be handled as having a feature corresponding to the one feature information. That is, when the information processing apparatus 1 determines that the similarity is close, if the user U has a feature corresponding to one of the feature information, the user U also uses the other feature information. Estimated as having the relevant characteristics.

  Thereby, the relationship between the individual analysis results of different services can be estimated, and the individual analysis results of different services can be used across the basis of the estimation results.

  For example, the information processing apparatus 1 can add and subtract vectorized feature information and estimate the feature information indicated by the calculation result. As a result, it is possible to estimate feature information that has been difficult to predict by hand alone from a combination of a plurality of feature information. Further, by newly extracting the user U included in the estimated feature information, for example, the coverage in marketing can be expanded.

  When there is a request from each service Y side, the information processing apparatus 1 provides the provision information based on the feature information shared as vector information and vectorized (steps S6-1, S6-2, S6-). 3). For example, there are the following two methods for providing the provision information.

  The first is a method in which the service providing server 10 side designates a specific user U and provides a feature (estimation result) of the user U based on shared information. The second is a method in which the service providing server 10 specifies a specific feature and provides a list of users U estimated to have the feature based on shared information and vector information.

  A specific example of the estimation process in step S5 and the provision process in steps S6-1, S6-2, and S6-3 will be described later with reference to FIGS. 7A to 7G.

  As described above, in the information processing method according to the embodiment, at least one of a model for estimating the characteristics of the user U using the attributes of the user U and an estimation result regarding the characteristics of the user U estimated using the attributes is obtained. The included feature information is acquired from each of the different services Y1, Y2, and Y3, the acquired feature information is converted into a vector value, and the relationship between the feature information is estimated based on the converted vector value.

  Therefore, according to the information processing method according to the embodiment, it is possible to estimate the relationship between individual analysis results of different services.

  Hereinafter, the information processing apparatus 1 that performs the information processing as described above and the information processing system S including the information processing apparatus 1 will be described in detail.

[2. Configuration of information processing system S]
Next, the configuration of the information processing system S according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating a configuration example of the information processing system S according to the embodiment. As illustrated in FIG. 2, the information processing system S according to the embodiment includes an information processing device 1, a plurality of service providing servers 10-1 to 10-n, a plurality of business operator terminals 20, and a plurality of user terminals. 100-1 to 100-n.

  These various devices are communicably connected via a network N by wire or wireless. The network N is a communication network such as a LAN (Local Area Network), a WAN (Wide Area Network), a telephone network (a mobile phone network, a fixed telephone network, etc.), a regional IP (Internet Protocol) network, and the Internet. The network N may include a wired network or a wireless network.

  In FIG. 2, the plurality of business operator terminals 20 are illustrated as being directly connected to the service providing servers 10-1 to 10-n, but the business operator terminal 20 is connected to the service providing servers 10-1 to 10-1. It schematically shows that it is under 10-n, and does not limit the connection form.

  The user terminal 100 is a terminal device used by the user U as described above. The user terminal 100 is an information processing apparatus such as a mobile phone including a smartphone, a tablet terminal, a desktop PC, a notebook PC, or a PDA (Personal Digital Assistant). The user terminal 100 also includes a wearable device that is an eyeglass-type or clock-type information processing terminal.

  The user terminal 100 acquires and acquires various types of information in accordance with operations performed by the user U and functions of the user terminal 100 (for example, a function for executing an application for using the service Y, a browser function, and the like). Generate and send information according to the information. For example, the user terminal 100 accesses the site of the service Y provided by the service providing server 10 via the network N. The user terminal 100 transmits a service provision request to the service provision server 10 by designating information necessary for the user U to receive the service, for example.

  The information processing apparatus 1 is a server apparatus that functions as an upper server of the service providing server 10. The information processing apparatus 1 acquires feature information from each service providing server 10 to generate shared information, and vectorizes each feature information in the shared information. Further, the information processing apparatus 1 estimates the relationship between the feature information based on the vectorized feature information, and provides various types of information to the service providing server 10 side based on the estimation result.

  The service providing server 10 is a server device that provides the service Y to the user U as described above. For example, the service providing server 10 analyzes the attribute of the user U using the service Y and generates feature information. Further, the service providing server 10 requests the information processing apparatus 1 for providing information. The request for the provision information is transmitted from the service provision server 10 when, for example, the provider of the service Y operates the provider terminal 20 connected to the service provision server 10.

  The business entity terminal 20 is a terminal device used by a business operator who operates the service Y. The business entity terminal 20 is, for example, a desktop PC, a notebook PC, or the like, but may be a mobile phone, a tablet terminal, a PDA, a wearable device, or the like. The business entity terminal 20 requests the information to be provided to the information processing apparatus 1 by designating a specific user U or a specific feature or the like from the shared information use service screen provided by the information processing apparatus 1.

  In the information processing system S, the information processing apparatus 1, the plurality of service providing servers 10-1 to 10-n, and the plurality of business operator terminals 20 are configured to have, for example, one business operator. That is, one provider can provide a plurality of services Y. In addition, one business operator aggregates the feature information generated by each service Y into the information processing apparatus 1 as shared information, obtains provided information based on the shared information, and performs marketing of the user U in each service Y, for example. Used for surveys, etc.

  The information processing apparatus 1 and the plurality of service providing servers 10-1 to 10-n may be configured by different operators. In other words, another service Y when viewed from a certain service Y of a certain business may be another service Y of the same business, or may be a service Y of another business. Further, the plurality of service providing servers 10-1 to 10-n may be configured by different business operators. In addition to the information processing system S described above, a server of a business operator that only uses the provided information may be connected.

  In FIG. 2, the information processing apparatus 1 is configured separately from the service providing server 10 and has functions of generating shared information, vectorizing, estimating the relationship between feature information, and providing information. Although the case has been shown, for example, any one of the service providing servers 10 may perform part or all of the functions of the information processing apparatus 1 instead.

  In FIG. 2, the information processing apparatus 1 shows a case in which all functions of shared information generation, vectorization, estimation of the relationship between feature information, and information provision are combined. The processing device 1 may perform the above functions in a shared manner.

[3. Information processing apparatus 1]
Next, a configuration example of the information processing apparatus 1 will be described with reference to FIG. FIG. 3 is a block diagram illustrating a configuration example of the information processing apparatus 1 according to the embodiment. In FIG. 3, only components necessary for the description of the information processing apparatus 1 are shown, and descriptions of general components are omitted.

  As illustrated in FIG. 3, the information processing apparatus 1 includes a communication unit 2, a storage unit 3, and a control unit 4. The information processing apparatus 1 includes an input unit (for example, a keyboard and a mouse) that receives various operations from an administrator who uses the information processing apparatus 1, and a display unit (for example, a liquid crystal display) that displays various types of information. ).

(About communication unit 2)
The communication unit 2 is realized by, for example, a NIC (Network Interface Card). The communication unit 2 is connected to the network N in a wired or wireless manner, and transmits / receives information to / from the service providing server 10, the provider terminal 20 connected to the service providing server 10, and the user terminal 100 via the network N. I do.

(About storage unit 3)
The storage unit 3 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. In the example of FIG. Shared information 31, service information 32, model program information 33, and vector information 34 are stored.

(About shared information 31)
The shared information 31 is information including feature information acquired from the service providing server 10. FIG. 4 is a diagram illustrating an example of the shared information 31. As illustrated in FIG. 4, the shared information 31 is, for example, a table in which the column item is “UID” and the line item is “feature information”. A score that is an output of the model in the feature information is input to each set of data items of the column item and the row item.

  “UID” is identification information (user ID) for identifying the user U. “Characteristic information” is information including at least one of a model and a score generated by each service providing server 10, and is information indicating the characteristics of the user U. In the example illustrated in FIG. 4, a model name such as “fishing rod” is input to each item of “feature information”.

  As shown in FIG. 4, the shared information 31 is information including “UID” and “feature information” scores. In the example illustrated in FIG. 4, the user U whose UID is “U1” indicates that the score of the model “fishing rod” in the feature information is “0.5”.

  For the user U with the UID “U2”, the score of the model “fishing rod” in the feature information is “not input”. This indicates that the score of the model “fishing rod” of the user U who is “U2” has not been acquired from the service providing server 10.

  For example, the score calculated by the service providing server 10 may be input as the score of the shared information 31 as it is, or the score calculated by the service providing server 10 is normalized according to a predetermined standard. The shared information 31 may be input.

  In FIG. 4, the shared information 31 is shown as a score. However, the shared information 31 is not limited to the score, and is divided into stages such as “◯, Δ and x”, “excellent, good and acceptable”, and the like. Information may be provided.

(About service information 32)
Next, the service information 32 is information on the service Y provided by each service providing server 10. FIG. 5 is a diagram illustrating an example of the service information 32. As shown in FIG. 5, the service information 32 includes items such as “service ID”, “provided service”, and “number of registered models”.

  “Service ID” is identification information for identifying the service Y (or the service providing server 10). “Provided service” is information indicating the contents of the service Y. “Number of registered models” is information indicating the number of models of feature information registered in the shared information 31.

(About model program information 33)
Next, the model program information 33 is information including model program data in the feature information. FIG. 6 is a diagram illustrating an example of the model program information 33. The model program information 33 is information including model program data in the feature information acquired by the acquisition unit 41 described later. As shown in FIG. 6, the model program information 33 includes items such as “model ID”, “model name”, “service ID”, and “model data”.

  “Model ID” is identification information for identifying a model in the feature information. “Model name” is identification information indicating the name of the model in the feature information. “Service ID” is identification information for identifying the service Y (or the service providing server 10). “Model data” is program data of a model in feature information.

  “Model data” is, for example, a regression model that uses the attribute of the user U as an explanatory variable. For example, “model data” is a regression model represented by the following equation. That is, it can be expressed as a regression model of y = ω1 · x1 + ω2 · x2 +... + Ωn · xn.

In the above formula, “x” is an explanatory variable corresponding to the attribute of the user U. “Y” is an objective variable corresponding to a score indicating the characteristics of the user U. “Ω” is a coefficient of “x” and indicates a predetermined weight value. Specifically, “ω1” is a weight value of “x1”, “ω2” is a weight value of “x2”, and “ωn” is a weight value of “xn”. As described above, the above formula is created by combining the variable (for example, “ω1 · x1 _” ) including the explanatory variable “x” corresponding to the attribute of the user U and the predetermined weight value “ω”. .

  The “model data” may be feature information generated using SVM (Support Vector Machine) or other machine learning methods. The feature information may be generated using a deep learning technique. For example, the service Y can generate feature information by appropriately using various deep learning techniques such as DNN (Deep Neural Network), RNN (Recurrent Neural Network), and CNN (Convolutional Neural Network).

(About vector information 34)
Returning to the description of FIG. 3, the vector information 34 will be described. The vector information 34 is information related to vectorization of feature information registered in the shared information 31. The vector information 34 is generated by a conversion unit 42 described later, and includes, for example, a vector value for each feature information.

(About the control unit 4)
The control unit 4 is a controller, for example, various programs (information processing programs) stored in a storage device inside the information processing apparatus 1 by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. (Corresponding to an example) is implemented by using the RAM as a work area. The control unit 4 is a controller, for example, and is realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

  As illustrated in FIG. 3, the control unit 4 includes an acquisition unit 41, a conversion unit 42, a reception unit 43, an estimation unit 44, an extraction unit 45, and a provision unit 46. Information described below Implement or execute a processing function or action. The internal configuration of the control unit 4 is not limited to the configuration illustrated in FIG. 3, and may be other configurations as long as the information processing described below can be performed. Further, the connection relationship between the processing units included in the control unit 4 is not limited to the connection relationship illustrated in FIG. 3, and may be another connection relationship.

  The control unit 4 registers the feature information acquired from the service providing server 10 in the shared information 31 and vectorizes each feature information. Further, the control unit 4 estimates the relationship between the feature information based on the vectorized feature information, and provides various information to the service Y.

(About the acquisition unit 41)
The acquisition unit 41 uses a different service to obtain feature information including at least one of a model for estimating the features of the user U using the attributes of the user U and an estimation result regarding the features of the user U estimated using the attributes of the user U. Obtain from each Y. For example, the acquisition unit 41 acquires feature information generated in a format specified in advance for the service Y. Thereby, since the format of the feature information can be made uniform among the services Y, it is possible to acquire the feature information based on the standard.

  Moreover, the acquisition part 41 acquires the information containing an above-described model and the initial value of the score output from this model as feature information. For example, the initial value of the score may be calculated in advance by the service providing server 10 or may be calculated by the acquisition unit 41 acquiring model data. When the acquisition unit 41 calculates the initial value, only the model data of the feature information is acquired first, and then the attribute of the user U is acquired from the service providing server 10 to calculate the score.

  The acquisition unit 41 may have a check function for checking whether or not the feature information is generated according to a specified format when the feature information is acquired.

  The acquisition unit 41 registers the acquired feature information as the shared information 31 and performs an update process for updating the registered feature information as necessary. Specifically, at the first time, the acquisition unit 41 first acquires the model data of the feature information and the initial value of the score and registers it as the shared information 31. For the second and subsequent times, the service Y uses the user as necessary. The attribute of U is acquired, and the score of the shared information 31 is updated based on this. When the model data is changed, new model data is acquired from the service Y and the model of the shared information 31 is updated.

  In addition, the acquisition unit 41 may acquire the attribute of the user U as an input of the model in the feature information from a service Y different from the service Y from which the model is acquired. That is, the user's U attribute may be acquired from the service Y of the model acquisition source registered as the shared information 31 and the score may be output, or the user U acquired from the service Y different from the acquisition source service Y A score may be output using the attribute of. Thereby, for example, the model registered in the shared information 31 can be shared between services Y of different operators.

(About the converter 42)
The conversion unit 42 converts each feature information in the shared information 31 into a vector value. For example, as described above, the converting unit 42 uses the user ID and score of each feature information as a data set, and converts each feature information into a multi-dimensional real value vector using a machine learning method using a neural network. Convert.

  In such a case, for example, if there are 200 million user IDs, the conversion unit 42 does not calculate a vector value consisting of 200 million dimensions as it is, but extracts only significant dimensions, for example, hundreds to tens of dimensions. A vector value that is dimensionally compressed up to is calculated. Thereby, noise components and the like can be eliminated, and the feature information can be vectorized with high accuracy.

(About the reception unit 43)
The accepting unit 43 accepts various requests from the service providing server 10. For example, the accepting unit 43 accepts a feature designation request for designating a specific feature from the service Y. The feature designation designates the model ID, model name, etc. of the feature information included in the shared information 31. The designation by feature designation may be one piece of feature information or a plurality of pieces of feature information. Alternatively, the feature designation may be a designation including a plurality of feature information such as a field designation.

  The accepting unit 43 accepts a user designation request for designating a specific user U from the service Y. One or more users U may be designated by user designation. Alternatively, the user designation may be a user group designation including a plurality of users U.

  In addition, the reception unit 43 may include a period specification in the above request. For example, in the case of feature designation, feature information registered in the period designated by the period designation may be used as the feature designation target. In the case of user designation, the user U who uses the service Y during the period designated by the period designation may be the user designation target.

  In addition, the reception unit 43 receives a calculation designation request for adding and subtracting vectorized feature information from the service Y.

(About the estimation unit 44)
The estimation unit 44 estimates the relationship between the feature information based on the vector value of each feature information included in the vector information 34. For example, the estimation unit 44 calculates a cosine similarity from the vector value of each feature information, and estimates the proximity of each feature information based on the similarity.

  Moreover, the estimation part 44 adds and subtracts the vectorized feature information, and estimates the feature information which the calculation result shows. At this time, if the feature information indicated by the calculation result is registered in the shared information 31, the estimation unit 44 derives the feature information as the estimation result. In addition, if the corresponding feature information is not registered in the shared information 31, the estimation unit 44 derives the feature information closest to the calculation result as the estimation result.

(About the extraction unit 45)
The extraction unit 45 extracts the user U corresponding to the specific feature designated by the feature designation. For example, the extraction unit 45 extracts a user U whose score is a predetermined value or more in the feature information corresponding to the designated feature.

  In addition, even if the user U is not included in the feature information corresponding to the specified feature, the extraction unit 45 has a score in other feature information estimated to be close to the feature information by the estimation unit 44. It is also possible to extract a user U having a predetermined value or more.

  In other words, when a specific feature is specified by the above-described feature specification, the extraction unit 45 has, in addition to feature information corresponding to the specific feature, other features estimated by the estimation unit 44 as close to the feature information. The information can be combined and set as an information extraction target.

  In addition, the extraction unit 45 extracts the feature of the specific user U designated by the user designation as the extraction target corresponding feature information. For example, the extraction unit 45 extracts the feature information regarding the specified user U that has a score equal to or greater than a predetermined value as indicating the feature of the user U.

  At this time, even if the feature information that does not include the specified specific user U is included in the extraction unit 45, as long as it is estimated by the estimation unit 44 to be close to the feature information indicating the feature of the user U, These can also be extracted as indicating the characteristics of the user U.

  That is, when a specific user U is specified by the above user specification, the extraction unit 45 is estimated to be close to the feature information by the estimation unit 44 in addition to the feature information including the specific user U. In addition, other feature information can be combined as information extraction targets.

  In addition, the extraction unit 45 extracts the feature information estimated by the estimation unit 44 as a result of the calculation specified by the calculation specification.

(About the providing unit 46)
The providing unit 46 provides provided information to the service Y based on various information extracted by the extracting unit 45. For example, the providing unit 46 generates providing information including a list of the users U extracted by the extracting unit 45 based on the feature designation and provides the providing information to the service Y. The list may be sorted in any order such as score order, alphabetical order, alphabetical order, or the like.

  Alternatively, the providing unit 46 may provide only some of the users U extracted by the extracting unit 45 to the service Y as providing information. Specifically, the providing unit 46 may provide providing information in which a user U or the like having a higher score is selected.

  Further, the providing unit 46 provides the service Y with provided information including information indicating the characteristics of the user U extracted by the extracting unit 45 based on the user designation. For example, the providing unit 46 provides providing information including information in which the features of the user U extracted by the extracting unit 45 are arranged in a list.

  The providing unit 46 may provide providing information including only some of the features of the user U extracted by the extracting unit 45. For example, the providing unit 46 may provide providing information in which features or the like indicated by feature information having higher scores are selected.

  Further, the providing unit 46 provides the service Y with the provided information including the feature information extracted by the extracting unit 45 based on the calculation designation.

(Specific example of information processing according to the embodiment)
Next, information processing according to the embodiment described so far will be described more specifically with reference to FIGS. 7A to 7G. 7A to 7G are explanatory diagrams (No. 1) to (No. 7) of information processing according to the embodiment.

  FIG. 7A very schematically shows the positional relationship between the pieces of feature information estimated by the estimation unit 44. In the following, as shown in FIG. 7A, each piece of feature information may be identified by the above model name (such as “owns a luxury car”).

  As shown in FIG. 7A, it is assumed that there are models “Luxury Car Ownership”, “Low Annual Income”, and “High Annual Income”. Each of these is vectorized by the conversion unit 42, and based on the vector value, as shown in FIG. 7A, it is estimated that the models “Luxury Car Ownership” and “High Annual Income” are closely related by the estimation unit 44. Shall. As shown in FIG. 7A, it is assumed that the model “owns a luxury car” includes data of the user U1, and the models “low income” and “high income” do not include data of the user U1.

  In such a case, the models “low annual income” and “high annual income” without user U1 data cannot indicate whether the user U1 has the characteristics of “low annual income” or “high annual income”. .

  However, if the user U1 has a high score in a model “Luxury Car Owner” with data of the user U1 and the models “Luxury Car Owner” and “High Income” are close as described above, the user U1 It can be estimated that the model “high annual income” also shows a high score, that is, shows the characteristics of the model “high annual income”.

(Specific example of feature specification)
Based on such an example, a specific example of designation by the above-described feature designation will be described with reference to FIGS. 7B and 7C. 7B to 7D, FIG. 7F, and FIG. 7G show a part of the usage screen image of “inter-service usage service” displayed on the operator terminal 20, for example.

  For example, the designation method by feature designation is performed by checking the check box of the model name of each feature information displayed as a category, as shown in FIG. 7B. For example, the reception unit 43 refers to the shared information 31, generates a usage screen in which the model name of each registered feature information can be specified with a check box, transmits the usage screen to the service providing server 10 side, and transmits it to the operator terminal 20. Display.

  7B shows an example in which the models “Luxury Car Ownership” and “High Annual Income” illustrated in FIG. 7A can be designated. As shown in FIG. 7B, for example, when the check box of “Luxury car ownership” is checked and the “Extract” button is pressed on the operator terminal 20, the extraction unit 45 obtains a score for the model “Luxury car ownership”. A user U that is equal to or greater than a predetermined value is extracted.

  Then, in response to the extraction result of the extraction unit 45, the providing unit 46 generates, for example, provision information “<< user extraction result >>” including a list of the extracted users U, and transmits it to the service providing server 10 side. And displayed on the operator terminal 20 (see the lower part of the usage screen in FIG. 7B).

  Here, according to the example of FIG. 7A, since the model “owns a luxury car” includes the data of the user U1, as shown in FIG. 7B, the user U1 with the UID “U1” is included in the list of provided information. Will be included. As shown in the figure, the score “0.9” of the user U1 may be displayed together.

  On the other hand, as illustrated in FIG. 7C, for example, when the check box of “high income” is checked and the “extract” button is pressed on the operator terminal 20, the extraction unit 45 determines the score in the model “high income”. The user U whose is greater than or equal to a predetermined value is extracted.

  Here, as illustrated in FIG. 7A, the model “high income” does not include the data of the user U1. However, since it is estimated by the estimation unit 44 that the models “high annual income” and “luxury car ownership” are close to each other, the extraction unit 45 determines not only the designated model “high annual income” but also the model “ The user U whose score is equal to or higher than a predetermined value in “owning a luxury car” is extracted.

  As a result, as shown in FIG. 7C, for example, the providing unit 46 includes the user U1 that is not included in the model “high income” but is estimated to have the characteristics of “high income” in the list. Is transmitted to the service providing server 10 and displayed on the operator terminal 20. As shown in the figure, the model name “* owning luxury car” from which the user U1 is extracted may be displayed together.

  Thus, by extracting together the user U having another feature close to the feature designated by the feature designation, for example, the operator operating the service Y can increase the coverage in marketing or the like. Become.

(Specific example of user designation)
Subsequently, a specific example of the designation by the user will be described with reference to FIG. 7D. For example, the designation method by user designation is performed by designating identification information of the user U as a feature extraction target, as shown in FIG. 7D.

  For example, the reception unit 43 generates a usage screen that can specify a user ID, transmits the usage screen to the service providing server 10 side, and causes the provider terminal 20 to display the usage screen. As illustrated in FIG. 7D, for example, when the user ID “U1” is designated and the “extract” button is pressed on the operator terminal 20, the extraction unit 45 includes the feature information including the data of the user U1. , Those having a score equal to or greater than a predetermined value are extracted as indicating the characteristics of the user U1.

  Further, even if the feature information has no data of the user U1, the extraction unit 45 combines the feature information estimated by the estimation unit 44 with the feature information indicating the feature of the user U1 together with the feature of the user U1. Extract as indicating.

  Therefore, according to the example shown in FIG. 7A, the extraction unit 45 extracts “high income” in addition to the model “owning a luxury car” as indicating the characteristics of the user U1.

  Then, in response to the extraction result of the extraction unit 45, the providing unit 46 provides the provided information “<< feature extraction result >>” including the model names of the extracted models “luxury car possession” and “high income”, for example. It is generated, transmitted to the service providing server 10 side, and displayed on the provider terminal 20 (see the lower part of the usage screen in FIG. 7D).

  Thus, by extracting another feature close to the feature of the user U designated by the user designation, it is possible to expand the coverage in marketing or the like.

(Overview of operation specification)
Next, a specific example of designation by the above calculation designation will be described with reference to FIGS. 7E to 7G. First, the outline of calculation designation will be described with reference to FIG. 7E, taking Word2Vec known as natural language processing as an example.

  Word2Vec is known as an example of embedding information as in the information processing method according to the embodiment. Word2Vec reads a large amount of text into a neural network and expresses a word by, for example, a vector of several hundred dimensions (hereinafter referred to as “word vector”). In such Word2Vec, it is possible to add and subtract meanings by adding and subtracting word vectors.

  Specifically, as illustrated in FIG. 7E, for example, a word vector representing “queen” is obtained by subtracting a word vector representing “male” from a word vector representing “king” and adding a word vector representing “female”. It is known that a word vector close to can be obtained. It is also known that when a word vector representing “Paris” is subtracted from a word vector representing “France” and a word vector representing “Tokyo” is added, a word vector close to the word vector representing “Japan” is obtained. ing.

(Specific example of operation specification)
Since the information processing method according to the embodiment is obtained by vectorizing feature information as in the Word2Vec, it is possible to specify calculation between the feature information.

  For example, the designation method by calculation designation is performed by designating each item of the calculation formula as shown in FIG. 7F. For example, the reception unit 43 generates a usage screen that can specify each item of the arithmetic expression, transmits the usage screen to the service providing server 10 side, and causes the operator terminal 20 to display the usage screen.

  As shown in FIG. 7F, on such a usage screen, a model name and a calculation symbol of feature information to be calculated can be selected and specified by a GUI (Graphical User Interface) such as a drop-down list. The model name and operation symbol designation fields can be added by clicking “Add” in the figure, for example.

  When each item of the arithmetic expression is specified and the “execute” button is pressed, the estimation unit 44 executes a vector operation using the specified arithmetic expression. Then, when the feature information indicated by the calculation result is registered in the shared information 31, the estimation unit 44 notifies the providing unit 46 of the feature information. In addition, when the feature information indicated by the calculation result is not registered in the shared information 31, the estimation unit 44 notifies the providing unit 46 of the feature information closest to the calculation result.

  In response to the notification from the estimation unit 44, the providing unit 46 generates provided information “<< calculation result >>” including, for example, the model name of the notified feature information, and transmits the generated information to the service providing server 10 side. It is displayed on the provider terminal 20 (see the lower part of the usage screen in FIG. 7F).

  For example, as shown in FIG. 7F, it is assumed that an arithmetic expression “fishing lover-river fishing” is executed. In such a case, considering the meaning of each feature, for example, “sea fishing” is predicted as one of the calculation results, but the estimation unit 44 searches the vector information 34 for feature information closest to the execution result of the calculation, This is notified to the providing unit 46.

  Here, if the vector information 34 has feature information that is closest to the execution result of the calculation and represents “sea fishing” (see “with sea fishing” in the figure), “<< operation result >>” "Sea fishing" will be displayed.

  On the other hand, as shown in FIG. 7G, if the feature information indicating “sea fishing” does not exist in the vector information 34 by the same calculation as in FIG. 7F (see “no sea fishing” in the figure), Characteristic information closest to the execution result of the calculation (here, “purchase of marine fishing gear” as an example) is displayed as “<< calculation result >>”.

  By such calculation designation, for example, feature information suitable for a desired prediction result can be extracted from any combination of feature information, but on the contrary, feature information that could not be predicted manually can also be extracted. It becomes. Therefore, it is possible to enhance and expand coverage in marketing and the like by such calculation designation.

[4. Processing procedure of information processing apparatus 1]
Next, a processing procedure executed by the information processing apparatus 1 according to the embodiment will be described. FIG. 8 is a flowchart (part 1) illustrating a processing procedure executed by the information processing apparatus 1 according to the embodiment. FIG. 9 is a flowchart (part 2) illustrating a processing procedure executed by the information processing apparatus 1 according to the embodiment. FIG. 8 shows a processing procedure until the feature information is acquired and the relationship between the feature information is estimated. FIG. 9 shows a processing procedure of the providing process.

  As illustrated in FIG. 8, first, the acquisition unit 41 at least one of a model that estimates the characteristics of the user U using the attributes of the user U and an estimation result related to the characteristics of the user U estimated using the attributes. Is acquired from each of the different services Y (step S101).

  Then, the conversion unit 42 converts the feature information acquired by the acquisition unit 41 into a vector value (step S102).

  And the estimation part 44 estimates the relationship between feature information based on the vector value converted by the conversion part 42 (step S103), and complete | finishes a process.

  In the providing process, as illustrated in FIG. 9, first, the receiving unit 43 determines whether a feature designation request is received from the service Y (step S <b> 201). Here, when the reception unit 43 receives the feature designation (step S201, Yes), the extraction unit 45 selects the user U estimated to have the feature designated by the feature designation based on the estimation result of the estimation unit 44. Extract (step S202).

  Then, the providing unit 46 provides the providing information including the list of the users U extracted by the extracting unit 45 to the service Y (Step S203), and ends the process.

  On the other hand, if the reception unit 43 has not received a feature specification request (No in step S201), the reception unit 43 determines whether a user specification request has been received (step S204). Note that the processing order of steps S201 and S204 may be switched.

  Here, when the reception unit 43 receives a user-specified request (step S204, Yes), the extraction unit 45 extracts the feature of the user U specified by the user specification based on the estimation result of the estimation unit 44 ( Step S205).

  Subsequently, the providing unit 46 provides providing information including information indicating the characteristics of the user U extracted by the extracting unit 45 to the service Y (step S206), and ends the process.

  On the other hand, when the reception unit 43 does not receive the user-specified request (No at Step S204), the process proceeds to Step S201.

  Although not shown in FIG. 9, when the reception unit 43 has not received a feature specification or user specification request (No in step S <b> 204), the reception unit 43 determines whether a calculation specification request has been received. it can.

  Here, when such a request is received, the estimation unit 44 performs a vector operation with the content of the specified arithmetic expression, and notifies the providing unit 46 of feature information closest to the calculation result. The providing unit 46 provides the provided information including the notified feature information to the service Y, and ends the process. On the other hand, when the calculation designation request is not accepted, the accepting unit 43 shifts the processing to step S201.

[5. Hardware configuration)
The information processing apparatus 1, the service providing server 10, the provider terminal 20, and the user terminal 100 according to the above-described embodiments are realized by a computer 1000 having a configuration as shown in FIG. Hereinafter, the information processing apparatus 1 will be described as an example. FIG. 10 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing apparatus 1 according to the embodiment. The computer 1000 includes a central processing unit (CPU) 1100, a random access memory (RAM) 1200, a read only memory (ROM) 1300, a hard disk drive (HDD) 1400, a communication interface (I / F) 1500, an input / output interface (I). / F) 1600 and a media interface (I / F) 1700.

  The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400 and controls each unit. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, a program depending on the hardware of the computer 1000, and the like.

  The HDD 1400 stores a program executed by the CPU 1100, data used by the program, and the like. The communication interface 1500 receives data from other devices via the communication network 500 (corresponding to the network N shown in FIG. 2) and sends the data to the CPU 1100, and the data generated by the CPU 1100 is transferred to other devices via the communication network 500. Send to device.

  The CPU 1100 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. Further, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the information processing apparatus 1 according to the embodiment, the CPU 1100 of the computer 1000 implements each function of the control unit 4 by executing a program loaded on the RAM 1200. The HDD 1400 stores data in the storage unit 3. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them, but as another example, these programs may be acquired from other devices via the communication network 500.

[6. Others]
In addition, among the processes described in the above embodiment, all or part of the processes described as being automatically performed can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified. For example, the various types of information illustrated in each drawing is not limited to the illustrated information.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  For example, the acquisition unit 41 and the reception unit 43 illustrated in FIG. 3 may be integrated. Further, for example, information stored in the storage unit 3 may be stored in a predetermined storage device provided outside via the network N.

  Moreover, in the said embodiment, the information processing apparatus 1 is the model which estimates the characteristic of the user U using the attribute of the user U, for example, and at least the estimation result regarding the characteristic of the user U estimated using the said attribute An acquisition process for acquiring feature information including one from each of different services Y, a conversion process for converting the feature information acquired by the acquisition process into a vector value, and a feature value based on the vector value converted by the conversion process The example which performs the estimation process which estimates a relationship was shown. However, in the information processing apparatus 1 described above, an acquisition apparatus that performs an acquisition process, a conversion apparatus that performs a conversion process, and an estimation apparatus that performs an estimation process may be separated. In this case, the acquisition device includes at least the acquisition unit 41. The conversion device has at least a conversion unit 42. The estimation device has at least an estimation unit 44. And the process by said information processing apparatus 1 is implement | achieved by information processing system S which has each apparatus of an acquisition apparatus, a conversion apparatus, and an estimation apparatus.

  In addition, the above-described embodiments and modifications can be combined as appropriate within a range that does not contradict processing contents.

[7. effect〕
The information processing apparatus 1 of the information processing system S according to the embodiment includes an acquisition unit 41, a conversion unit 42, and an estimation unit 44. The acquisition unit 41 obtains feature information including at least one of a model for estimating the feature of the user U using the attribute of the user U and an estimation result regarding the feature of the user U estimated using the attribute from each of the different services Y. Get from. The conversion unit 42 converts the feature information acquired by the acquisition unit 41 into a vector value. The estimation unit 44 estimates the relationship between the feature information based on the vector value converted by the conversion unit 42. Thereby, the relationship between the individual analysis results of different services can be estimated.

  Moreover, the estimation part 44 determines the similarity of feature information based on the said vector value. This makes it possible to determine the proximity of feature information.

  Further, when the estimation unit 44 determines that the similarity is close, if the user U has a feature corresponding to one of the feature information, the user U is also applicable to the other feature information. It presumes that it has the characteristic to do. Thereby, it is possible to perform estimation using the individual analysis results of different services in a cross-sectional manner.

  Moreover, the estimation part 44 estimates the feature information which the calculation result which calculated the feature information based on the vector value shows. Thereby, for example, feature information suitable for a desired prediction result can be extracted, but on the contrary, feature information that could not be predicted manually can be extracted. Therefore, it is possible to enhance and expand coverage in marketing and the like.

  The estimation unit 44 derives feature information having a vector value closest to the calculation result. As a result, among the feature information registered in the shared information 31, the feature information closest to the calculation result can be derived accurately and easily.

  Further, the feature information model outputs the feature of the user U as a score by inputting the attribute, and the conversion unit 42 is based on at least learning using the attribute and the score as a feature. Each piece of information is converted into a multi-dimensional vector value. Thereby, vectorization of the feature information based on the attribute and score of the user U is possible.

  The conversion unit 42 performs dimensional compression by extracting only significant dimensions from the multi-dimensions. Thereby, noise components and the like can be eliminated, and the feature information can be vectorized with high accuracy.

  The embodiment of the present application has been described in detail with reference to the drawings. However, this is an exemplification, and various modifications and improvements are made based on the knowledge of those skilled in the art including the aspects described in the column of the disclosure of the invention. The present invention can be implemented in other forms.

  In addition, the “section (module, unit)” described above can be read as “means” or “circuit”. For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

DESCRIPTION OF SYMBOLS 1 Information processing apparatus 2 Communication part 3 Memory | storage part 4 Control part 10 Service provision server 20 Provider terminal 31 Shared information 32 Service information 33 Model program information 34 Vector information 41 Acquisition part 42 Conversion part 43 Acceptance part 44 Estimation part 45 Extraction part 46 Providing unit 100 User terminal S Information processing system

Claims (9)

An acquisition unit for acquiring feature information including at least one of a model for estimating the feature of the user using the attribute of the user and an estimation result regarding the feature of the user estimated using the attribute from each of different services;
A conversion unit that converts the feature information acquired by the acquisition unit into a vector value;
An information processing apparatus comprising: an estimation unit that estimates a relationship between the feature information based on the vector value converted by the conversion unit. The estimation unit includes
The information processing apparatus according to claim 1, wherein the similarity between the feature information is determined based on the vector value. The estimation unit includes
When it is determined that the similarity is close, if the user has a feature corresponding to one of the feature information, the user is estimated to have a feature corresponding to the other feature information. The information processing apparatus according to claim 2, wherein: The estimation unit includes
The information processing apparatus according to claim 1, wherein the feature information indicated by a calculation result obtained by calculating the feature information based on the vector value is estimated. The estimation unit includes
The information processing apparatus according to claim 4, wherein the feature information having the vector value closest to the calculation result is derived. The model of the feature information is
By inputting the attribute, the feature of the user is output as a score,
The converter is
6. The information processing apparatus according to claim 1, wherein each of the feature information is converted into the multi-dimensional vector value based on at least learning using the attribute and the score as features. . The converter is
The information processing apparatus according to claim 6, wherein dimension compression is performed by extracting only significant dimensions among the multi-dimensions. An information processing method executed by a computer,
An acquisition step of acquiring feature information including at least one of a model for estimating the user's feature using the user's attribute and an estimation result regarding the user's feature estimated using the attribute from each of different services;
A conversion step of converting the feature information acquired by the acquisition step into a vector value;
An estimation step of estimating a relationship between the feature information based on the vector value converted by the conversion step. An acquisition procedure for acquiring feature information including at least one of a model for estimating the user's feature using the user's attribute and an estimation result regarding the user's feature estimated using the attribute from each of different services;
A conversion procedure for converting the feature information acquired by the acquisition procedure into a vector value;
An information processing program causing a computer to execute an estimation procedure for estimating a relationship between the feature information based on the vector value converted by the conversion procedure.

Images