JP2020129192A - Information processing device, information processing method, and information processing program - Google Patents

Abstract

To recommend appropriate information to a user.SOLUTION: An information processing device comprises an extraction unit and a determination unit. The extraction unit extracts feature information indicating features of a prescribed query, assuming that a plurality of queries entered by a same user in a prescribed time should have similar features, using a learning model having learned the features of the plurality of queries. The determination unit determines recommendation information to be recommended to the user who entered the prescribed query on the basis of the feature information extracted by the extraction unit.SELECTED DRAWING: Figure 1

Description

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

2. Description of the Related Art Conventionally, there is known a technique of recommending a product or service that is of interest to a user. For example, the input natural language request is semantically analyzed to generate branch information including the user's intention. Then, a technique for ranking the item candidates to be presented to the user based on the generated context information has been proposed.

JP, 2016-91535, A

However, in the above-mentioned related art, it is not always possible to recommend appropriate information to the user. For example, in the above-mentioned conventional technique, the input natural language request is only semantically analyzed to generate context information, and it is not always possible to recommend appropriate information to the user.

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 that can recommend appropriate information to a user.

The information processing apparatus according to the present application uses a learning model in which the features of the plurality of search queries are learned, assuming that the plurality of search queries input by the same user within a predetermined time have similar features, An extraction unit that extracts characteristic information indicating characteristics of a predetermined query, and a determination unit that determines recommendation information recommended to a user who has input the predetermined query, based on the characteristic information extracted by the extraction unit. It is characterized by having.

According to the aspect of the embodiment, it is possible to recommend appropriate information to the user.

FIG. 1 is a diagram illustrating an example of information processing according to the first embodiment. FIG. 2 is a diagram illustrating a configuration example of the information processing system according to the first embodiment. FIG. 3 is a diagram illustrating a configuration example of the information processing device according to the first embodiment. FIG. 4 is a diagram illustrating an example of the model information storage unit according to the first embodiment. FIG. 5 is a diagram illustrating an example of the vector information storage unit according to the first embodiment. FIG. 6 is a diagram illustrating an example of the search information storage unit according to the first embodiment. FIG. 7 is a flowchart showing a generation processing procedure according to the first embodiment. FIG. 8 is a flowchart showing an information processing procedure according to the first embodiment. FIG. 9 is a diagram illustrating an example of information processing according to the modification. FIG. 10 is a diagram illustrating an example of information processing according to the modification. FIG. 11 is a diagram illustrating an example of a process in which the user terminal according to the modification switches the content. FIG. 12 is a diagram illustrating a configuration example of the information processing device according to the second embodiment. FIG. 13 is a diagram illustrating an example of the category information storage unit according to the second embodiment. FIG. 14 is a flowchart showing a prediction processing procedure according to the second embodiment. FIG. 15 is a flowchart showing an information processing procedure according to the second embodiment. FIG. 16 is a diagram illustrating an example of the first learning model generation process according to the embodiment. FIG. 17 is a diagram illustrating an example of the first learning model generation process according to the embodiment. FIG. 18 is a diagram illustrating an example of a second learning model generation process according to the embodiment. FIG. 19 is a diagram illustrating a configuration example of the generation device according to the embodiment. FIG. 20 is a diagram illustrating an example of the query information storage unit according to the embodiment. FIG. 21 is a diagram illustrating an example of the vector information storage unit according to the embodiment. FIG. 22 is a diagram illustrating an example of the classification definition storage unit according to the embodiment. FIG. 23 is a diagram illustrating an example of the category information storage unit according to the embodiment. FIG. 24 is a diagram illustrating an example of the model information storage unit according to the embodiment. FIG. 25 is a diagram illustrating an example of the first learning model according to the embodiment. FIG. 26 is a diagram illustrating an example of the second learning model according to the embodiment. FIG. 27 is a flowchart showing a procedure for generating the first learning model according to the embodiment. FIG. 28 is a flowchart showing a procedure for generating the second learning model according to the embodiment. FIG. 29 is a diagram illustrating an example of the hardware configuration of a computer that executes a program.

Hereinafter, modes (hereinafter, referred to as “embodiments”) for implementing an information processing apparatus, an information processing method, and an information processing program according to the present application will be described in detail with reference to the drawings. The information processing apparatus, the information processing method, and the information processing program according to the present application are not limited to this embodiment. Further, in each of the following embodiments, the same parts are designated by the same reference numerals, and overlapping description will be omitted.

[1. First Embodiment]
[1-1. Example of information processing]
First, an example of information processing according to the first embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of information processing according to the first embodiment. The information processing shown in FIG. 1 is performed by the user terminal 10, the search server 20 (see FIG. 2), the generation device 50 (see FIG. 2), and the information processing device 100.

The user terminal 10 is an information processing device used by a user. The user terminal 10 is realized by, for example, a smartphone, a tablet terminal, a notebook PC (Personal Computer), a mobile phone, a PDA (Personal Digital Assistant), or the like. In the following, the user terminal 10 may be identified with the user. That is, in the following, the user can be read as the user terminal 10.

In the following, the user identified by the user ID “U11” may be referred to as “user U11”. As described above, in the following, when "user U* (* is an arbitrary numerical value)" is described, that user is a user specified by the user ID "U*". For example, when described as “user U21”, the user is the user specified by the user ID “U21”.

Moreover, below, depending on the user who uses the user terminal 10, the user terminal 10 may be described as the user terminals 10-1 and 10-2. For example, the user terminal 10-1 is the user terminal 10 used by the user U11. Further, for example, the user terminal 10-2 is the user terminal 10 used by the user U21. Further, in the following, the user terminals 10-1 and 10-2 will be referred to as the user terminal 10 if they are described without any particular distinction.

The search server 20 (see FIG. 2) is a server device that provides a search service. For example, the search service provided by the search server 20 is a comprehensive search service capable of searching all kinds of information. The search server 20 stores information about the search query input by the user. Specifically, the search server 20 stores information regarding the search history of the user.

The generation device 50 (see FIG. 2) is a server device that generates the first learning model. Here, the outline of the generation processing of the first learning model by the generation device 50 will be described. The details of the generation process of the first learning model by the generation device 50 will be described later. Specifically, the generation device 50 acquires information regarding the search query input by the user from the search server 20. Then, the generation device 50 extracts a plurality of search queries input from the search server 20 by the same user within a predetermined time. Here, the generation device 50 treats the entire character string input in the search box for each search by the user as one search query input by the user. For example, when the search query including a plurality of character strings such as “Roppongi pasta” is input to the search box in one search by the user U1, the generation device 50 makes one search query for the entire “Roppongi pasta”. Treat as. In addition, the generating device 50 may generate a plurality of search queries within a predetermined time by the same user such that the time intervals at which the respective search queries are input by the same user are within a predetermined time (for example, within 2 minutes). Extract as multiple search queries entered in.

Subsequently, the generating device 50 generates a first learning model that predicts the characteristic information of the predetermined search query from the predetermined search query by learning that the extracted plurality of search queries have similar characteristics. Specifically, the generation device 50 trains the first learning model so that the extracted distributed expressions of the plurality of search queries are similar to each other, so that the distributed expression including the characteristic information of the predetermined search query from the predetermined search query. A first learning model that outputs (vector) is generated. More specifically, the generation device 50 uses a technique of DSSM (Deep Structured Semantic Model) that uses an LSTM (Long Short-Term Memory), which is a type of RNN (Recurrent Neural Network), for distributed expression generation, and performs a search. A first learning model that outputs a distributed expression (vector) is generated from the query. For example, the generation device 50 assumes that, as the correct answer data of the first learning model, a pair of search queries input by the same user within a predetermined time have similar characteristics, and the distributed expression (vector) of the predetermined search query is used. ) And a distributed expression (vector) of another search query that forms a pair with a predetermined search query are learned so that they exist near each other in the distributed expression space. Note that learning so that two vectors exist close to each other in the distributed expression space can be rephrased as learning so that the two vectors are similar in the distributed expression space.

The information processing device 100 is a server device that provides a real estate information search service (hereinafter, appropriately referred to as “real estate information search service R1”). The information processing device 100 acquires the model data of the first learning model from the generation device 50. In the following, the model data of the first learning model may be simply referred to as the first learning model. The information processing apparatus 100 uses the first learning model to recommend another real estate area having characteristics similar to the real estate area corresponding to the place name received from the user as the recommended area.

From here, the flow of information processing is demonstrated using FIG. In FIG. 1, the information processing apparatus 100 uses a first learning model to generate a distributed expression (vector) corresponding to a character string indicating a place name or a station name (hereinafter, referred to as “place name query” as appropriate) throughout the country. Generate (step S1). In the balloon indicated by a dotted line on the right side of FIG. 1, a state in which the distributed expression (vector) corresponding to the place name query generated by the information processing apparatus 100 is mapped in the distributed expression space is shown. For example, the point indicating the distributed expression (vector) corresponding to the place name query “place name #11” and the point indicating the distributed expression (vector) corresponding to the place name queries “place name #12” to “place name #14” are the distributed expressions. It is shown to be located relatively close in space. That is, this figure means that the place name query “place name #11” and the place name queries “place name #12” to “place name #14” have similar features. On the other hand, the point indicating the distributed expression (vector) corresponding to the place name query “place name #11” and the point indicating the distributed expression (vector) corresponding to the place name queries “place name #21” to “place name #22” are the distributed expressions. It is shown to exist relatively far in space. That is, this figure means that the place name query “place name #11” and the place name queries “place name #21” to “place name #22” have different characteristics. In FIG. 1, for the sake of explanation, the place name query is expressed by an abstract symbol such as “place name #11”. However, when implementing the present invention, the place name query includes “Musashi Kosugi” and “Kichijoji”. A specific station name such as "" or a specific place name such as "Minato Ward" or "Tokyo" is used.

In addition, the information processing apparatus 100 accepts the place name query “place name #11” from the user U11 via the content C11 that includes the search box for inputting the search query for searching the city of interest (step S2). Subsequently, when the information processing apparatus 100 receives the place name query “place name #11”, the distributed expression (vector) corresponding to the place name query “place name #11” that has been generated in advance and the place name query “place name #11” are excluded. The degree of similarity with the distributed expression (vector) corresponding to another place name query is calculated (step S3). The information processing apparatus 100 may calculate the similarity between the place name queries in advance, instead of calculating the similarity each time the place name query is received. Then, the information processing apparatus 100, after calculating the similarity, determines whether the calculated similarity exceeds a predetermined threshold value. Subsequently, when the information processing apparatus 100 determines that the degree of similarity exceeds a predetermined threshold value, the information processing apparatus 100 extracts another place name query as a similar query having characteristics similar to the place name query “place name #11” (step S4). ).

For example, the information processing apparatus 100 calculates the similarity between the distributed expression (vector) corresponding to the place name query “place name #11” and the distributed expression (vector) corresponding to the place name query “place name #12” as 0.9. .. Subsequently, in the information processing apparatus 100, the similarity between the distributed expression (vector) corresponding to the place name query “place name #11” and the distributed expression (vector) corresponding to the place name query “place name #12” is a predetermined threshold value (eg, , 0.8) is exceeded. For example, the information processing apparatus 100 extracts the place name query “place name #12” as the similar query of the place name query “place name #11” because the similarity 0.9 exceeds the predetermined threshold of 0.8. .. Similarly, the information processing apparatus 100 calculates the degree of similarity with the distributed expression (vector) corresponding to the place name query “place name #11” for all other place name queries other than the place name query “place name #11”. Then, the information processing apparatus 100 determines whether or not the degree of similarity exceeds a predetermined threshold for all place name queries other than the place name query “place name #11”.

Subsequently, when the information processing apparatus 100 extracts the similar query, the information processing apparatus 100 determines a recommended area recommended for the user U11 based on the similar query (step S5). Specifically, the information processing apparatus 100 determines to recommend the real estate area corresponding to the place name query extracted as the similar query as the recommended area. For example, the information processing apparatus 100 determines to recommend the real estate area corresponding to the place name query “place name #12” extracted as a similar query to the place name query “place name #11” as a recommended area recommended to the user U11. To do. Note that the information processing apparatus 100 may determine to recommend, as a recommended area, a real estate area corresponding to a place name query having a similarity within a higher predetermined number than the real estate areas corresponding to the similar query. ..

Subsequently, when the information processing apparatus 100 determines the recommended area, the information processing apparatus 100 transmits information on the determined recommended area (for example, information on real estate properties in the recommended area) to the user U11 (step S6).

As described above, the information processing apparatus 100 according to the first embodiment has a feature that a plurality of search queries have the similar feature that a plurality of search queries input by the same user within a predetermined time period have similar features. Feature information indicating the features of a predetermined query is extracted using the learning model that has learned. In addition, the information processing apparatus 100 determines recommendation information that is recommended to the user who has input the predetermined query, based on the extracted characteristic information. With this, the information processing apparatus 100 can recommend information based on the characteristic information indicating the characteristics of the predetermined search query to the user who is interested in the predetermined search query. That is, the information processing apparatus 100 can recommend the interest of the user or information that matches the interest. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

In addition, in general, a user who is interested in a specific field such as a user who visits a search service, but has little knowledge about the field, who has little knowledge of the field, is likely to find an appropriate search. The problem is that I can't think of a query. The information processing apparatus 100 according to the present invention can recommend recommended information based on an appropriate search query according to a search intention, based on a search query input by a user who has little knowledge of the input search query. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

Further, conventionally, a technique related to concept retrieval is known. For example, a technique for constructing a concept from a log of user actions is known. For example, focusing on the correlation between the action of turning on the cooler and the temperature, there is known a technique of constructing a concept of hot (high temperature) (for example, a search query) from a log of the action of the user turning on the cooler. There is. However, in the past, humans have registered the rule that the act of putting on a cooler is related to the concept of being hot (high temperature) (rule base). In addition, if machine learning is used without depending on the rule base, it is necessary to manually create learning data by labeling a large number of queries. Therefore, conventionally, only a concept search of a narrowly defined query can be performed. That is, conventionally, only a concept search of a query (a narrowly defined query) to which a human being has previously given an answer concept can be performed. Therefore, the information processing apparatus 100 according to the present invention models a query string in a user's search session without manually labeling a large number of queries. As a result, the information processing apparatus 100 according to the present invention can learn a niche query (long tail query) that is searched by only a small number of users by associating the query with the search intention. That is, the information processing apparatus 100 according to the present invention can perform a concept search of a query in a broad sense that also covers a niche query that the user freely inputs. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

[1-2. Information processing system configuration]
Next, the configuration of the information processing system according to the first embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating a configuration example of the information processing system according to the first embodiment. As shown in FIG. 2, the information processing system 1 includes a user terminal 10, a search server 20, a generation device 50, and an information processing device 100. The user terminal 10, the search server 20, the generation device 50, and the information processing device 100 are connected via a predetermined network N so as to be communicable by wire or wirelessly. Note that the information processing system 1 illustrated in FIG. 2 includes an arbitrary number of user terminals 10, an arbitrary number of search servers 20, an arbitrary number of generation devices 50, and an arbitrary number of information processing devices 100. Good.

The user terminal 10 transmits the search query input by the user to the search server 20. Specifically, the user terminal 10 acquires a search page including a search box for inputting a search query from the search server 20 according to the operation by the user. Subsequently, when the user terminal 10 performs an operation of transmitting a search query following an operation of inputting characters in the search box by the user, the character input in the search box via the search page is used as the search query. It is transmitted to the search server 20. For example, when the user presses the send button of the search query or presses the enter key after the user inputs characters in the search box, the user terminal 10 displays the search page through the search page. The characters input in the search box are transmitted to the search server 20 as a search query.

When the search server 20 receives the search query from the user terminal 10, the search server 20 selects the content corresponding to the received search query and output as the search result. Then, the search server 20 delivers the search result page including the selected content to the user terminal 10. Here, the content distributed by the search server 20 is not limited to the web page displayed by the web browser. For example, the content distributed by the search server 20 may be content displayed by a dedicated application installed in the user terminal 10. The contents distributed by the search server 20 are music contents, images (including still images as well as moving images) contents, and text contents (including news articles and articles posted on SNS (Social Networking Service)). , Any content including a combination of images and text, game content, and the like.

In addition, when the search server 20 receives a search query from the user terminal 10, the search server 20 registers the received search query, the user ID that identifies the user who is the transmission source of the search query, and the transmission date and time of the search query in association with each other. .. The search server 20 transmits the information about the search query input by the user to the generation device 50 in response to the request from the generation device 50.

The generation device 50 generates the first learning model by executing the processing described below. Details of the generation processing of the first learning model by the generation device 50 will be described later.

In addition, the user terminal 10 transmits the search query input by the user to the information processing device 100. Specifically, the user terminal 10 acquires, from the information processing apparatus 100, the content C11 including a search box for inputting a search query for searching a city that the user wants to know according to an operation performed by the user. Subsequently, when the user terminal 10 performs an operation of transmitting a search query after the operation of inputting characters in the search box by the user, the characters such as the place name and the station name input in the search box via the content C11. Is transmitted to the information processing device 100 as a search query. For example, when the operation of pressing the send button of the search query or the operation of pressing the enter key is performed subsequent to the operation of the user inputting a character in the search box, the user terminal 10 transmits the content C11 via the content C11. The characters input in the search box are transmitted to the information processing device 100 as a search query.

The information processing device 100 is a server device that performs the information processing described in FIG. The information processing apparatus 100 determines that a plurality of search queries input by the same user within a predetermined time period have similar features, and use a first learning model that learns the features of the plurality of search queries to determine a predetermined value. As the feature information indicating the feature of the query, a similar query that is a search query having a feature similar to the predetermined query is extracted. In addition, the information processing apparatus 100 determines recommendation information that is recommended to the user who has input a predetermined query, based on the similar query that is the extracted feature information.

[1-3. Configuration of information processing device]
Next, the configuration of the information processing device 100 according to the first embodiment will be described with reference to FIG. FIG. 3 is a diagram illustrating a configuration example of the information processing device 100 according to the first embodiment. As illustrated in FIG. 3, the information processing device 100 includes a communication unit 110, a storage unit 120, and a control unit 130. The information processing apparatus 100 includes an input unit (for example, a keyboard and a mouse) that receives various operations from an administrator of the information processing apparatus 100, and a display unit (for example, a liquid crystal display) for displaying various information. You may have.

(Communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like. Then, the communication unit 110 is connected to the network by wire or wirelessly, and transmits and receives information between the user terminal 10, the search server 20, and the generation device 50, for example.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory (Flash Memory), or a storage device such as a hard disk or an optical disk. As shown in FIG. 3, the storage unit 120 has a model information storage unit 121, a vector information storage unit 122, a search information storage unit 123, and a content storage unit 124.

(Model information storage unit 121)
The model information storage unit 121 stores various information regarding the learning model generated by the generation device 50. FIG. 4 shows an example of the model information storage unit according to the first embodiment. In the example shown in FIG. 4, the model information storage unit 121 has items such as “model ID” and “model data”.

The “model ID” indicates identification information for identifying the learning model generated by the generation device 50. The “model data” indicates model data of the learning model generated by the generation device 50. For example, “model data” stores data for converting a search query into a distributed expression.

In the example shown in the first record of FIG. 4, the learning model identified by the model ID “M1” corresponds to the first learning model M1 shown in FIG. The model data “MDT1” indicates model data (model data MDT1) of the first learning model M1 generated by the generation device 50.

The model data MDT1 includes an input layer to which a search query is input, an output layer, a first element belonging to any layer from the input layer to the output layer and other than the output layer, the first element, and the first element. A second element whose value is calculated based on the weight of one element, and a distributed expression of the search query input to the input layer is output from the output layer according to the search query input to the input layer. Thus, the generation device 50 may be operated.

Here, it is assumed that the model data MDT1 is realized by the regression model represented by “y=a1*x1+a2*x2+...+ai*xi”. In this case, the first element included in the model data MDT1 corresponds to the input data (xi) such as x1 and x2. The weight of the first element corresponds to the coefficient ai corresponding to xi. Here, the regression model can be regarded as a simple perceptron having an input layer and an output layer. When each model is regarded as a simple perceptron, the first element can be regarded as a node included in the input layer, and the second element can be regarded as a node included in the output layer.

Further, it is assumed that the model data MDT1 is realized by a neural network having one or a plurality of intermediate layers such as DNN (Deep Neural Network). In this case, the first element included in the model data MDT1 corresponds to any node included in the input layer or the intermediate layer. The second element corresponds to the node at the next stage, which is the node to which the value is transmitted from the node corresponding to the first element. The weight of the first element corresponds to the connection coefficient, which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

The generation device 50 calculates the distributed expression using a model having an arbitrary structure such as the above-described regression model or neural network. Specifically, the model data MDT1 is set with a coefficient so as to output a distributed expression when a search query is input. The generation device 50 calculates a distributed expression using such model data MDT1.

In the above example, the model data MDT1 is the model (hereinafter referred to as model X1) that outputs the distributed expression of the search query when the search query is input. However, the model data MDT1 according to the embodiment may be a model generated based on a result obtained by repeating input/output of data to/from the model X1. For example, the model data MDT1 may be a model (hereinafter, referred to as model Y1) learned by inputting the distributed expression output by the model X1 when the search query is input. Alternatively, the model data MDT1 may be a model learned so that the search query is input and the output value of the model Y1 is output.

When the generation device 50 performs the estimation process using GAN (Generative Adversarial Networks), the model data MDT1 may be a model forming a part of GAN.

(Vector information storage unit 122)
The vector information storage unit 122 stores various kinds of information regarding vectors, which are distributed expressions of search queries. FIG. 5 shows an example of the vector information storage unit according to the first embodiment. In the example shown in FIG. 5, the vector information storage unit 122 has items such as “search query” and “vector information”.

"Search query" indicates a search query input by the user. “Vector information” indicates an N-dimensional vector that is a distributed expression of a search query. The vector that is the distributed expression of the search query is, for example, a 128-dimensional vector.

In the example shown in the first record of FIG. 5, the search query “place name #11” corresponds to the place name query “place name #11” shown in FIG. The vector information “V11” indicates a distributed expression (vector) corresponding to the place name query “place name #11” shown in FIG.

(Search information storage unit 123)
The search information storage unit 123 stores various kinds of information regarding a user's search history in the real estate information search service R1 provided by the information processing apparatus 100. FIG. 6 shows an example of the search information storage unit according to the first embodiment. In the example shown in FIG. 6, the search information storage unit 123 has items such as “user ID”, “date and time”, and “search query”.

The “user ID” indicates identification information for identifying the user who has input the search query. The “date and time” indicates the date and time when the information processing apparatus 100 accepted the search query from the user. "Search query" indicates a search query input by the user.

In the example shown in the first record of FIG. 6, the search query “place name #11” corresponds to the place name query “place name #11” shown in FIG. Further, the user ID “U11” indicates that the user who inputs the place name query “place name #11” is the user (user U11) identified by the user ID “U11”. The date and time “2019/1/1 PM 17:00” means that the date and time when the information processing apparatus 100 receives the place name query “place name #11” from the user U11 is 17:00 pm on January 1, 2019. Show.

(Content storage unit 124)
The content storage unit 124 stores various kinds of information regarding content. Specifically, the content storage unit 124 stores content related to the real estate information search service R1 provided by the information processing apparatus 100. For example, the content storage unit 124 stores the content C11 including a search box for inputting a search query shown in FIG.

(Control unit 130)
Returning to the description of FIG. 3, the control unit 130 is a controller, and is stored in a storage device inside the information processing device 100 by, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). It is realized by executing various programs (corresponding to an example of an information processing program) stored in the RAM as a work area. The control unit 130 is a controller, 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 130 includes an acquisition unit 131, a generation unit 132, a provision unit 133, a calculation unit 134, an extraction unit 135, and a determination unit 136, and information described below. Realize or execute the action of processing. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in FIG.

(Acquisition unit 131)
The acquisition unit 131 acquires the first learning model. More specifically, the acquisition unit 131 acquires the first learning model generated by the generation device 50 from the generation device 50. When acquiring the first learning model, the acquisition unit 131 stores the acquired first learning model in the model information storage unit 121.

(Generator 132)
The generation unit 132 determines that a plurality of search queries input by the same user within a predetermined time have similar features, and uses a learning model that learns the features of the plurality of search queries to perform a predetermined search query. Generate a distributed representation of. Specifically, the generation unit 132 inputs the search query into the first learning model acquired by the acquisition unit 131 and generates a distributed expression (vector) corresponding to the search query. For example, the generation unit 132 acquires list data of place names and station names from all over the country from an open database or a dictionary. In this way, the generation unit 132 acquires a place name query that is a character string indicating a place name or a station name nationwide. Subsequently, the generation unit 132 inputs the place name query into the first learning model acquired by the acquisition unit 131, and generates a distributed expression (vector) corresponding to the place name query. After generating the distributed expression (vector), the generation unit 132 stores the vector information of the generated distributed expression in the vector information storage unit 122 in association with the search query.

(Providing section 133)
The providing unit 133 provides the real estate information search service R1. Specifically, the providing unit 133 distributes the content related to the real estate information search service R1 to the user terminal 10. For example, the providing unit 133 distributes the content C11 including a search box for inputting a search query for searching a city that the user wants to know. The providing unit 133 also receives a search query from the user via the content C11. For example, the providing unit 133 receives a place name query from the user. When the providing unit 133 receives a search query from a user, the providing unit 133 stores the received search query in the search information storage unit 123 in association with the date and time when the query was received and the user ID of the user who is the transmission source of the query.

The providing unit 133 provides a service that recommends another real estate area having characteristics similar to the real estate area corresponding to the place name received from the user as a recommended area. Specifically, the providing unit 133 transmits information on the recommended area determined by the determining unit 136 to the user terminal 10.

(Calculation unit 134)
The calculation unit 134 calculates the similarity between the distributed expression of the predetermined search query generated by the generation unit 132 and the distributed expression of another search query different from the predetermined search query generated by the generation unit 132. Specifically, when the providing unit 133 receives the predetermined search query, the calculating unit 134 receives the distributed expression (vector) corresponding to the predetermined search query generated in advance by the generating unit 132 and the received predetermined search query. Other than, the similarity with the distributed expression (vector) corresponding to the search query is calculated. For example, the calculation unit 134 calculates the cosine similarity between the distributed expressions (vectors). Note that the calculation unit 134 may calculate the similarity between distributed expressions (vectors) based on any index as long as it is an index applicable as a distance measure between vectors, not limited to the cosine similarity. Good. For example, the calculation unit 134 may calculate a value of a predetermined distance function such as a Euclidean distance between distributed expressions (vectors), a distance in a non-Euclidean space such as a hyperbolic space, a Manhattan distance, a Mahalanobis distance, or the like. The calculating unit 134 may calculate the similarity between the distributed expressions (vectors) corresponding to the search query in advance, instead of calculating the similarity each time the providing unit 133 accepts a predetermined search query. Good. For example, the calculation unit 134 calculates the degree of similarity between the distributed expressions (vectors) corresponding to the place name query in advance.

(Extractor 135)
The extraction unit 135 determines that a plurality of search queries input by the same user within a predetermined time have similar features by using a learning model that has learned the features of the plurality of search queries. Feature information indicating features is extracted. Specifically, the extraction unit 135 extracts, as the characteristic information, a similar query that is a search query having characteristics similar to a predetermined query. More specifically, when the calculating unit 134 calculates the similarity between the distributed expressions (vectors), the extracting unit 135 determines whether the calculated similarity exceeds a predetermined threshold value. For example, it is determined whether the cosine similarity between the distributed expressions (vectors) calculated by the calculation unit 134 exceeds a predetermined threshold. Subsequently, when the extraction unit 135 determines that the similarity calculated by the calculation unit 134 exceeds a predetermined threshold, the extraction unit 135 extracts another place name query as a similar query having similar characteristics to the predetermined search query. .. The calculating unit 134 may determine whether or not the value of the predetermined distance function between the distributed expressions (vectors) (that is, the distance in the distributed expression space) is below a predetermined threshold. Subsequently, when the extraction unit 135 determines that the similarity calculated by the calculation unit 134 is lower than a predetermined threshold value, the extraction unit 135 extracts another place name query as a similar query having similar characteristics to the predetermined search query. ..

In addition, the extraction unit 135 extracts a similar query having the same attribute as the predetermined query. For example, the extraction unit 135 extracts a similar query indicating a real estate area, which is a search query having similar characteristics to the predetermined query indicating the real estate area, as the similar query having the same attribute as the predetermined query.

The extraction unit 135 also extracts the characteristic information by using a learning model that outputs a distributed expression of the predetermined search query as the output information when the predetermined search query is input as the input information. For example, the extraction unit 135 extracts the similar query as the characteristic information by using the first learning model that outputs the distributed expression of the predetermined search query as the output information when the predetermined search query is input as the input information. ..

In addition, the extraction unit 135 uses a learning model in which the characteristics of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are continuously input within a predetermined time are similar to each other. Extract feature information. For example, the extraction unit 135 uses the first learning model in which the features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries continuously input within a predetermined time are similar to each other. Then, a similar query is extracted as the characteristic information.

The extraction unit 135 also learns that a plurality of search queries that are input by the same user within a predetermined time have similar characteristics to each other that include a character string delimited by a predetermined delimiter. By doing so, the feature information is extracted using the learning model in which the features of the plurality of search queries are learned. For example, the extraction unit 135 learns that a plurality of search queries input by the same user within a predetermined time period have a similar feature to a plurality of search queries including a character string delimited by a predetermined delimiter. By doing so, the similar query is extracted as the feature information using the first learning model in which the features of the plurality of search queries are learned.

Further, the extraction unit 135 extracts the characteristic information by using the learning model in which the characteristics of the plurality of search queries are learned by learning the plurality of randomly extracted search queries as having different characteristics. .. For example, the extraction unit 135 uses the first learning model in which the features of the plurality of search queries are learned by learning the plurality of randomly extracted search queries as having different features, and uses the first learning model as the feature information. Extract similar queries.

In addition, the extraction unit 135 extracts feature information using a learning model in which the features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are randomly extracted differ. .. For example, the extraction unit 135 uses the first learning model in which the characteristics of the plurality of search queries are learned by learning so that the distributed expressions of the pair of search queries that are randomly extracted are different from each other. Extract similar queries.

(Determination unit 136)
The determination unit 136 determines recommendation information to be recommended to the user who has input the predetermined query, based on the characteristic information extracted by the extraction unit 135. Specifically, the determination unit 136 determines recommendation information to be recommended to the user who has input the predetermined query, based on the similar query extracted by the extraction unit 135. More specifically, the determining unit 136 determines the information regarding the real estate area, which is the recommendation information, based on the similar query extracted by the extracting unit 135. For example, the determination unit 136 determines to recommend the real estate area indicated by the similar query extracted by the extraction unit 135 as the recommended area.

[1-4. Flow of generation processing]
Next, the procedure of the generation process according to the first embodiment will be described with reference to FIG. 7. FIG. 7 is a flowchart showing a generation processing procedure according to the first embodiment. In the example illustrated in FIG. 7, the information processing device 100 acquires the search query and the first learning model (step S101). Subsequently, when the information processing apparatus 100 acquires the search query and the first learning model, the information processing apparatus 100 uses the first learning model to generate a distributed expression (vector) of the search query (step S102).

[1-5. Information processing flow]
Next, the procedure of information processing according to the first embodiment will be described with reference to FIG. FIG. 8 is a flowchart showing an information processing procedure according to the first embodiment. In the example illustrated in FIG. 8, the information processing device 100 determines whether a search query has been accepted (step S201). If the search query has not been received (step S201; No), the information processing apparatus 100 waits until the search query is received.

On the other hand, when the information processing apparatus 100 receives the search query (step S201; Yes), the information processing apparatus 100 calculates the similarity between the distributed expressions (vectors) corresponding to the search query (step S202). Specifically, the information processing apparatus 100 calculates, for each vector, the degree of similarity between the vector corresponding to the received search query and the vector corresponding to another search query.

Then, the information processing apparatus 100, after calculating the similarity between the vectors, determines whether the calculated similarity exceeds a predetermined threshold value (step S203). If the calculated similarity does not exceed the predetermined threshold (step S203; No), the information processing apparatus 100 ends the process.

On the other hand, when the calculated similarity exceeds a predetermined threshold (step S203; Yes), the information processing apparatus 100 extracts a similar query having similar characteristics to the received search query (step S204). Specifically, when the similarity between the vector corresponding to the received search query and the vector corresponding to another search query exceeds a predetermined threshold, the information processing apparatus 100 extracts another search query as a similar query. To do. Subsequently, when the information processing apparatus 100 extracts the similar query, the information processing apparatus 100 determines a recommended area based on the extracted similar query (step S205).

[1-6. Modified example)
The information processing system 1 according to the above-described first embodiment may be implemented in various different forms other than the above-described embodiment. Therefore, other embodiments of the information processing system 1 will be described below. The same parts as those in the embodiment are designated by the same reference numerals and the description thereof will be omitted.

[1-6-1. Recommendation of real estate area based on conceptual query]
Next, information processing according to the modification will be described with reference to FIG. FIG. 9 is a diagram illustrating an example of information processing according to the modification. In FIG. 9, the generation unit 132 uses the first learning model to generate a distributed expression (vector) corresponding to a conceptual keyword (hereinafter, appropriately referred to as “concept query”) assumed in real estate search. (Step S1-A). Here, examples of conceptual keywords that can be assumed in real estate search include "safety", "school city", "sunny", "beautiful fireworks", "wide layout", etc. ..

In the balloon indicated by the dotted line on the right side of FIG. 9, in addition to the distributed expression (vector) corresponding to the place name query generated by the generation unit 132 in FIG. 1, the concept query generated by the generation unit 132 in FIG. It is shown that the corresponding distributed representation (vector) is mapped in the distributed representation space. For example, the point indicating the distributed expression (vector) corresponding to the concept query “safety” and the point indicating the distributed expression (vector) corresponding to the place name query “place name #21” are relatively in the distributed expression space. It is shown to be located nearby. That is, this figure means that the concept query “safety is good” and the place name query “place name #21” have similar features. On the other hand, the point indicating the distributed expression (vector) corresponding to the concept query “school city” and the point indicating the distributed expression (vector) corresponding to the place name query “place name #21” are relatively far in the distributed expression space. It is shown to exist in. That is, this figure means that the concept query “safety is good” and the place name query “place name #21” have different characteristics.

Further, the providing unit 133 receives the concept query “safety is good” from the user U21 via the content C21 including the search box for inputting the search query in the free word format (step S2-A). Subsequently, when the providing unit 133 receives the concept query “safety is good”, the calculating unit 134 corresponds to the distributed expression (vector) and the place name query corresponding to the concept query “safety is good” generated in advance. The degree of similarity with the distributed expression (vector) is calculated (step S3-A). Subsequently, when the extraction unit 135 calculates the similarity, it determines whether the calculated similarity exceeds a predetermined threshold value. Subsequently, when the extraction unit 135 determines that the degree of similarity exceeds a predetermined threshold, the extraction unit 135 extracts the place name query as a similar query having characteristics similar to the conceptual query “safety” (step S4-A). ).

For example, the calculation unit 134 calculates the similarity between the distributed expression (vector) corresponding to the concept query “safety is good” and the distributed expression (vector) corresponding to the place name query “place name #21” as 0.9. Then, the extraction unit 135 determines that the similarity between the distributed expression (vector) corresponding to the concept query “safety is good” and the distributed expression (vector) corresponding to the place name query “place name #21” is a predetermined threshold value (for example, 0.8) is determined. For example, the extraction unit 135 extracts the place name query “place name #21” as a similar query to the concept query “safety” because the similarity 0.9 exceeds a predetermined threshold of 0.8. Similarly, the calculation unit 134 calculates the degree of similarity with the distributed expression (vector) corresponding to the conceptual query “safety is good” for all the place name queries. Then, the extraction unit 135 determines whether or not the degree of similarity exceeds a predetermined threshold for all the place name queries.

Subsequently, when the extraction unit 135 extracts a similar query, the determination unit 136 determines a recommended area recommended for the user U21 based on the similar query extracted by the extraction unit 135 (step S5- A). Specifically, the determination unit 136 determines to recommend the real estate area corresponding to the place name query extracted as the similar query as the recommended area. For example, the determination unit 136 determines to recommend the real estate area corresponding to the place name query “place name #21” extracted as a similar query to the concept query “safety” as a recommended area recommended to the user U21. ..

Subsequently, when the determining unit 136 determines the recommended area, the providing unit 133 provides the user U21 with information on the recommended area determined by the determining unit 136 (for example, information on real estate properties in the recommended area). And transmits (step S6-A).

[1-6-2. Recommendation of refinement condition based on conceptual query]
Next, information processing according to the modification will be described with reference to FIG. FIG. 10 is a diagram illustrating an example of information processing according to the modification. In FIG. 10, the determination unit 136 determines a query candidate for re-search, which is recommendation information, based on the similar query extracted by the extraction unit 135. Specifically, the generation unit 132 uses the first learning model, and is a keyword indicating a narrowing-down condition used when the user narrows down properties in the real estate search (hereinafter, appropriately referred to as “re-search query”). A distributed expression (vector) corresponding to is generated (step S1-B). Here, as an example of the keyword indicating the narrowing condition used when the user narrows down the property in the real estate search, "high-rise condominium" or "low-rise condominium", which is a keyword indicating the characteristic of the property, or a keyword indicating the location condition of the property "Riverside", "within 5 minutes walk from the station" and so on.

In the balloon indicated by the dotted line on the right side of FIG. 10, a distributed expression (vector) corresponding to the place name query generated by the generation unit 132 in FIG. 1 and a distribution corresponding to the conceptual query generated by the generation unit 132 in FIG. In addition to the expressions (vectors), FIG. 10 illustrates that the distributed expressions (vectors) corresponding to the re-search query generated by the generation unit 132 are mapped in the distributed expression space. For example, the point indicating the distributed expression (vector) corresponding to the concept query “fireworks is beautiful” and the point indicating the distributed expression (vector) corresponding to the re-search query “high-rise apartment” are relative in the distributed expression space. It is shown to be located close to. That is, this figure means that the concept query “fireworks are beautiful” and the re-search query “high-rise apartment” have similar features. On the other hand, the point indicating the distributed expression (vector) corresponding to the concept query “fireworks is beautiful” and the point indicating the distributed expression (vector) corresponding to the re-search query “low-rise apartment” are relative in the distributed expression space. It is shown to exist in the distance. That is, this figure means that the concept query “fireworks are beautiful” and the re-search query “low-rise apartment” have different characteristics.

Further, the providing unit 133 receives the conceptual query “fireworks are beautiful” from the user U31 via the content C21 including the search box for inputting the search query in the free word format (step S2-B). Subsequently, when the providing unit 133 receives the conceptual query “fireworks are beautiful”, the calculating unit 134 causes the distributed expression (vector) corresponding to the previously generated conceptual query “fireworks is beautiful” and a re-search query. The degree of similarity with the distributed expression (vector) corresponding to is calculated (step S3-B). Subsequently, when the extraction unit 135 calculates the similarity, it determines whether the calculated similarity exceeds a predetermined threshold value. Subsequently, when the extraction unit 135 determines that the degree of similarity exceeds a predetermined threshold value, the extraction unit 135 extracts the re-search query as a similar query having characteristics similar to the conceptual query “fireworks are beautiful” (step S4). -B).

For example, the calculation unit 134 calculates the similarity between the distributed expression (vector) corresponding to the concept query “fireworks are beautiful” and the distributed expression (vector) corresponding to the re-search query “high-rise apartment” as 0.9. .. Then, the extraction unit 135 determines that the similarity between the distributed expression (vector) corresponding to the concept query “fireworks is beautiful” and the distributed expression (vector) corresponding to the re-search query “high-rise apartment” is a predetermined threshold value (eg, , 0.8) is exceeded. For example, the extraction unit 135 extracts the re-search query “high-rise apartment” as a similar query of the conceptual query “fireworks are beautiful” because the similarity 0.9 exceeds the predetermined threshold of 0.8. .. Similarly, the calculation unit 134 calculates the degree of similarity with the distributed expression (vector) corresponding to the conceptual query “fireworks are beautiful” for all the re-search queries. Then, the extraction unit 135 determines whether or not the degree of similarity exceeds a predetermined threshold for all re-search queries.

Subsequently, when the extraction unit 135 extracts the similar query, the determination unit 136 determines the narrowing-down condition recommended for the user U31 based on the similar query extracted by the extraction unit 135 (step S5-B). ). Specifically, the determination unit 136 determines to recommend the narrowing-down condition corresponding to the re-search query extracted as the similar query. For example, the determination unit 136 determines to recommend the narrowing-down condition corresponding to the re-search query “high-rise apartment” extracted as a similar query of the conceptual query “fireworks is beautiful” to the user U31.

Subsequently, when the determination unit 136 determines the recommended narrowing-down condition, the providing unit 133 provides information on the narrowing-down condition determined by the determination unit 136 (for example, a check box corresponding to the recommended narrowing-down condition is checked. Content) is transmitted to the user U31 (step S6-B).

Next, a content switching process according to the modification will be described with reference to FIG. FIG. 11 is a diagram illustrating an example of a process in which the user terminal according to the modification switches the content. The left diagram of FIG. 11 shows an example of the content C21 shown in FIG. When the search button B1 displayed on the content C21 is pressed in response to the operation of the user U31, the user terminal 10 transmits the concept query “fireworks are beautiful” to the information processing device 100. The information processing apparatus 100 receives the concept query “fireworks are beautiful” from the user U31 (step S2-B shown in FIG. 9). Subsequently, the information processing apparatus 100 executes the processes of steps S3-B to S6-B shown in FIG.

The right diagram of FIG. 11 illustrates an example of the content C22 transmitted by the information processing apparatus 100 to the user U31 in step S6-B illustrated in FIG. 9. The information processing apparatus 100 transmits, to the user U31, the content C22 in which the check box corresponding to the recommended narrowing-down condition is checked among the narrowing-down conditions of the condition search. Upon receiving the content C22, the user terminal 10 displays the content C22 on the screen. When the search button B2 displayed in the content C22 is pressed in response to the operation of the user U31, the user terminal 10 issues a search request to the information processing apparatus 100 to search for real estate information under the narrowing-down condition checked in the check box. Send.

[1-6-3. Application to other fields besides real estate]
In the above example, the information processing apparatus 100 has described the example of recommending the real estate area to the user who has input a predetermined query in the real estate search service. Not limited to the above example, the information processing apparatus 100 is characteristic information indicating characteristics of a predetermined query regarding general content in fields other than real estate, such as products, videos, music, restaurants, food, companies (stock prices, job hunting). To extract.

Specifically, the information processing apparatus 100 is a search service that searches for content in fields other than real estate, such as products, videos, music, restaurants, foods, companies (stock prices, job hunting), in products, videos, music. Get predetermined queries for other areas besides real estate, such as restaurants, food, companies (stock prices, job hunting). Subsequently, the information processing apparatus 100 uses the first learning model in which the features of the plurality of search queries are learned, assuming that the plurality of search queries input by the same user within a predetermined time have similar features. Then, a similar query having characteristics similar to the predetermined query is extracted. Then, the information processing apparatus 100 determines recommendation information based on the extracted similar query.

[2. Second Embodiment]
[2-1. Example of information processing]
Next, a second embodiment will be described. In the first embodiment described above, an example of information processing in which the information processing apparatus 100 extracts a similar query similar to a predetermined query received from a user and determines recommendation information based on the extracted similar query will be described. did. The second embodiment shows an example of information processing in which the information processing apparatus 100A extracts a category to which a predetermined query received from a user belongs, and determines recommendation information based on the extracted category. In addition, in the second embodiment, the same components as those in the first embodiment are designated by the same reference numerals, and the description thereof will be omitted.

The generation device 50 generates the second learning model. Here, the outline of the generation processing of the second learning model by the generation device 50 will be described. The details of the generation processing of the second learning model by the generation device 50 will be described later. Specifically, the generation device 50 uses the first learning model to generate a second learning model that predicts a category to which a predetermined search query belongs from a predetermined search query. More specifically, when generating the first learning model, the generation device 50 acquires the generated first learning model (model data MDT1 of the first learning model M1). When the generation device 50 acquires the first model M1, the generation device 50 generates the second learning model M2 using the acquired first model M1. The generation device 50 re-learns the first model M1 to generate a second learning model M2 having a connection coefficient that is a weight of the learning model different from that of the first model M1. For example, when the search query is input to the learning model, the generation device 50 learns such that the classification result of the distributed expression output by the learning model corresponds to the category to which the search query belongs, thereby generating a predetermined search query. A second learning model M2 that predicts a category to which a predetermined search query belongs is generated.

The information processing device 100A is a server device that provides the real estate information search service R1. The information processing device 100A acquires the model data of the second learning model from the generation device 50. In the following, the model data of the second learning model may be simply referred to as the second learning model. For example, the information processing apparatus 100A recommends a property belonging to the real estate area corresponding to the category into which the predetermined place name query received from the user is classified, as a recommended property, using the second learning model.

[2-2. Configuration of information processing device]
Next, the configuration of the information processing apparatus 100A according to the second embodiment will be described with reference to FIG. FIG. 12 is a diagram illustrating a configuration example of the information processing apparatus 100A according to the second embodiment. As illustrated in FIG. 12, the information processing device 100A includes a communication unit 110, a storage unit 120A, and a control unit 130A. The information processing apparatus 100A includes an input unit (for example, a keyboard and a mouse) that receives various operations from an administrator of the information processing apparatus 100A, and a display unit (for example, a liquid crystal display) for displaying various information. You may have.

(Storage unit 120A)
The storage unit 120A is realized by, for example, a semiconductor memory device such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. As shown in FIG. 12, the storage unit 120A includes a model information storage unit 121, a category information storage unit 122A, a search information storage unit 123, and a content storage unit 124.

(Category information storage unit 122A)
122 A of category information storage parts memorize|store various information regarding the category to which a search query belongs. Specifically, the category information storage unit 122A stores various kinds of information regarding the categories output by the second learning model when the search query is input to the learned second learning model. FIG. 13 shows an example of the category information storage unit according to the second embodiment. In the example illustrated in FIG. 13, the category information storage unit 122A has items such as “search query”, “major classification”, “small classification”, and “probability (%)”.

The “search query ID” indicates a search query input by the user. The “major classification” indicates the major classification of the categories into which the search query is classified. “Minor classification” indicates a minor classification of the category into which the search query is classified. The “probability (%)” indicates the probability for each subclass output by the second learning model when the search query is input to the learned second learning model.

In the example illustrated in FIG. 13, the large classification “search for real estate area” indicates that the large classification of the categories for classifying the search query is the search intention to search for the real estate area. In the example shown in FIG. 13, the large category “search for real estate area” has four smaller categories. The sub-category “Find a high-class residential area” is a category that belongs to the main category “Search for a real estate area”, and the search query classified into the sub-category is entered by the user with the intention of searching for a high-class residential area. Is shown. In addition, the subcategory “Find a downtown area” is a category that belongs to the large category “Find a real estate area”, and a search query that is classified into the subcategory is a search query that is input by the user with the intention of searching for a downtown area. Is shown. In addition, the subcategory “Search for Gulf area” is a category that belongs to the large category “Search for real estate area”, and the search query classified into the subclass is input by the user with the intention of searching for the Gulf area. Is shown. In addition, the sub-category “Search for suburb area” is a category that belongs to the subcategory “Search for real estate area”, and the search query classified into the sub-category is input by the user with the intention of searching the suburb area. Is shown.

In the example illustrated in FIG. 13, the probability (90) of the search query “place name #11” is 90, and the probability that the search query “place name #11” is classified as a query input with the intention of searching for a high-class residential area is 90. % Is shown. Further, the probability (%) “0” of the search query “place name #11” indicates that the probability that the search query “place name #11” is classified into the query input with the intention of searching the downtown area is 0%. .. The probability (%) “10” of the search query “place name #11” indicates that the probability that the search query “place name #11” is classified into the query input with the intention of searching the Gulf area is 10%. .. In addition, the probability (%) “0” of the search query “place name #11” indicates that the probability that the search query “place name #11” is classified into the query input with the intention of searching the suburb area is 0%. ..

(Control unit 130A)
Returning to the description of FIG. 12, the control unit 130A is a controller, and various programs (corresponding to an example of the information processing program) stored in the storage device inside the information processing apparatus 100A by, for example, a CPU, an MPU, and the like are stored. It is realized by executing the RAM as a work area. The control unit 130A is a controller and is realized by an integrated circuit such as ASIC or FPGA.

As illustrated in FIG. 12, the control unit 130A includes an acquisition unit 131, a generation unit 132, a provision unit 133, a calculation unit 134A, an extraction unit 135A, and a determination unit 136A, and information described below. Realize or execute the action of processing. Note that the internal configuration of the control unit 130A is not limited to the configuration shown in FIG. 12, and may be another configuration as long as it is a configuration for performing information processing described later.

(Acquisition unit 131)
The acquisition unit 131 acquires the second learning model. More specifically, the acquisition unit 131 acquires the second learning model generated by the generation device 50 from the generation device 50. When acquiring the second learning model, the acquisition unit 131 stores the acquired second learning model in the model information storage unit 121.

(Calculator 134A)
The calculation unit 134A calculates the probability that the search query belongs to a predetermined category for each category. Specifically, when the provision unit 133 accepts a predetermined search query, the calculation unit 134A inputs the search query into the second learning model acquired by the acquisition unit 131 so that the predetermined query falls into a predetermined category. Probability of belonging is calculated for each category. For example, when the provision unit 133 receives the predetermined place name query, the calculation unit 134A inputs the received place name query into the second learning model acquired by the acquisition unit 131 and places the received place name query into a predetermined category. Probability of belonging is calculated for each category. For example, the calculation unit 134A has four categories (small categories) of the received predetermined place name queries, “search for high-class residential area”, “search for downtown area”, “search for coastal area”, “search for suburban area”. The probability of belonging to each category (small classification) is calculated for each category (small classification).

(Extractor 135A)
The extraction unit 135A extracts, as the characteristic information, a category to which a predetermined query belongs. For example, the extraction unit 135A determines, for each category (small classification), whether or not the probability of each category (small classification) calculated by the calculation unit 134A exceeds a predetermined threshold value. Subsequently, when the probability that the received search query belongs to the predetermined category exceeds a predetermined threshold, the extraction unit 135A extracts the predetermined category as the category into which the received search query is classified. For example, when the received predetermined place name query has a probability of 90% belonging to the “search for a high-class residential area” category and the predetermined threshold value is 80%, the extraction unit 135A determines that the received predetermined place name query is As a category to be classified, a category "search for a high-class residential area" is extracted.

Further, the extraction unit 135A extracts feature information by using a learning model that outputs a distributed expression of a predetermined search query as output information when a predetermined search query is input as input information. For example, the extraction unit 135A uses the second learning model generated by using the first learning model that outputs the distributed expression of the predetermined search query as the output information when the predetermined search query is input as the input information. Then, the category to which the predetermined query belongs is extracted as the characteristic information.

In addition, the extraction unit 135A uses a learning model that learns the characteristics of a plurality of search queries by learning so that the distributed expressions of a pair of search queries that are continuously input within a predetermined time are similar, Extract feature information. For example, the extraction unit 135A uses the first learning model in which the characteristics of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries continuously input within a predetermined time period are similar to each other. A category to which a predetermined query belongs is extracted as the characteristic information using the second learning model generated as described above.

Further, the extraction unit 135A learns that a plurality of search queries including a character string delimited by a predetermined delimiter have similar characteristics as the plurality of search queries input by the same user within a predetermined time. By doing so, the feature information is extracted using the learning model in which the features of the plurality of search queries are learned. For example, the extraction unit 135A learns that a plurality of search queries including a character string delimited by a predetermined delimiter have similar characteristics as the plurality of search queries input by the same user within a predetermined time period. By doing so, the category to which the predetermined query belongs is extracted as the feature information by using the second learning model generated by using the first learning model in which the features of the plurality of search queries are learned.

In addition, the extraction unit 135A extracts the characteristic information by using the learning model in which the characteristics of the plurality of search queries are learned by learning that the plurality of search queries that are randomly extracted have different characteristics. .. For example, the extraction unit 135A learns that a plurality of search queries that are randomly extracted have different characteristics, and thus the first learning model that is generated by using the first learning model that has learned the characteristics of the plurality of search queries. A category to which a predetermined query belongs is extracted as feature information using the two learning model.

In addition, the extraction unit 135A extracts feature information using a learning model in which the features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are randomly extracted differ. .. For example, the extraction unit 135A learns such that the distributed expressions of a pair of search queries that are randomly extracted differ from each other, and thus the first learning model generated by using the first learning model that has learned the features of the plurality of search queries. A category to which a predetermined query belongs is extracted as feature information using the two learning model.

(Decision section 136A)
The determination unit 136A determines recommendation information to be recommended to the user who has input the predetermined query, based on the category extracted by the extraction unit 135A. For example, the determination unit 136A determines to recommend the property belonging to the high-class residential area to the user who has input a predetermined place name query based on the “search for high-class residential area” category extracted by the extraction unit 135A. To do.

[2-3. Prediction processing flow]
Next, the procedure of the prediction process according to the second embodiment will be described with reference to FIG. FIG. 14 is a flowchart showing a prediction processing procedure according to the second embodiment. In the example illustrated in FIG. 14, the information processing device 100A acquires the search query and the second learning model (step S301). Subsequently, when the information processing apparatus 100A acquires the search query and the second learning model, the information processing apparatus 100A estimates the category into which the search query is classified, using the second learning model (step S302).

[2-4. Information processing flow]
Next, a procedure of information processing according to the second embodiment will be described with reference to FIG. FIG. 15 is a flowchart showing an information processing procedure according to the second embodiment. In the example illustrated in FIG. 15, the information processing device 100A determines whether a search query has been accepted (step S401). When the search query is not accepted (step S401; No), the information processing apparatus 100A waits until the search query is accepted.

On the other hand, when the information processing device 100A receives the search query (step S401; Yes), the information processing device 100A calculates the probability of the search query belonging to a predetermined category for each category (step S402).

Subsequently, when the probability that the search query belongs to the predetermined category is calculated, the information processing apparatus 100A determines whether the calculated probability exceeds a predetermined threshold value (step S403). When the calculated probability does not exceed the predetermined threshold value (step S403; No), the information processing apparatus 100A ends the process.

On the other hand, when the calculated probability exceeds the predetermined threshold value (step S403; Yes), the information processing apparatus 100A extracts the category into which the received search query is classified (step S404). Specifically, when the probability that the received search query belongs to a predetermined category exceeds a predetermined threshold, information processing apparatus 100A extracts the predetermined category as the category into which the received search query is classified. Subsequently, when the information processing apparatus 100A extracts the category, the information processing apparatus 100A determines recommended information based on the extracted category (step S405).

[3. Learning model generation processing]
[3-1. Generation processing of first learning model]
Next, the flow of generation processing of the first learning model will be described with reference to FIG. FIG. 16 is a diagram illustrating an example of the first learning model generation process according to the embodiment. In the example illustrated in FIG. 16, the generation device 50 includes a pair of search queries Q11 “Roppongi pasta” and a search query Q12 “Roppongi Italian” that are continuously input by the same user U1 within a predetermined time. A search query is extracted (step S11).

Subsequently, the generation device 50 inputs the extracted search query Q11 into the first model M1 and outputs a vector BQV11 that is a distributed expression of the search query Q11. Here, the vector BQV11 is a distributed expression of the search query Q11 just output from the output layer of the first model M1, and represents a distributed expression before feedback is given to the first model M1 (before learning). Further, the generation device 50 inputs the extracted search query Q12 into the first model M1 and outputs the vector BQV12 which is a distributed expression of the search query Q12. Here, the vector BQV12 is a distributed expression of the search query Q12 just output from the output layer of the first model M1, and represents a distributed expression before feedback is applied to the first model M1 (before learning). In this way, the generation device 50 outputs the vector BQV11, which is the distributed expression of the search query Q11, and the vector BQV12, which is the distributed expression of the search query Q12 (step S12).

Subsequently, the generation device 50 includes a pair of search queries including a search query Q11 (“Roppongi pasta”) and a search query Q12 (“Roppongi Italian”) that are continuously input by the same user U1 within a predetermined time. Is presumed to be a search query input with a predetermined search intent (for example, a search intent "to search for a restaurant at a certain place"), and therefore, it is assumed that search queries Q11 have similar features to each other. The first model M1 is trained so that the distributed expression (vector QV11) and the distributed expression (vector QV12) of the search query Q12 paired with the search query Q11 are similar in the distributed expression space. For example, the magnitude of the angle between the vector BQV11, which is the distributed expression of the search query Q11 before feedback (before learning) to the first model M1, and the vector BQV12, which is the distributed expression of the search query Q12, is Θ. Further, the size of the angle formed by the vector QV11, which is the distributed expression of the search query Q11 after feedback is applied to the first model M1 (after learning), and the vector QV12, which is the distributed expression of the search query Q12, is Φ. At this time, the generation device 50 trains the first model M1 so that Φ becomes smaller than Θ. For example, the generation device 50 calculates the value of the cosine similarity between the vector BQV11 and the vector BQV12. Further, the generation device 50 calculates the value of the cosine similarity between the vector QV11 and the vector QV12. Then, the generation device 50 sets the first model so that the value of the cosine similarity between the vectors QV11 and QV12 is larger than the value of the cosine similarity between the vectors BQV11 and BQV12 (so that the value approaches 1). Train M1. In this way, the generation device 50 learns the first model M1 so that the two vectors, which are a pair of distributed expressions corresponding to the pair of search queries, are similar in the distributed expression space, and thus the distributed expression from the search query. A first model M1 that outputs (vector) is generated (step S13). Note that the generation device 50 may calculate the similarity between distributed expressions (vectors) based on any index as long as it is an index applicable as a distance measure between vectors, not limited to the cosine similarity. Good. Further, the generation device 50 may train the first model M1 based on any index as long as it is an index applicable as a distance measure between vectors. For example, the generating device 50 calculates a value of a predetermined distance function such as a Euclidean distance between distributed expressions (vectors), a distance in a non-Euclidean space such as a hyperbolic space, a Manhattan distance, a Mahalanobis distance, or the like. Subsequently, the generation device 50 may train the first model M1 so that the value of the predetermined distance function between the distributed expressions (vectors) (that is, the distance in the distributed expression space) becomes small.

Next, the flow of the generation process of the first learning model will be described in more detail with reference to FIG. Note that in the description of FIG. 17, the portions overlapping the description of FIG. 16 will be omitted as appropriate. FIG. 17 is a diagram showing a generation process of the first learning model according to the embodiment. In the following, the first learning model will be appropriately referred to as the first model (or the first model M1). In the example illustrated in FIG. 17, the distributed representation (vector) output by the first model M1 generated by the generation device 50 is mapped to the distributed representation space. The generation device 50 trains the first model M1 so that the distributed expression of the predetermined search query and the distributed expressions of the other search queries paired with the predetermined search query are closely mapped in the distributed expression space. ..

In the example illustrated in the upper part of FIG. 17, the generation device 50 includes a search query Q11 (“Roppongi pasta”) and a search query Q12, which are four search queries continuously input by the same user U1 within a predetermined time. (“Roppongi Italian”), search query Q13 (“Akasaka pasta”), and search query Q14 (“Azabu pasta”) are extracted. The generation device 50 extracts four search queries whose time intervals at which each search query is input by the same user U1 are within a predetermined time. The generation device 50 extracts a plurality of search queries in which a time interval at which a pair of search queries described below is input by the same user U1 is within a predetermined time. When the generating device 50 arranges the search queries in the input order, the generating device 50 extracts the four search queries input in the order of the search query Q11, the search query Q12, the search query Q13, and the search query Q14. When the four search queries are extracted, the generation device 50 defines two search queries that are adjacent in time series as a pair of search queries, and is a pair of three search queries (search query Q11, search query Q12), (Search query Q12, search query Q13) and (search query Q13, search query Q14) are extracted (step S21-1). Note that the generation device 50 may extract a plurality of search queries in which all the search queries are input by the same user U1 within a predetermined time. Then, the generation device 50 selects two search queries from the plurality of extracted search queries, regardless of whether they are adjacent in time series, and sets the selected two search queries as a pair of search queries. You may extract.

Then, the generation device 50 inputs the extracted search query Q1k (k=1, 2, 3, 4) into the first model M1 and distributes the search query Q1k (k=1, 2, 3, 4). The vector BQV1k (k=1, 2, 3, 4) that is the expression is output. Here, the vector BQV1k (k=1, 2, 3, 4) is a distributed representation of the search query Q1k (k=1, 2, 3, 4) just output from the output layer of the first model M1. , A distributed expression before feedback is given to the first model M1 (before learning) (step S22-1).

Then, the generation device 50 determines that the pair of search queries continuously input by the same user U1 within a predetermined time period has a predetermined search intention (for example, “a certain place (in the vicinity of Minato-ku, Tokyo) at a restaurant). It is presumed that the search queries are input according to the search intention of "searching for". Therefore, the distributed representation (vector QV11) of the search query Q11 is paired with the search query Q11 as having similar features. The first model M1 is trained so that the distributed expression (vector QV12) of the search query Q12 is similar in the distributed expression space. In addition, the generating device 50 makes the distributed expression of the search query Q12 (vector QV12) and the distributed expression of the search query Q13 (vector QV13) paired with the search query Q12 similar to each other in the distributed expression space. The model M1 is trained. In addition, the generation device 50 makes the distributed expression (vector QV13) of the search query Q13 and the distributed expression (vector QV14) of the search query Q14 paired with the search query Q13 similar to each other in the distributed expression space. The model M1 is trained. In this way, the generation device 50 learns the first model M1 so that the two vectors, which are a pair of distributed expressions corresponding to the pair of search queries, are similar in the distributed expression space, and thus the distributed expression from the search query. A first model M1 that outputs (vector) is generated (step S23-1).

As a result of the information processing shown in the upper part of FIG. 17, the vector QV1k (k=1, 2, 3, 4), which is the distributed expression of the search query Q1k (k=1, 2, 3, 4), is close to the distributed expression space. It is shown that the position is mapped as the cluster CL11. For example, the search query Q1k (k=1, 2, 3, 4) is a set of search queries searched by the user U1 under the search intention of “searching for a restaurant in a certain place (near Minato-ku, Tokyo)”. Is estimated to be That is, the search query Q1k (k=1, 2, 3, 4) is a search query that is searched under the search intent of “searching for a restaurant in a certain place (in the vicinity of Minato-ku, Tokyo)”. , Are presumed to be search queries having similar features to each other. Here, when the predetermined search query input with the search intention of “search for a restaurant in a certain location (Near Minato-ku, Tokyo)” is input to the first model, the generation device 50 sets the position of the cluster CL11. A distributed representation as mapped can be output. As a result, for example, the generating device 50 extracts the search query corresponding to the distributed expression mapped to the position of the cluster CL11 to search for a restaurant in a certain place (in the vicinity of Minato-ku, Tokyo). It is possible to extract a search query according to. Therefore, the generation device 50 can appropriately interpret the meaning of the search query.

In the example illustrated in the lower part of FIG. 17, the generation device 50 includes a search query Q21 (“refrigerator 400L”) and a search query Q22, which are three search queries continuously input by the same user U2 within a predetermined time. (“Refrigerator Medium-sized”) and search query Q23 (“Refrigerator Medium-sized recommended”) are extracted. After arranging the search queries in the input order, the generation device 50 extracts the three search queries input in the order of the search query Q21, the search query Q22, and the search query Q23. When the three search queries are extracted, the generation device 50 regards two search queries that are adjacent in time series as a pair of search queries, and is a pair of two search queries (search query Q21, search query Q22), (Search query Q22, search query Q23) is extracted (step S21-2).

Subsequently, the generation device 50 inputs the extracted search query Q2m (m=1, 2, 3) into the first model M1 and outputs a vector that is a distributed expression of the search query Q2m (m=1, 2, 3). BQV2m (m=1, 2, 3) is output. Here, the vector BQV2m (m=1, 2, 3) is a distributed representation of the search query Q2m (m=1, 2, 3) just output from the output layer of the first model M1, and is the first model. A distributed expression before feedback is given to M1 (before learning) is shown (step S22-2).

Subsequently, the generation device 50 inputs a pair of search queries continuously input by the same user U2 within a predetermined time period with a predetermined search intention (for example, a search intention of “searching a medium-sized refrigerator”). Since it is estimated that the search query is a search query that has been generated, it is assumed that the search query Q21 has a distributed expression (vector QV21) and a search query Q22 that forms a pair with the search query Q21 (a vector QV22). ) And the first model M1 are learned so that they are similar to each other in the distributed expression space. In addition, the generation device 50 makes the distributed expression (vector QV22) of the search query Q22 and the distributed expression (vector QV23) of the search query Q23 paired with the search query Q22 similar to each other in the distributed expression space. The model M1 is trained. In this way, the generation device 50 learns the first model M1 so that the two vectors, which are a pair of distributed expressions corresponding to the pair of search queries, are similar in the distributed expression space, and thus the distributed expression from the search query. A first model M1 that outputs (vector) is generated (step S23-2).

As a result of the information processing shown in the lower part of FIG. 17, the vector QV2m (m=1, 2, 3), which is the distributed expression of the search query Q2m (m=1, 2, 3), is located in the cluster CL21 near the distributed expression space. Is mapped as. For example, the search query Q2m (m=1, 2, 3) is presumed to be a set of search queries searched by the user U2 under the search intention of “searching for a medium-sized refrigerator”. That is, Q2m (m=1, 2, 3) is a search query having similar characteristics in that it is a search query searched under the search intention of “searching for a medium-sized refrigerator”. Presumed. Here, when the predetermined search query input with the search intention of “search for a medium-sized refrigerator” is input to the first model, the generation device 50 outputs a distributed expression that is mapped to the position of the cluster CL21. can do. Thereby, for example, the generating device 50 extracts the search query corresponding to the distributed expression mapped to the position of the cluster CL21, thereby extracting the search query according to the search intention of “inspecting the medium-sized refrigerator”. You can Therefore, the generation device 50 can appropriately interpret the meaning of the search query.

Further, the generation device 50 according to the present invention is a search query having mutually different characteristics in that the plurality of search queries that are randomly extracted are search queries that are searched under different search intentions. And the first model M1 is learned. Specifically, the generation device 50 causes the distributed expression of the predetermined search query and the distributed expression of the search query randomly extracted irrespective of the predetermined search query to be mapped far in the distributed expression space. The first model M1 is trained. In the example illustrated in FIG. 17, it is assumed that the generation device 50 extracts the search query Q21 when the search query is randomly extracted regardless of the search query Q11. In this case, the generation device 50 causes the distributed representation of the search query Q11 (vector QV11) and the distributed representation of the search query Q21 (vector QV21) randomly extracted irrespective of the search query Q11 to be far in the distributed representation space. The first model M1 is trained so as to be mapped to As a result, the vector QV1k, which is a distributed expression of the search query Q1k (k=1, 2, 3, 4), which is searched under the search intention of “searching for a restaurant in a certain place (Near Minato-ku, Tokyo)”. It is a distributed representation of a cluster CL11 including (k=1, 2, 3, 4) and a search query Q2m (m=1, 2, 3) searched under the search intention of “searching for a medium-sized refrigerator”. The cluster CL21 including the vector QV2m (m=1, 2, 3) is mapped far in the distributed representation space. That is, the generation device 50 according to the present invention trains the first model M1 so that the distributed expressions of a plurality of search queries that are randomly extracted are different, and thus the distributed expressions of the search queries having different search intentions are distributed expressions. It is possible to output to a distant position in space.

As a result of the distributed representation (vector) output by the first model M1 generated by the generation device 50 being mapped in the distributed representation space, in addition to the cluster CL11 and the cluster CL21 described above, a predetermined user may specify a predetermined expression. A cluster CL12 or a cluster CL22, which is a set of distributed expressions (vectors) of a plurality of search queries input in time, is generated.

As described above, the generation device 50 acquires the search query input by the user. In addition, the generation device 50 learns that a plurality of search queries input by the same user within a predetermined time have a similar feature among the acquired search queries, and thus a predetermined search query can be used. A first model that predicts the characteristic information of the search query is generated. That is, the generation device 50 according to the present invention is similar to each other in that the plurality of search queries continuously input within a predetermined time period are search queries searched under a predetermined search intention. The first model is trained by regarding it as a search query having characteristics. Specifically, the generation device 50 learns the first model such that the distributed expressions of the plurality of search queries input by the same user within a predetermined time are similar to each other, and thereby the predetermined model is obtained from the predetermined search query. A first model that outputs a distributed expression that includes the characteristic information of the search query is generated. That is, the generation device 50 according to the present invention learns the first model M1 so that the distributed expressions of a plurality of search queries continuously input within a predetermined time are similar to each other. It is possible to output the distributed expression of the search query searched by in a near position on the distributed expression space. Thereby, the generation device 50 enables the meaning (search intention) of the search query to be output (interpreted) according to the context of the user who inputs the search query. Therefore, the generation device 50 can appropriately interpret the meaning of the search query. Furthermore, the generation device 50 extracts the search query corresponding to the distributed expression that is mapped in the vicinity of the distributed expression that includes the characteristic information of the predetermined search query, and thus the predetermined search query responds to the search intention. Search queries can be extracted. That is, the generation device 50 makes it possible to analyze the search trend of the user in consideration of the search intention and context of the user who inputs the search query. Therefore, the generation device 50 can improve the accuracy of analysis of the search trend of the user. In addition, the first model M1 generated by the generation device 50 can be made to function as a part of the search system. Alternatively, the generation device 50 uses the distributed representation of the search query output by the first model M1 as input information to another system (for example, a search engine) that uses the feature information of the search query predicted by the first model M1. Can also be provided. Accordingly, the search system can select the content output as the search result based on the characteristic information of the search query predicted by the first model M1. That is, the search system can select the content output as the search result in consideration of the search intention or context of the user who inputs the search query. Furthermore, the search system calculates the similarity between the distributed expression of the character string included in the content output as the search result and the distributed expression of the search query based on the characteristic information of the search query predicted by the first model M1. It will be possible. Then, the search system can determine the display order of the content output as the search result based on the calculated similarity. That is, the search system can determine the display order of the content output as the search result in consideration of the search intention or context of the user who inputs the search query. Therefore, the generation device 50 can improve usability in the search service.

[3-2. Second learning model generation process]
Next, the flow of the second learning model generation process will be described with reference to FIG. FIG. 18 is a diagram illustrating an example of a second learning model generation process according to the embodiment. In the following, the second learning model will be appropriately referred to as the second model (or the second model M2). In the example illustrated in the upper part of FIG. 18, the generation device 50 includes a search query Q11 (“Roppongi pasta”) and a search query Q12, which are four search queries continuously input by the same user U1 within a predetermined time. (“Roppongi Italian”), search query Q13 (“Akasaka pasta”), and search query Q14 (“Azabu pasta”) are extracted. The generation device 50 extracts a plurality of search queries in which the time intervals at which the search queries are input by the same user U1 are within a predetermined time. In addition, the generation device 50 extracts a plurality of search queries in which the time interval at which each search query pair is input by the same user U1 is within a predetermined time. Here, it is assumed that the four search queries are search queries in which the search query Q11, the search query Q12, the search query Q13, and the search query Q14 are input in this order by the user U1 within a predetermined time. When the four search queries are extracted, the generation device 50 defines two search queries that are adjacent in time series as a pair of search queries, and is a pair of three search queries (search query Q11, search query Q12), (Search query Q12, search query Q13), (search query Q13, search query Q14) are extracted. When the generator 50 extracts three pairs of search queries, it inputs the extracted search queries Q1k (k=1, 2, 3, 4) to the first model M1 (step S31). Note that the generation device 50 may extract a plurality of search queries in which all the search queries are input by the same user U1 within a predetermined time. Then, the generation device 50 selects two search queries from the plurality of extracted search queries, regardless of whether they are adjacent in time series, and sets the selected two search queries as a pair of search queries. You may extract.

Then, the generation device 50 outputs the vector BQV1k (k=1, 2, 3, 4), which is the distributed expression of the search query Q1k (k=1, 2, 3, 4), as the output data of the first model M1. Yes (step S32). Here, the vector BQV1k (k=1, 2, 3, 4) is a distributed representation of the search query Q1k (k=1, 2, 3, 4) just output from the output layer of the first model M1. , A distributed expression before feedback is applied to the first model M1 (before learning).

Here, the search query Q1k (k=1, 2, 3, 4) continuously input by the same user U1 within a predetermined time is, for example, “some place (near Minato-ku, Tokyo)” by the user U1. It is presumed to be a set of search queries searched under the search intention of "searching for restaurants in." That is, the search query Q1k (k=1, 2, 3, 4) is a search query that is searched under the search intent of “searching for a restaurant in a certain place (in the vicinity of Minato-ku, Tokyo)”. , Are presumed to be search queries having similar features to each other. Therefore, the generating device 50 generates the first model that predicts the characteristic information of the predetermined search query from the predetermined search query by learning that the successively input search queries have similar characteristics. Step S33). Specifically, the generation device 50 learns that the distributed expressions of continuously input search queries are similar to each other, and thus the first model M1 predicts the distributed expression of the predetermined search query from the predetermined search query. To generate. For example, the generation device 50 may make the distributed expression (vector QV11) of the search query Q11 and the distributed expression (vector QV12) of the search query Q12 paired with the search query Q11 similar to each other in the distributed expression space. The model M1 is trained. In addition, the generating device 50 makes the distributed expression of the search query Q12 (vector QV12) and the distributed expression of the search query Q13 (vector QV13) paired with the search query Q12 similar to each other in the distributed expression space. The model M1 is trained. In addition, the generation device 50 makes the distributed expression (vector QV13) of the search query Q13 and the distributed expression (vector QV14) of the search query Q14 paired with the search query Q13 similar to each other in the distributed expression space. The model M1 is trained.

On the right side of the upper part of FIG. 18, the variance of the search query Q1k (k=1, 2, 3, 4) input by the same user U1 within a predetermined time as the output result of the learned first model M1. It is shown that the expression vector QV1k (k=1, 2, 3, 4) is mapped as the cluster CL11 in the distributed expression space. In this way, the generation device 50 generates the first learning model M1 that has learned the characteristics of the plurality of search queries input by the same user within a predetermined time.

After generating the first model M1, the generation device 50 acquires the generated first model M1 (model data MDT1 of the first model M1). When the generation device 50 acquires the first model M1, the generation device 50 generates the second learning model M2 using the acquired first model M1. Specifically, the generation device 50 re-learns the first model M1 to generate the second model M2 having a connection coefficient that is a weight of the learning model different from that of the first model M1. More specifically, the generation device 50 uses the first model M1 to generate a second learning model M2 that predicts a category to which a predetermined search query belongs from a predetermined search query (step S34).

In the example illustrated in the lower part of FIG. 18, when the search query is input to the second model M2, the generation device 50 CAT11 (“search for a restaurant”), CAT12 (“search for a product”), CAT13 (“ A second model M2 that predicts which one of the four categories of "reserve a restaurant") and CAT14 ("purchase a product") is generated. Specifically, when the search query is input to the second model M2 as the input information, the generation device 50 generates the second model M2 that outputs the probability that the search query belongs to the category for each category as the output information. To do. For example, the generation device 50 learns a set of a search query and a category (one of CAT11 to CAT14) to which the search query belongs as the correct answer data of the second model M2.

It should be noted that the fact that the search query belongs to CAT11 (“search for restaurants”) indicates that the search query is a search query that was input with the intention of searching for restaurants. Belonging to CAT12 (“search for product”) indicates that the search query is a search query input with the intention of searching for a product. Further, the fact that the search query belongs to CAT13 (“reserve a restaurant”) indicates that the search query is a search query input with the intention of reserving a restaurant. Further, the fact that the search query belongs to CAT14 (“purchase a product”) indicates that the search query is a search query input with the intention of purchasing the product.

Specifically, when the search query is input to the learning model, the generation device 50 learns such that the classification result of the distributed expressions output by the learning model corresponds to the category to which the search query belongs, thereby making A second model M2 that predicts a category to which a predetermined search query belongs is generated from the search query of. Then, for example, when the search query is input to the second model M2 as the input information, the generation device 50 outputs the probability that the search query belongs to the category as the output information for each of the categories CAT11 to CAT14. To generate.

For example, when the search query Q11 (“Roppongi pasta”) is input to the second model M2 as the input information (step S35), the generation device 50 outputs the distributed representation of the search query Q11 (“Roppongi pasta”) as the output information. The vector BQV11 is output. Here, the vector BQV11 is a distributed expression of the search query Q11 just output from the output layer of the second model M2, and represents a distributed expression before feedback is applied to the second model M2 (before learning). Here, it is assumed that the correct category to which the search query Q11 (“Roppongi pasta”) belongs is CAT11 (“Find a restaurant”). In this case, the generation device 50 sets the probability that the vector BQV11, which is the distributed expression of the output search query Q11 (“Roppongi pasta”), is classified into CAT11 (“Search for restaurants”) exceeds a predetermined threshold. The second model M2 is trained. The generation device 50 uses the correct answer data prepared in advance to train the second model. The generation device 50 may generate correct answer data of the second model M2. Then, the generation device 50 may train the second model M2 using the generated correct answer data. Specifically, the generation device 50 determines the correct category to which the search query belongs, based on the post-search behavior of the user who searched for the search query. More specifically, the generation device 50 associates, with respect to the user who has searched for a predetermined search query, a predetermined action in which the ratio of users who have performed a predetermined action after the search exceeds a predetermined threshold, to the correct category. Decide as an action. For example, as a ratio of users who searched for the search query Q11 (“Roppongi pasta”) to take a predetermined action after the search, a ratio of users who took action to search for a restaurant was 90%, and to search for a product after the search. It is assumed that the percentage of users who have caused a search is 0%, the percentage of users who have made an action to reserve a restaurant after the search is 10%, and the percentage of users who have made an action to purchase a product after the search is 0%. In this case, since the ratio of the users who have taken the action to search for the restaurant exceeds the predetermined threshold value (for example, 90%), the generation device 50 searches for the action to search for the restaurant in the search query Q11 (“Roppongi pasta”). Determined as the action corresponding to the category. Then, since the generation device 50 has determined that the action corresponding to the correct answer category is the action of searching for a restaurant, the correct answer category to which the search query Q11 (“Roppongi pasta”) belongs to is CAT11 (“search for a restaurant”). decide.

For example, when the search query Q11 (“Roppongi pasta”) is input to the second model M2 before learning, the generator 50 classifies the vector BQV11, which is a distributed expression, into CAT11 (“Find a restaurant”). 80% probability, 0% probability of being classified as CAT12 (“find a product”), 20% probability of being classified as CAT13 (“reserve a restaurant”), CAT14 (“purchase product”) It is assumed that the probability of being classified as is output as 0%. In this case, the generation device 50 trains the second model M2 so that the probability that the vector BQV11, which is a distributed representation, is classified into CAT11 (“search for a restaurant”) exceeds a predetermined threshold value (for example, 90%). .. In addition, the generation device 50 performs dispersion so that the probability that the vector BQV11, which is a distributed expression, is classified into CAT11 (“search for a restaurant”) exceeds a predetermined threshold value (for example, 90%). The second model M2 is trained so as to reduce the probability that the expression vector BQV11 is classified into another category CAT13 (“reserve restaurant”) to 10%.

As described above, when the predetermined search query is input as the input information, the generation device 50 is configured such that the probability that the distributed expression of the predetermined search query is classified into the correct answer category as the output information exceeds the predetermined threshold value. Train the model. Then, when a predetermined search query is input as input information, the generation device 50 sets a category whose probability that the distributed expression of the predetermined search query belongs to the category exceeds a predetermined threshold as a category of the predetermined search query. Output. For example, when the search query Q11 (“Roppongi pasta”) is input as input information to the learned second model M2, the generation device 50 generates the vector BQV11 that is a distributed expression of the search query Q11 (“Roppongi pasta”). Since the probability of belonging to the category CAT11 (“search for restaurants”) exceeds 90%, the category to which the search query belongs is output as output information as CAT11 (“searches for restaurants”) (step S36). In this way, the generation device 50 generates the second model that predicts the category of the predetermined search query from the predetermined search query by learning the set of the search query and the correct category of the search query (step S37). ..

Generally, it is considered that a user performs a search a plurality of times with a certain intention, and therefore it is assumed that search queries continuously input within a predetermined time have similar search intentions. Therefore, the generation device 50 according to the present invention is similar to each other in that the plurality of search queries continuously input within a predetermined time period are search queries searched under a predetermined search intention. The first model M1 is trained by considering it as a search query having characteristics. Thereby, the generation device 50 can cause the first model M1 to learn the characteristics of the search query in consideration of the search intention. Then, the generation device 50 efficiently generates the second model that predicts the category of the predetermined search query from the predetermined search query by utilizing the first model M1 that has learned the characteristics of the search query in consideration of the search intention. can do. As a result, the generation device 50 enables the search query to be classified into categories in consideration of the search intention of the user who inputs the search query. Further, conventionally, it has been necessary to prepare a sufficient amount of correct answer data in order to classify search queries into categories and obtain high classification accuracy. However, since the search queries themselves are diverse and have long tail characteristics, it is very troublesome and difficult to label the correct answer categories corresponding to a large number of search queries. Here, instead of labeling the correct answer category, the generation device 50 uses the user's search intention (the context of the user who entered the search query) as a kind of correct answer to learn the second model that predicts the category of the search query. be able to. Thereby, the generation device 50 can train the second model without manually labeling the correct category of the search query. That is, the generation device 50 can obtain sufficient classification accuracy even when the correct answer data is small. In addition, the generation device 50 can obtain higher classification accuracy when there are many correct data. Therefore, the generation device 50 can improve the classification accuracy of the search query.

[3-3. Configuration of information processing device]
Next, the configuration of the generation device 50 according to the embodiment will be described with reference to FIG. FIG. 19 is a diagram illustrating a configuration example of the generation device 50 according to the embodiment. As illustrated in FIG. 19, the generation device 50 includes a communication unit 51, a storage unit 53, and a control unit 52. The generating device 50 has an input unit (for example, a keyboard or a mouse) that receives various operations from an administrator of the generating device 50, and a display unit (for example, a liquid crystal display) for displaying various information. May be.

(Communication unit 51)
The communication unit 51 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 51 is connected to the network by wire or wirelessly, and transmits and receives information between the user terminal 10 and the search server 20, for example.

(Storage unit 53)
The storage unit 53 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory (Flash Memory), or a storage device such as a hard disk or an optical disk. As shown in FIG. 19, the storage unit 53 includes a query information storage unit 531, a vector information storage unit 532, a classification definition storage unit 533, a category information storage unit 534, and a model information storage unit 535.

(Query information storage unit 531)
The query information storage unit 531 stores various information regarding the search query input by the user. FIG. 20 shows an example of the query information storage unit according to the embodiment. In the example illustrated in FIG. 20, the query information storage unit 531 has items such as “user ID”, “date and time”, “search query”, and “search query ID”.

The “user ID” indicates identification information for identifying the user who has input the search query. The “date and time” indicates the date and time when the search server accepted the search query from the user. "Search query" indicates a search query input by the user. The “search query ID” indicates identification information for identifying the search query input by the user.

In the example shown in the first record of FIG. 20, the search query (search query Q11) identified by the search query ID “Q11” corresponds to the search query Q11 shown in FIG. The user ID “U1” indicates that the user who inputs the search query Q11 is the user (user U1) identified by the user ID “U1”. Further, the date and time "2018/9/1 PM 17:00" indicates that the date and time when the search server receives the search query Q11 from the user U1 is 17:00 pm on September 1, 2018. The search query "Roppongi pasta" indicates the search query Q11 input by the user U1. Specifically, the search query “Roppongi pasta” is a character string in which “Roppongi” indicating a place name and “pasta” indicating the type of food are separated by a space, which is a delimiter.

(Vector information storage unit 532)
The vector information storage unit 532 stores various kinds of information regarding vectors, which are distributed expressions of search queries. FIG. 21 shows an example of the vector information storage unit according to the embodiment. In the example illustrated in FIG. 21, the vector information storage unit 532 has items such as “vector ID”, “search query ID”, and “vector information”.

"Vector ID" indicates identification information for identifying a vector that is a distributed expression of a search query. The “search query ID” indicates identification information for identifying the search query corresponding to the vector. “Vector information” indicates an N-dimensional vector that is a distributed expression of a search query. The vector that is the distributed expression of the search query is, for example, a 128-dimensional vector.

In the example shown in the first record of FIG. 21, the vector identified by the vector ID “QV11” (vector QV11) corresponds to the vector QV11 that is the distributed expression of the search query Q11 shown in FIG. The search query (search query Q11) identified by the search query ID “Q11” indicates that the search query corresponding to the vector QV11 is the search query Q11. The vector information “QVDT11” indicates an N-dimensional vector that is a distributed expression of the search query Q11.

(Classification definition storage unit 533)
The classification definition storage unit 533 stores various kinds of information regarding the definition of the category into which the search query is classified. FIG. 22 shows an example of the classification definition storage unit according to the embodiment. In the example shown in FIG. 22, the classification definition storage unit 533 has items such as “large classification ID”, “large classification”, “small classification ID”, and “small classification”.

The “major classification” indicates the major classification of the categories into which the search query is classified. The “major classification ID” indicates identification information for identifying the major classification. In the example illustrated in FIG. 22, the large classification “purchasing behavior type” corresponds to the large classification described in the example illustrated in the lower part of FIG. 1. The large classification "purchasing behavior system" indicates a large classification of categories into which the search query is classified based on the purchasing behavior of the user. In the example shown in FIG. 22, the large classification “purchasing behavior type” further has four small classifications. The large classification ID “CAT1” indicates identification information for identifying the large classification “purchasing behavior type”.

“Minor classification” indicates a minor classification of the category into which the search query is classified. The “small category ID” indicates identification information for identifying the small category. In the example shown in FIG. 22, the small category “search for restaurants” is a category belonging to the large category “purchasing behavior system”, and the search query classified into the small categories is input by the user with the intention of searching for restaurants. Indicates that the search query was performed. The small classification ID “CAT11” indicates identification information for identifying the small classification “search for a restaurant”.

The small category “search for products” is a category belonging to the large category “purchasing behavior type”, and indicates that the search queries classified into the small categories are search queries input by the user with the intention of searching for products. .. The small classification ID “CAT12” indicates identification information for identifying the small classification “search for products”.

The small category “reserve a restaurant” is a category belonging to the large category “purchasing behavior type”, and the search query classified into the small category is a search query input by the user with the intention of reserving a restaurant. Indicates that. The small classification ID “CAT13” indicates identification information for identifying the small classification “reserve a restaurant”.

The small category “Purchase goods” is a category belonging to the large category “Purchasing behavior type”, and the search queries classified into the small categories are search queries input by the user with the intention of purchasing the product. Show. The small classification ID “CAT14” indicates identification information for identifying the small classification “purchase a product”.

(Category information storage unit 534)
The category information storage unit 534 stores various kinds of information regarding the category to which the search query belongs. Specifically, the category information storage unit 534 stores various kinds of information regarding the category output by the second learning model when the search query is input to the learned second learning model. FIG. 23 shows an example of the category information storage unit according to the embodiment. In the example illustrated in FIG. 23, the category information storage unit 534 has items such as “search query ID”, “large classification ID”, “small classification ID”, and “probability (%)”.

The “search query ID” indicates identification information for identifying the search query input by the user. In the example shown in FIG. 23, the search query (search query Q11) identified by the search query ID “Q11” corresponds to the search query Q11 shown in FIG.

The “major classification ID” indicates identification information for identifying the major classification. The “small category ID” indicates identification information for identifying the small category. The “probability (%)” indicates the probability for each subclass output by the second learning model when the search query is input to the learned second learning model. In the example shown in FIG. 23, the probability (%) “90” indicates that the probability that the search query Q11 is classified into the category CAT11 is 90%.

(Model information storage unit 535)
The model information storage unit 535 stores various information regarding the learning model generated by the generation device 50. FIG. 24 shows an example of the model information storage unit according to the embodiment. In the example shown in FIG. 24, the model information storage unit 535 has items such as “model ID” and “model data”.

The “model ID” indicates identification information for identifying the learning model generated by the generation device 50. The “model data” indicates model data of the learning model generated by the generation device 50. For example, “model data” stores data for converting a search query into a distributed expression.

In the example shown in the first record of FIG. 24, the learning model identified by the model ID “M1” corresponds to the first model M1 shown in FIG. The model data “MDT1” indicates the model data (model data MDT1) of the first model M1 generated by the generation device 50.

The model data MDT1 includes an input layer to which a search query is input, an output layer, a first element belonging to any layer from the input layer to the output layer and other than the output layer, the first element, and the first element. A second element whose value is calculated based on the weight of one element, and a distributed expression of the search query input to the input layer is output from the output layer according to the search query input to the input layer. Thus, the generation device 50 may be operated.

Here, it is assumed that the model data MDT1 is realized by the regression model represented by “y=a1*x1+a2*x2+...+ai*xi”. In this case, the first element included in the model data MDT1 corresponds to the input data (xi) such as x1 and x2. The weight of the first element corresponds to the coefficient ai corresponding to xi. Here, the regression model can be regarded as a simple perceptron having an input layer and an output layer. When each model is regarded as a simple perceptron, the first element can be regarded as a node included in the input layer, and the second element can be regarded as a node included in the output layer.

Further, it is assumed that the model data MDT1 is realized by a neural network having one or a plurality of intermediate layers such as DNN (Deep Neural Network). In this case, the first element included in the model data MDT1 corresponds to any node included in the input layer or the intermediate layer. The second element corresponds to the node at the next stage, which is the node to which the value is transmitted from the node corresponding to the first element. The weight of the first element corresponds to the connection coefficient, which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

The generation device 50 calculates the distributed expression using a model having an arbitrary structure such as the above-described regression model or neural network. Specifically, the model data MDT1 is set with a coefficient so as to output a distributed expression when a search query is input. The generation device 50 calculates a distributed expression using such model data MDT1.

In the above example, the model data MDT1 is the model (hereinafter referred to as model X1) that outputs the distributed expression of the search query when the search query is input. However, the model data MDT1 according to the embodiment may be a model generated based on a result obtained by repeating input/output of data to/from the model X1. For example, the model data MDT1 may be a model (hereinafter, referred to as model Y1) learned by inputting the distributed expression output by the model X1 when the search query is input. Alternatively, the model data MDT1 may be a model learned so that the search query is input and the output value of the model Y1 is output.

When the generation device 50 performs the estimation process using GAN (Generative Adversarial Networks), the model data MDT1 may be a model forming a part of GAN.

In the example shown in the second record of FIG. 24, the learning model identified by the model ID “M2” corresponds to the second model M2 shown in FIG. The model data “MDT2” indicates the model data (model data MDT2) of the second model M2 generated by the generation device 50.

The model data MDT2 includes an input layer to which a search query is input, an output layer, a first element belonging to any layer from the input layer to the output layer and other than the output layer, the first element, and the first element. A second element whose value is calculated based on the weight of one element, and a probability that the search query input to the input layer belongs to each category according to the search query input to the input layer The generator 50 may be operated to output from.

Here, it is assumed that the model data MDT2 is realized by the regression model represented by “y=a1*x1+a2*x2+...+ai*xi”. In this case, the first element included in the model data MDT2 corresponds to the input data (xi) such as x1 and x2. The weight of the first element corresponds to the coefficient ai corresponding to xi. Here, the regression model can be regarded as a simple perceptron having an input layer and an output layer. When each model is regarded as a simple perceptron, the first element can be regarded as a node included in the input layer, and the second element can be regarded as a node included in the output layer.

It is also assumed that the model data MDT2 is realized by a neural network having one or a plurality of intermediate layers such as DNN (Deep Neural Network). In this case, the first element included in the model data MDT2 corresponds to any node included in the input layer or the intermediate layer. The second element corresponds to the node at the next stage, which is the node to which the value is transmitted from the node corresponding to the first element. The weight of the first element corresponds to the connection coefficient, which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

The generation device 50 calculates the probability that the search query belongs to each category using a model having an arbitrary structure such as the regression model and the neural network described above. Specifically, the model data MDT2 is set with a coefficient so as to output the probability that the search query belongs to each category when the search query is input. The generation device 50 uses such model data MDT2 to calculate the probability that the search query belongs to each category.

In the above example, the model data MDT2 is the model (hereinafter, referred to as model X2) that outputs the distributed expression of the search query when the search query is input. However, the model data MDT2 according to the embodiment may be a model generated based on a result obtained by repeating input/output of data to/from the model X2. For example, the model data MDT2 may be a model (hereinafter, referred to as model Y2) learned by inputting the distributed expression output by the model X2 when the search query is input. Alternatively, the model data MDT2 may be a model learned so that the search query is input and the output value of the model Y2 is output.

Further, when the generation device 50 performs the estimation process using GAN (Generative Adversarial Networks), the model data MDT2 may be a model forming a part of GAN.

(Control unit 52)
Returning to the description of FIG. 19, the control unit 52 is a controller, and is stored in a storage device inside the generation device 50 by, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). It is realized by executing various programs (corresponding to an example of the generation program) using the RAM as a work area. The control unit 52 is a controller, and is realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

In addition, the control unit 52 performs information processing according to the first model M1 (model data MDT1) stored in the model information storage unit 535, in response to the search query input in the input layer, in each layer other than the output layer. The computer is made to function so that the distributed representation is output from the output layer by performing the calculation based on the first element and the weight of the first element with each element that belongs to the first element.

In addition, the control unit 52 performs information processing in accordance with the second model M2 (model data MDT2) stored in the model information storage unit 535 on each layer other than the output layer with respect to the search query input to the input layer. The computer is operated so that the probability that the search query belongs to each category is output from the output layer by performing an operation based on the first element and the weight of the first element with each element that belongs to the first element.

As illustrated in FIG. 19, the control unit 52 includes an acquisition unit 521, an extraction unit 522, and a generation unit 523, and realizes or executes the operation of information processing described below. Note that the internal configuration of the control unit 52 is not limited to the configuration shown in FIG. 19, and may be another configuration as long as it is a configuration for performing information processing described later.

(Acquisition unit 521)
The acquisition unit 521 acquires various information. Specifically, the acquisition unit 521 acquires the search query input by the user from the search server 20. When acquiring the search query input by the user, the acquisition unit 521 stores the acquired search query in the query information storage unit 531. The acquisition unit 521 also acquires vector information about a vector that is a distributed expression of the search query. When acquiring the vector information, the acquisition unit 521 stores the acquired vector information in the vector information storage unit 532. In addition, the acquisition unit 521 acquires information defining the classification of the search query and the category to which the search query belongs. When acquiring the search query and the classification definition information that defines the classification of the category to which the search query belongs, the acquisition unit 521 stores the acquired classification definition information in the classification definition storage unit 533. The acquisition unit 521 also acquires category information regarding the category to which the search query belongs. When acquiring the category information, the acquisition unit 521 stores the acquired category information in the category information storage unit 534.

(Extractor 522)
The extraction unit 522 extracts various information. Specifically, the extraction unit 522 extracts a plurality of search queries input by the same user within a predetermined time, from the search queries acquired by the acquisition unit 521. For example, the extraction unit 522 extracts a plurality of search queries in which the time intervals at which the search queries are input by the same user are within a predetermined time. Subsequently, the extraction unit 522 extracts a pair of search queries continuously input by the same user within a predetermined time from a plurality of search queries input by the same user within a predetermined time. For example, the extraction unit 522 extracts a plurality of search queries in which the time interval at which each search query pair is input by the same user is within a predetermined time. For example, the extraction unit 522, among the search queries acquired by the acquisition unit 521, is a search query Q11 (“Roppongi pasta”) that is four search queries continuously input by the same user U1 within a predetermined time. ), search query Q12 (“Roppongi Italian”), search query Q13 (“Akasaka pasta”), and search query Q14 (“Azabu pasta”). When the extraction unit 522 arranges the search queries in the input order, the extraction unit 522 extracts the four search queries input in the order of the search query Q11, the search query Q12, the search query Q13, and the search query Q14. Subsequently, when the four search queries are extracted, the extraction unit 522 defines two search queries that are adjacent in time series as a pair of search queries, and is a pair of three search queries (search query Q11, search query). Q12), (search query Q12, search query Q13), and (search query Q13, search query Q14) are extracted. The extraction unit 522 may extract a plurality of search queries in which all the search queries have been input by the same user within a predetermined time. Then, the extraction unit 522 selects two search queries from the plurality of extracted search queries, regardless of whether they are adjacent in time series, and sets the selected two search queries as a pair of search queries. You may extract.

Further, the extraction unit 522 extracts a predetermined search query and other search queries unrelated to the predetermined search query from the search queries acquired by the acquisition unit 521. For example, the extraction unit 522 extracts a predetermined search query from the search queries acquired by the acquisition unit 521. Subsequently, the extraction unit 522 randomly extracts another search query from the search queries acquired by the acquisition unit 521, regardless of the predetermined search query.

(Generator 523)
The generation unit 523 generates various information. Specifically, the generation unit 523 learns, among the search queries acquired by the acquisition unit 521, that a plurality of search queries input by the same user within a predetermined time have similar characteristics. , A learning model for predicting the characteristic information of the predetermined search query from the predetermined search query is generated. Specifically, the generation unit 523 learns the learning model so that the distributed expressions of the plurality of search queries input by the same user within a predetermined time are similar to each other, thereby performing a predetermined search from the predetermined search query. A learning model for predicting query feature information is generated. For example, the generation unit 523 generates a learning model by learning so that the distributed expressions of a pair of search queries continuously input within a predetermined time period are similar to each other. For example, the generation unit 523 calculates the similarity value of the distributed representation (vector) before learning of the pair of search queries. In addition, the generation unit 523 calculates the value of the similarity of the distributed expressions (vectors) after learning of the pair of search queries. Subsequently, the generation unit 523 trains the learning model such that the similarity value of the distributed expression (vector) after learning is larger than the similarity value of the distributed expression (vector) before learning. In this way, the generating unit 523 learns the learning model so that the two vectors, which are a pair of distributed expressions corresponding to the pair of search queries, are similar in the distributed expression space, and thus the distributed expression (vector ) Is output to generate a learning model. More specifically, the generation unit 523 generates a learning model that outputs a distributed expression (vector) from a search query using the DSSM technique that uses LSTM, which is a type of RNN, for distributed expression generation. For example, the generation unit 523 determines that, as the correct data of the learning model, a pair of search queries input by the same user within a predetermined time have similar characteristics, and a distributed expression (vector) of the predetermined search query is used. , A predetermined search query and a distributed expression (vector) of another search query paired with each other are learned so as to exist close to each other in the distributed expression space. Further, when the generation unit 523 generates the first learning model, the generation unit 523 stores the generated first learning model (model data MDT1) in the model information storage unit 535 in association with the identification information for identifying the first learning model.

[3-4. Example of first learning model]
Here, an example of the first learning model generated by the generation device 50 will be described with reference to FIG. 25. FIG. 25 is a diagram illustrating an example of the first learning model according to the embodiment. In the example shown in FIG. 25, the first learning model M1 generated by the generation device 50 is composed of three layers of LSTM RNNs. In the example illustrated in FIG. 25, the extraction unit 522 includes a pair of search queries Q11 “Roppongi pasta” and a search query Q12 “Roppongi Italian” that are continuously input by the same user U1 within a predetermined time. Extract the search query. The generation unit 523 inputs the search query Q11 extracted by the extraction unit 522 into the input layer of the first learning model M1 (step S41).

Then, the generation unit 523 outputs a 256-dimensional vector BQV11, which is a distributed expression of the search query Q11, from the output layer of the first learning model M1. Further, the generation unit 523 inputs the search query Q12 extracted by the extraction unit 522 into the input layer of the first learning model M1. Subsequently, the generation unit 523 outputs the 256-dimensional vector BQV12, which is the distributed expression of the search query Q12, from the output layer of the first learning model M1 (step S42).

Subsequently, the generation unit 523 learns so that the distributed expressions (vectors) of the two consecutively input search queries are similar to each other, and thus the first learning model M1 that outputs the distributed expressions (vectors) from the search query. Is generated (step S43). For example, the magnitude of the angle between the vector BQV11, which is the distributed expression of the search query Q11 before feedback (before learning) to the first learning model M1, and the vector BQV12, which is the distributed expression of the search query Q12, is Θ. In addition, the size of the angle formed by the vector QV11, which is the distributed expression of the search query Q11 after feedback is applied to the first learning model M1 (after learning), and the vector QV12, which is the distributed expression of the search query Q12, is Φ. .. At this time, the generation unit 523 trains the first learning model M1 so that Φ becomes smaller than Θ. For example, the generation unit 523 calculates the value of the cosine similarity between the vector BQV11 and the vector BQV12. Further, the generation unit 523 calculates the value of the cosine similarity between the vector QV11 and the vector QV12. Then, the generation unit 523 sets the learning model M1 so that the value of the cosine similarity between the vectors QV11 and QV12 is larger than the value of the cosine similarity between the vectors BQV11 and BQV12 (so that the value approaches 1). To learn. In this way, the generation unit 523 learns the first learning model M1 so that the two vectors, which are a pair of distributed expressions corresponding to the pair of search queries, are similar in the distributed expression space, and thus the search query is distributed. A first learning model M1 that outputs an expression (vector) is generated. The generation unit 523 is not limited to the cosine similarity, and may calculate the similarity between the distributed expressions (vectors) based on any index as long as it is an index applicable as a distance measure between vectors. Good. Further, the generation unit 523 may train the learning model M1 based on any index as long as it is an index applicable as a distance measure between vectors. For example, the generation unit 523 calculates a value of a predetermined distance function such as a Euclidean distance between distributed expressions (vectors), a distance in a non-Euclidean space such as a hyperbolic space, a Manhattan distance, a Mahalanobis distance, or the like. Subsequently, the generation unit 523 may train the learning model M1 so that the value of the predetermined distance function between the distributed expressions (vectors) (that is, the distance in the distributed expression space) becomes small.

Further, the generation unit 523 learns that a plurality of search queries including a character string delimited by a predetermined delimiter have similar characteristics as the plurality of search queries input by the same user within a predetermined time. By doing so, the first learning model is generated. For example, the generation unit 523 uses a search query “Roppongi pasta” in which “Roppongi” indicating a place name and “pasta” indicating the type of food are separated by a space that is a delimiter, and “Roppongi” indicating a place name. The first learning model is generated by learning that the search query “Roppongi Italian”, which is separated by a space that is a delimiter between the characters “Italian” indicating the type of food, has similar characteristics.

In addition, the generation unit 523 generates the first learning model by learning that among the search queries acquired by the acquisition unit 521, a plurality of randomly extracted search queries have different characteristics. Specifically, the generation unit 523 generates a first learning model by learning so that the distributed expressions of a pair of search queries that are randomly extracted among the search queries acquired by the acquisition unit 521 are different. To do. For example, the generation unit 523 causes the distributed representation of the predetermined search query extracted by the extraction unit 522 and the distributed representation of the search query randomly extracted irrespective of the predetermined search query to be far apart in the distributed representation space. The first learning model M1 is trained so as to be mapped.

The generation unit 523 also generates a second learning model. Specifically, the generation unit 523 refers to the model information storage unit 535 and acquires the first learning model (model data MDT1 of the first learning model M1) generated by the generation unit 523. Subsequently, the generation unit 523 uses the acquired first learning model to generate a second learning model that predicts the category to which the predetermined search query belongs from the predetermined search query. When the generation unit 523 acquires the first model M1, the generation unit 523 generates the second learning model M2 using the acquired first model M1. The generating unit 523 re-learns the first model M1 to generate the second model M2 having a connection coefficient that is a weight of the learning model different from that of the first model M1. Specifically, when the search query is input to the learning model, the generation unit 523 learns such that the classification result of the distributed expressions output by the learning model corresponds to the category to which the search query belongs, and thus the predetermined result is obtained. A second model M2 that predicts a category to which a predetermined search query belongs is generated from the search query of.

Specifically, when the search query is input to the learning model, the generation unit 523 learns such that the classification result of the distributed expressions output by the learning model corresponds to the category to which the search query belongs, and thus the predetermined result is obtained. A second learning model that predicts a category to which a predetermined search query belongs is generated from the search query. The generation unit 523 generates a second learning model that outputs the probability of each category to which the search query belongs as output information when the search query is input as input information to the learning model. For example, the generation unit 523 uses the first model M1 to calculate the probability that the distributed expression of the search query is classified as the output information into the category when the predetermined search query is input as the input information into the learning model. The second model M2 to be output for each is generated. When a predetermined search query is input as the input information, the generation unit 523 trains the second model so that the probability that the distributed expression of the predetermined search query is classified into the correct category as the output information exceeds the predetermined threshold. .. Then, when the predetermined search query is input as the input information, the generation unit 523 sets the category whose probability that the distributed expression of the predetermined search query belongs to the category exceeds the predetermined threshold as the category of the predetermined search query. The second model M2 to be output is generated. Further, when the generation unit 523 generates the second learning model, the generation unit 523 stores the generated second learning model (model data MDT2) in the model information storage unit 535 in association with the identification information for identifying the second learning model.

For example, the generation unit 523 refers to the model information storage unit 535 illustrated in FIG. 24 and acquires the first model M1 (model data MDT1 of the first model M1). Subsequently, the generation unit 523 refers to the classification definition storage unit 533 illustrated in FIG. 22 and selects the large classification of the categories into which the search query is classified. Subsequently, when selecting the large classification, the generation unit 523 learns a set of the search query and the small classification to which the search query belongs as the learning data of the second model M2.

For example, it is assumed that the correct category to which the search query Q11 (“Roppongi pasta”) belongs is CAT11 (“Find a restaurant”). When the search query Q11 (“Roppongi pasta”) is input to the second model M2 as input information, the generation unit 523 uses the distributed representation of the search query Q11 (“Roppongi pasta”) from the output layer of the second model M2. It outputs a certain vector BQV11. Here, the vector BQV11 is a distributed expression of the search query Q11 just output from the output layer of the second model M2, and represents a distributed expression before feedback is given to the second model M2 (before learning). In this case, the generation unit 523 causes the probability that the vector BQV11, which is the distributed expression of the output search query Q11 (“Roppongi pasta”), is classified into the correct answer category CAT11 (“Search for restaurants”) exceeds a predetermined threshold. Thus, the second model M2 is trained.

For example, when the search query Q11 (“Roppongi pasta”) is input to the second model M2 before learning, the generation unit 523 classifies the vector BQV11, which is a distributed expression, into CAT11 (“Find a restaurant”). 80% probability, 0% probability of being classified as CAT12 (“find a product”), 20% probability of being classified as CAT13 (“reserve a restaurant”), CAT14 (“purchase product”) It is assumed that the probability of being classified as is output as 0%. In this case, the generation unit 523 trains the second model M2 so that the probability that the vector BQV11, which is a distributed expression, is classified into CAT11 (“search for a restaurant”) exceeds a predetermined threshold value (for example, 90%). .. In addition, the generation unit 523 performs dispersion so that the probability that the vector BQV11, which is a distributed expression, is classified into CAT11 (“search for a restaurant”) exceeds a predetermined threshold value (for example, 90%), and the distribution is performed. The second model M2 is trained so as to reduce the probability that the expression vector BQV11 is classified into another category CAT13 (“reserve restaurant”) to 10%. Subsequently, when the search query Q11 (“Roppongi pasta”) is input as input information to the learned second model M2, the generation unit 523 generates a vector BQV11 that is a distributed expression of the search query Q11 (“Roppongi pasta”). Belongs to the category CAT11 (“search for restaurants”) exceeds 90%, the category to which the search query belongs is output as output information as CAT11 (“searches for restaurants”).

Note that the generation unit 523 may select any number of large classifications as the large classifications. Then, when the search query is input to the second model M2 as the input information, the generation unit 523 determines the probability of belonging to each of the small categories belonging to the arbitrary number of the large categories selected by the search query as the output information for each of the small categories. The second model M2 to be output to may be generated. Further, the generation unit 523 may select all the large classifications as the large classifications. Then, the generation unit 523 may generate the second model M2 that outputs the probability of belonging to each of the small classifications for every small classification when the search query is input to the second model M2.

[3-5. Example of second learning model]
Here, an example of the second learning model generated by the generation device 50 will be described with reference to FIG. FIG. 26 is a diagram illustrating an example of the second learning model according to the embodiment. In the example shown in FIG. 26, the second learning model M2 generated by the generating device 50 is generated using the first learning model M1. That is, the generation device 50 re-learns the first learning model M1 to generate the second learning model M2 having a connection coefficient that is a weight of the learning model different from that of the first learning model M1.

More specifically, the second learning model M2 generated by the generating device 50 is composed of three layers of LSTM RNNs, like the first learning model M1. In the example illustrated in FIG. 26, the extraction unit 522 inputs the search query Q11 “Roppongi pasta” input by the user U1 into the input layer of the second learning model M2 (step S51).

Subsequently, the generation unit 523 outputs the 256-dimensional vector BQV11, which is the distributed expression of the search query Q11, from the output layer of the second learning model M2 (step S52).

Then, the generation unit 523 outputs the probability that the vector BQV11, which is the distributed expression of the search query Q11, is classified into each category (step S53).

Subsequently, the generation unit 523 learns the second learning model M2 so as to increase the probability that the vector BQV11, which is the distributed expression of the search query Q11, is classified into the correct answer category, thereby changing the category of the search query from the search query. A second model to be predicted is generated (step S54).

[3-6. Flow of processing for generating first learning model]
Next, the procedure of the first learning model generation process according to the embodiment will be described with reference to FIG. 27. FIG. 27 is a flowchart showing a procedure for generating the first learning model according to the embodiment.

In the example illustrated in FIG. 27, the generation device 50 acquires the search query input by the user (step S1001).

Subsequently, the generation device 50 extracts a plurality of search queries input by the same user within a predetermined time (step S1002).

Subsequently, the generation device 50 generates a first learning model that predicts the characteristic information of a predetermined search query from the predetermined search query by learning that the extracted plurality of search queries have similar characteristics ( Step S1003).

[3-7. Flow of second learning model generation processing]
Next, the procedure of the second learning model generation process according to the embodiment will be described with reference to FIG. FIG. 28 is a flowchart showing the procedure of the second learning model generation process according to the embodiment.

In the example illustrated in FIG. 28, the generation device 50 acquires the first learning model (model data MDT1 of the first learning model M1) (step S2001).

Then, the generation device 50 generates a second learning model that predicts a category of a predetermined search query from a predetermined search query using the first learning model (step S2002).

[4. effect〕
As described above, the information processing device 100 according to the first embodiment includes the extraction unit 135 and the determination unit 136. The extraction unit 135 determines that a plurality of search queries input by the same user within a predetermined time have similar features by using a learning model that has learned the features of the plurality of search queries. Feature information indicating features is extracted. The determination unit 136 determines recommendation information to be recommended to the user who has input the predetermined query, based on the characteristic information extracted by the extraction unit 135.

With this, the information processing apparatus 100 can recommend information based on the characteristic information indicating the characteristics of the predetermined search query to the user who is interested in the predetermined search query. That is, the information processing apparatus 100 can recommend the interest of the user or information that matches the interest. In addition, in general, a user who is interested in a specific field such as a user who visits a search service, but has little knowledge about the field, who has little knowledge of the field, is likely to find an appropriate search. The problem is that I can't think of a query. The information processing apparatus 100 according to the present invention can recommend recommended information based on an appropriate search query according to a search intention, based on a search query input by a user who has little knowledge. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

Further, the extraction unit 135 extracts, as the characteristic information, a similar query that is a search query having characteristics similar to a predetermined query. The determination unit 136 determines recommendation information to be recommended to the user who has input the predetermined query, based on the similar query extracted by the extraction unit 135.

As a result, the information processing apparatus 100 can recommend information based on a similar query having characteristics similar to the predetermined search query to a user who is interested in the predetermined search query. That is, the information processing apparatus 100 can recommend the interest of the user or information that matches the interest. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

In addition, the extraction unit 135 extracts a similar query having the same attribute as the predetermined query. The determining unit 136 determines information about the similar query, which is the recommendation information, based on the similar query extracted by the extracting unit 135.

As a result, the information processing apparatus 100 can recommend information based on a similar query having the same attributes as the predetermined query to users who are interested in the predetermined search query. Therefore, the information processing apparatus 100 can recommend more appropriate information to the user.

In addition, the extraction unit 135 extracts a similar query indicating a real estate area, which is a search query having characteristics similar to the predetermined query indicating the real estate area, as a similar query having the same attribute as the predetermined query. The determination unit 136 determines information regarding the real estate area, which is the recommendation information, based on the similar query extracted by the extraction unit 135.

Accordingly, the information processing apparatus 100 can recommend a real estate area having characteristics similar to the predetermined real estate area to a user who is interested in the predetermined real estate area. Therefore, the information processing apparatus 100 can recommend an appropriate real estate area to the user.

Further, the determining unit 136 determines a query candidate for re-search, which is recommendation information, based on the similar query extracted by the extracting unit 135.

As a result, the information processing apparatus 100 can recommend an appropriate narrowing-down condition candidate to a user who is interested in a predetermined search query but does not know the specific search condition.

The information processing apparatus 100 further includes a generation unit 132 and a calculation unit 134. The generation unit 132 determines that a plurality of search queries input by the same user within a predetermined time have similar features, and uses a learning model that learns the features of the plurality of search queries to perform a predetermined search query. Generate a distributed representation of. The calculation unit 134 calculates the similarity between the distributed expression of the predetermined search query generated by the generation unit 132 and the distributed expression of another search query different from the predetermined search query generated by the generation unit 132. The extraction unit 135 extracts another search query whose similarity calculated by the calculation unit 134 exceeds a predetermined threshold as a similar query.

The information processing apparatus 100 can recommend information based on characteristic information indicating characteristics of a predetermined search query to a user who is interested in the predetermined search query. That is, the information processing apparatus 100 can recommend the interest of the user or information that matches the interest. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

The extraction unit 135 also extracts the characteristic information by using a learning model that outputs a distributed expression of the predetermined search query as the output information when the predetermined search query is input as the input information. In addition, the extraction unit 135 uses a learning model in which the characteristics of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are continuously input within a predetermined time are similar to each other. Extract feature information. The extraction unit 135 also learns that a plurality of search queries that are input by the same user within a predetermined time have similar characteristics to each other that include a character string delimited by a predetermined delimiter. By doing so, the feature information is extracted using the learning model in which the features of the plurality of search queries are learned. Further, the extraction unit 135 extracts the characteristic information by using the learning model in which the characteristics of the plurality of search queries are learned by learning the plurality of randomly extracted search queries as having different characteristics. .. In addition, the extraction unit 135 extracts feature information using a learning model in which the features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are randomly extracted differ. ..

As a result, the information processing apparatus 100 can extract appropriate characteristic information in consideration of the user's search intention. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

The information processing apparatus 100A according to the second embodiment includes an extraction unit 135A and a determination unit 136A. The extraction unit 135A extracts, as the characteristic information, a category to which a predetermined query belongs. The determination unit 136A determines recommendation information to be recommended to the user who has input the predetermined query, based on the category extracted by the extraction unit 135A.

Accordingly, the information processing apparatus 100A can recommend information based on a category to which a predetermined search query belongs to a user who is interested in the predetermined search query. That is, the information processing apparatus 100A can recommend the interest of the user or information matching the interest. Therefore, the information processing apparatus 100 can recommend appropriate information to the user.

[5. Hardware configuration]
In addition, the information processing apparatus 100 according to the first embodiment, the information processing apparatus 100A according to the second embodiment, and the generation apparatus 50 according to the embodiments described above are implemented by a computer 1000 having a configuration illustrated in FIG. 29, for example. Will be realized. FIG. 29 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing device 100, the information processing device 100A, and the generation device 50. The computer 1000 includes a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I/F) 1500, input/output interface (I/F) 1600, and 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 starts up, a program dependent on the hardware of the computer 1000, and the like.

The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. The communication interface 1500 receives data from another device via a predetermined communication network, sends the data to the CPU 1100, and transmits the data generated by the CPU 1100 to another device via the predetermined communication network.

The CPU 1100 controls output devices such as a display and a printer and input devices 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 the program or data 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 device 100, the information processing device 100A, or the generation device 50, the CPU 1100 of the computer 1000 executes the program loaded on the RAM 1200 to cause the control unit 130, the control unit 130A, or The function of the control unit 52 is realized. 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 obtained from other devices via a predetermined communication network.

As described above, some of the embodiments of the present application have been described in detail based on the drawings, but these are examples, and various modifications based on the knowledge of those skilled in the art, including the modes described in the section of the disclosure of the invention, It is possible to implement the present invention in other forms with improvements.

[6. Other]
Further, of the processes described in the above-described embodiment and modification, all or part of the processes described as being automatically performed may be manually performed, or described as manually performed. It is also possible to automatically carry out all or part of the above-mentioned processing by a known method. In addition, the processing procedures, specific names, information including various data and parameters shown in the above-mentioned documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

Further, each component of each device shown in the drawings is functionally conceptual and does not necessarily have to be physically configured as shown. That is, the specific form of distribution/integration of each device is not limited to that shown in the figure, and all or part of the device may be functionally or physically distributed/arranged in arbitrary units according to various loads and usage conditions. It can be integrated and configured.

Further, the above-described embodiments and modified examples can be appropriately combined within a range in which the processing content is not inconsistent.

Also, the above-mentioned "section (module, unit)" can be read as "means" or "circuit". For example, the extraction unit can be read as an extraction unit or an extraction circuit.

1 Information Processing System 10 User Terminal 20 Search Server 50 Generation Device 100 Information Processing Device 110 Communication Unit 120 Storage Unit 121 Model Information Storage Unit 122 Vector Information Storage Unit 123 Search Information Storage Unit 124 Content Storage Unit 130 Control Unit 131 Acquisition Unit 132 Generation Part 133 Providing part 134 Calculation part 135 Extracting part 136 Determining part

Claims (14)

A feature indicating a feature of a predetermined query by using a learning model in which the features of the plurality of search queries are learned, assuming that the plurality of search queries input by the same user within a predetermined time have similar features An extraction unit for extracting information,
A determination unit that determines recommendation information recommended to the user who has input the predetermined query, based on the characteristic information extracted by the extraction unit;
An information processing apparatus comprising: The extraction unit is
As the characteristic information, a similar query that is a search query having characteristics similar to the predetermined query is extracted,
The determining unit is
The information processing apparatus according to claim 1, wherein recommendation information recommended for a user who has input the predetermined query is determined based on the similar query extracted by the extraction unit. The extraction unit is
Extract a similar query that has the same attributes as the predetermined query,
The determining unit is
The information processing apparatus according to claim 2, wherein information regarding the similar query that is the recommendation information is determined based on the similar query extracted by the extraction unit. The extraction unit is
As the similar query having the same attributes as the predetermined query, a search query having characteristics similar to the predetermined query indicating the real estate area, and extracting a similar query indicating the real estate area,
The determining unit is
The information processing apparatus according to claim 3, wherein information regarding the real estate area, which is the recommendation information, is determined based on the similar query extracted by the extraction unit. The determining unit is
The information processing apparatus according to any one of claims 1 to 4, wherein a candidate for a query for re-search that is the recommendation information is determined based on a similar query extracted by the extraction unit. As a plurality of search queries input by the same user within a predetermined time have similar features, a distributed representation of the predetermined search queries is calculated using a learning model that learns the features of the plurality of search queries. A generator to generate,
A calculation unit that calculates the degree of similarity between the distributed expression of the predetermined search query generated by the generation unit and the distributed expression of another search query that is different from the predetermined search query generated by the generation unit;
Further equipped with,
The extraction unit is
The information processing apparatus according to any one of claims 2 to 5, wherein another search query whose similarity calculated by the calculator exceeds a predetermined threshold is extracted as the similar query. The extraction unit is
The characteristic information is extracted by using a learning model that outputs a distributed expression of the predetermined search query as output information when a predetermined search query is input as input information. The information processing apparatus described in any one of 1. The extraction unit is
Extracting the feature information by using a learning model in which the features of the plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries that are continuously input within the predetermined time are similar to each other. The information processing apparatus according to any one of claims 1 to 7, characterized in that. The extraction unit is
By learning as a plurality of search queries input by the same user within a predetermined time period, a plurality of search queries including a character string delimited by a predetermined delimiter have similar characteristics. The information processing apparatus according to any one of claims 1 to 8, wherein the characteristic information is extracted using a learning model in which a characteristic included in the search query is learned. The extraction unit is
The feature information is extracted by using a learning model that learns the features of the plurality of search queries by learning as a plurality of randomly extracted search queries having different features. The information processing device according to claim 1. The extraction unit is
By learning so that the distributed expressions of a pair of search queries that are randomly extracted are different, the feature information is extracted using a learning model that has learned the features of the plurality of search queries. The information processing apparatus according to claim 1. The extraction unit is
As the characteristic information, a category to which the predetermined query belongs is extracted,
The determining unit is
The recommendation information recommended for the user who has input the predetermined query is determined based on the category extracted by the extraction unit. apparatus. An information processing method executed by a computer,
A feature indicating a feature of a predetermined query by using a learning model in which the features of the plurality of search queries are learned, assuming that the plurality of search queries input by the same user within a predetermined time have similar features An extraction step for extracting information,
A determination step of determining recommendation information to be recommended to the user who has input the predetermined query, based on the characteristic information extracted by the extraction step;
An information processing method comprising: A feature indicating a feature of a predetermined query by using a learning model in which the features of the plurality of search queries are learned, assuming that the plurality of search queries input by the same user within a predetermined time have similar features Extraction means for extracting information,
Determination means for determining recommendation information to be recommended to the user who has input the predetermined query, based on the characteristic information extracted by the extraction means,
An information processing program that causes a computer to execute.

Images