JP6553793B1 - Information processing apparatus, information processing method, and information processing program - Google Patents

Abstract

To improve the classification accuracy of a search query. An information processing apparatus according to the present application includes an acquisition unit and a generation unit. The acquisition unit acquires a first learning model in which features of a plurality of search queries are learned, assuming that the plurality of search queries input within a predetermined time by the same user have similar features. The generation unit generates a second learning model that predicts a category to which a predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit. [Selection] Figure 1

Description

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

  Conventionally, techniques for classifying various information into categories are known. Specifically, a learning model corresponding to each category is generated by a group of teacher data classified for each category. Then, there is known a technique of inputting data to be classified into the generated learning model and outputting the probability that the data to be classified belongs to a category corresponding to the learning model.

JP-A-2018-97397

  However, the above-mentioned prior art can not always improve the classification accuracy of the search query. Specifically, it is necessary to prepare a sufficient amount of teacher data in order to classify search queries into categories and obtain high classification accuracy. However, in the above-described prior art, a learning model corresponding to each category is merely generated from a group of teacher data classified for each category, so that a sufficient amount of teacher data cannot always be prepared. Absent. Therefore, the above-mentioned prior art can not necessarily improve the classification accuracy of the search query.

  The present application has been made in view of the above, and an object thereof is to provide an information processing apparatus, an information processing method, and an information processing program capable of improving the classification accuracy of a search query.

  The information processing apparatus according to the present application acquires a first learning model in which a plurality of search queries input within a predetermined time by the same user have similar characteristics and learns the characteristics of the plurality of search queries. And a generator configured to generate a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit. It is characterized by

  According to one aspect of the embodiment, there is an effect that the classification accuracy of the search query can be increased.

FIG. 1 is a diagram illustrating an example of information processing according to the embodiment. FIG. 2 is a diagram illustrating an exemplary configuration of an information processing system according to the embodiment. FIG. 3 is a diagram illustrating an exemplary configuration of the information processing apparatus according to the embodiment. FIG. 4 is a diagram illustrating an example of a query information storage unit according to the embodiment. FIG. 5 is a diagram illustrating an example of a vector information storage unit according to the embodiment. FIG. 6 is a diagram illustrating an example of a classification definition storage unit according to the embodiment. FIG. 7 is a diagram illustrating an example of the category information storage unit according to the embodiment. FIG. 8 is a diagram illustrating an example of a model information storage unit according to the embodiment. FIG. 9 is a diagram illustrating an example of the first learning model according to the embodiment. FIG. 10 is a diagram illustrating an example of the second learning model according to the embodiment. FIG. 11 is a flowchart illustrating a first learning model generation processing procedure according to the embodiment. FIG. 12 is a flowchart illustrating an information processing procedure according to the embodiment. FIG. 13 is a hardware configuration diagram showing an example of a computer for realizing the function of the information processing apparatus.

  Hereinafter, a mode for carrying out an information processing apparatus, an information processing method, and an information processing program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the information processing apparatus, the information processing method, and the information processing program according to the present application are not limited by this embodiment. Moreover, the same code | symbol is attached | subjected to the same site | part in each following embodiment, and the overlapping description is abbreviate | omitted.

[1. Example of information processing]
First, an example of information processing according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of information processing according to the embodiment. The information processing illustrated in FIG. 1 is performed by the user terminal 10, the search server 50, and the information processing apparatus 100.

[Configuration of information processing system]
Prior to the description of FIG. 1, the configuration of the information processing system 1 will be described with reference to FIG. FIG. 2 is a diagram illustrating an exemplary configuration of an information processing system according to the embodiment. As illustrated in FIG. 2, the information processing system 1 includes a user terminal 10, a search server 50, and an information processing apparatus 100. The user terminal 10, the search server 50, and the information processing apparatus 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 may include an arbitrary number of user terminals 10, an arbitrary number of search servers 50, and an arbitrary number of information processing apparatuses 100.

  The user terminal 10 is an information processing apparatus used by a user who uses a search service. 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, the user may be read as the user terminal 10 below.

  Also, in the following, the user specified by the user ID “U1” may be referred to as “user U1”. Thus, in the following, when “user U * (* is any numerical value)” is described, it indicates that the user is the user specified by the user ID “U *”. For example, when "user U2" is described, the user is a user specified by user ID "U2".

  Moreover, below, the user terminal 10 is demonstrated as user terminal 10-1, 10-2 according to the user who uses the user terminal 10. For example, the user terminal 10-1 is the user terminal 10 used by the user U1. Also, for example, the user terminal 10-2 is a user terminal 10 used by the user U2. Also, in the following description, the user terminals 10-1 and 10-2 are described as the user terminal 10 when the description is made without particular distinction.

  The user terminal 10 transmits a search query input by the user to the search server 50. Specifically, the user terminal 10 acquires a search page including a search box for inputting a search query from the search server 50 in accordance with an operation by the user. Subsequently, when an operation of transmitting a search query is performed following an operation in which the user inputs characters in the search box, the user terminal 10 uses the characters input in the search box via the search page as the search query. It transmits to the search server 50. For example, if the user terminal 10 performs an operation of pressing a search query transmission button or an enter key after an operation of inputting characters into a search box by the user, the user terminal 10 performs a search via a search page. The characters entered in the search box are transmitted to the search server 50 as a search query.

  The search server 50 is a server device that provides a search service. Specifically, when receiving a search query from the user terminal 10, the search server 50 selects content corresponding to the received search query and output as a search result. Subsequently, the search server 50 delivers the search result page including the selected content to the user terminal 10. Here, the content delivered by the search server 50 is not limited to the web page displayed by the web browser. For example, the content distributed by the search server 50 may be content displayed by a dedicated application installed in the user terminal 10. The content distributed by the search server 50 includes music content, images (including moving images as well as still images), and text content (including news articles and articles posted on SNS (Social Networking Service)). Any content such as content combining an image and text, game content, and the like may be used.

  Further, the search server 50 stores information related to the search query input by the user. Specifically, the search server 50 stores information related to the user search history. For example, when receiving a search query from the user terminal 10, the search server 50 registers the received search query, the user ID for identifying 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 50 transmits information related to the search query input by the user to the information processing apparatus 100 in response to a request from the information processing apparatus 100.

  The information processing apparatus 100 is a server apparatus that generates a second learning model that predicts a category to which a predetermined search query belongs from a predetermined search query, using a first learning model described later. Here, the first learning model has a plurality of search queries input by the same user within a predetermined time, assuming that the plurality of search queries input by the same user within a predetermined time have similar characteristics. This is a learning model that learns the features of. Hereinafter, the first learning model is described as a first model (or first model M1) as appropriate. In the following description, the second learning model is described as a second model (or first model M2) as appropriate.

  Generally, when a searcher performs a search, the searcher intends as a result of performing a search multiple times using different search queries than in the case where the information intended by the searcher is reached by one search. There are more cases where information is reached. That is, since it is considered that the user performs a search a plurality of times with a certain intention, it is assumed that a search query continuously input within a predetermined time has a close search intention. Therefore, in the information processing apparatus 100 according to the present invention, a plurality of search queries input by a single user within a predetermined period of time have a similar feature, and a plurality of search queries input by the same user within a predetermined period of time A first learning model in which features of the search query are learned is generated. Specifically, the information processing apparatus 100 acquires, from the search server 50, information on the search query input by the user. Subsequently, the information processing apparatus 100 extracts, from among the search queries acquired from the search server 50, a plurality of search queries input by the same user within a predetermined time. Subsequently, the information processing apparatus 100 generates a first learning model that predicts feature information of a predetermined search query from a predetermined search query by learning that the plurality of extracted search queries have similar features. . For example, the information processing apparatus 100 learns the first learning model so that the distributed expressions of the plurality of extracted search queries are similar to each other, so that the distributed expression (vector) including the characteristic information of the predetermined search query from the predetermined search query ) Is generated.

  More specifically, the information processing apparatus 100 uses the technology of Deep Structured Sematic Model (DSSM) that uses Long Short-Term Memory (LSTM), which is a type of Recurrent Neural Network (RNN), for distributed expression generation. A first learning model that outputs a distributed expression (vector) from the search query is generated. For example, it is assumed that the information processing apparatus 100 has a feature that the pair of search queries input by the same user within a predetermined time have similar characteristics as the correct data of the first learning model (a distributed search query (FIG. (Vector) and a distributed expression (vector) of another search query paired with a predetermined search query are learned so as to be present near each other on the distributed expression space. Note that learning so that two vectors are close to each other in the distributed representation space can be rephrased as learning so that the two vectors are similar in the distributed representation space.

  From here, the flow of information processing will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of information processing according to the embodiment. In the example illustrated in the upper part of FIG. 1, the information processing apparatus 100 is a search query Q11 ("Roppongi pasta"), which is four search queries continuously input within the predetermined time by the same user U1. Q12 (“Roppongi Italian”), search query Q13 (“Akasaka pasta”), and search query Q14 (“Azabu pasta”) are extracted. The information processing apparatus 100 extracts a plurality of search queries in which the time interval when each search query is input by the same user U1 is within a predetermined time. Further, the information processing apparatus 100 extracts a plurality of search queries in which an interval of time in 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 each search query is input within a predetermined time by the user U1 in the order of search query Q11, search query Q12, search query Q13, and search query Q14. When the information processing apparatus 100 extracts four search queries, it is a pair of three pairs of search queries with two search queries adjacent in time series as a pair of search queries (search query Q11, search query Q12) , (Search query Q12, search query Q13), (search query Q13, search query Q14) are extracted. After extracting the three search query pairs, the information processing apparatus 100 inputs the extracted search query Q1k (k = 1, 2, 3, 4) to the first model M1 (step S11). The information processing apparatus 100 may extract a plurality of search queries in which all the search queries are input within a predetermined time by the same user U1. Then, the information processing apparatus 100 selects two search queries from the plurality of extracted search queries regardless of whether they are adjacent in time series, and selects the two selected search queries as a pair of search queries. It may be extracted as

  Subsequently, the information processing apparatus 100 uses the vector BQV1k (k = 1, 2, 3, 4), which is a distributed representation of the search query Q1k (k = 1, 2, 3, 4), as output data of the first model M1. It outputs (step S12). 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. The distributed expression before applying feedback to the first model M1 (before learning) is shown.

  Here, for example, the search query Q1k (k = 1, 2, 3, 4) continuously inputted within the predetermined time by the same user U1 is, for example, “a place (near Minato-ku, Tokyo) by the user U1 It is inferred that it is a set of search queries searched under the search intention “search for restaurant”. That is, the search query Q1k (k = 1, 2, 3, 4) is a search query searched under a search intention of “finding a restaurant in a certain place (near Tokyo Minato-ku)”. , It is presumed that the search queries have characteristics similar to each other. Therefore, the information processing apparatus 100 generates a first model that predicts feature information of a predetermined search query from a predetermined search query by learning that search queries sequentially input have similar characteristics. (Step S13). Specifically, the first information processing apparatus 100 predicts a distributed expression of a predetermined search query from a predetermined search query by learning that the distributed expressions of the search query input continuously are similar to each other. Generate M1. For example, in the information processing apparatus 100, 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 are similar in the distributed expression space. 1 Train the model M1. Further, the information processing apparatus 100 is configured such that the distributed representation of the search query Q12 (vector QV12) and the distributed representation of the search query Q13 paired with the search query Q12 (vector QV13) are similar in the distributed representation space. 1 Train the model M1. In addition, the information processing apparatus 100 causes the distributed representation of the search query Q13 (vector QV13) and the distributed representation of the search query Q14 paired with the search query Q13 (vector QV14) to be similar in the distributed representation space. One model M1 is learned.

  In the upper right side of FIG. 1, as an output result of the learned first model M1, the distribution of the search query Q1k (k = 1, 2, 3, 4) input within a predetermined time by the same user U1 A state in which a vector QV1k (k = 1, 2, 3, 4) as an expression is mapped as a cluster CL11 in a distributed expression space is shown. As described above, the information processing apparatus 100 generates the first learning model M1 in which the features possessed by the plurality of search queries input by the same user within a predetermined time have been learned.

  When the information processing apparatus 100 generates the first model M1, the information processing apparatus 100 acquires the generated first model M1 (model data MDT1 of the first model M1). When acquiring the first model M1, the information processing apparatus 100 generates a second learning model M2 using the acquired first model M1. Specifically, the information processing apparatus 100 generates a second model M2 having a connection coefficient different from that of the first model M1 as a weight of the learning model by relearning the first model M1. More specifically, the information processing apparatus 100 generates a second learning model M2 that predicts, from the predetermined search query, a category to which the predetermined search query belongs, using the first model M1 (step S14).

  In the example illustrated in the lower part of FIG. 1, when the search query is input to the second model M2, the information processing apparatus 100 CAT11 ("search for restaurant"), CAT12 ("search for goods"), CAT13 ( A second model M2 that predicts which of the four categories of “Reserve restaurant” and CAT14 (“Purchase a product”) is generated. Specifically, when a search query is input to the second model M2 as input information, the information processing apparatus 100 outputs a second model M2 that outputs, for each category, a probability that the search query belongs to the category as output information. Generate For example, the information processing apparatus 100 learns a set of a search query and a category (any one of CAT11 to CAT14) to which the search query belongs as correct data of the second model M2.

  The fact that the search query belongs to CAT11 (“Find a restaurant”) indicates that the search query is a search query input with the intention of searching for a restaurant. Also, belonging to CAT12 ("search for a product") indicates that the search query is a search query that is input with the intention of searching for a product. Further, the fact that the search query belongs to CAT 13 (“Reserve restaurant”) indicates that the search query is a search query input with the intention of reserving a restaurant. Also, the fact that the search query belongs to the CAT 14 (“purchase the item”) indicates that the search query is a search query entered with the intention of purchasing the item.

  Specifically, when the search query is input to the learning model, the information processing apparatus 100 performs learning so that the classification result of the distributed expression output by the learning model corresponds to the category to which the search query belongs. A second model M2 that predicts a category to which the predetermined search query belongs is generated from the predetermined search query. Then, for example, when a search query is input to the second model M2 as input information, the information processing apparatus 100 outputs, as output information, a probability that the search query belongs to the category for each category CAT11 to CAT14. Generate M2.

  For example, when the search query Q11 ("Roppongi pasta") is input to the second model M2 as input information (step S15), the information processing apparatus 100 disperses the search query Q11 ("Roppongi pasta") as output information. A vector BQV11 which is an expression is output. Here, the vector BQV11 is a distributed representation of the search query Q11 that has just been output from the output layer of the second model M2, and indicates the distributed representation before applying feedback to the second model M2 (before learning). Here, it is assumed that the correct answer category to which the search query Q11 ("Roppongi pasta") belongs is CAT11 ("search for a restaurant"). In this case, in the information processing apparatus 100, the probability that the vector BQV11, which is a distributed expression of the output search query Q11 ("Roppongi pasta"), is classified into CAT 11 ("search for a restaurant") exceeds a predetermined threshold. To learn the second model M2. Note that the information processing apparatus 100 learns the second model using correct answer data prepared in advance. The information processing apparatus 100 may generate correct answer data of the second model M2. The information processing apparatus 100 may learn the second model M2 using the generated correct answer data. Specifically, the information processing apparatus 100 determines the correct answer category to which the search query belongs, based on the post-search behavior of the user who has searched for the search query. More specifically, the information processing apparatus 100 corresponds to a correct answer category for a user who has searched for a predetermined search query, a predetermined action in which the percentage of users who have performed a predetermined action after the search exceeds a predetermined threshold. Decide as the action to be taken. For example, the percentage of users who have searched for restaurants is 90%, and the percentage of users who have searched for restaurants as a percentage of users who have searched for a restaurant after searching for a search query Q11 (“Roppongi pasta”). It is assumed that the proportion of users who have caused is 0%, the proportion of users who have made an action to reserve a restaurant after a search is 10%, and the proportion of those who have made an action to purchase a product after a search is 0%. In this case, since the information processing apparatus 100 exceeds the predetermined threshold (for example, 90%) with the percentage of users who have made an action to search for a restaurant, the information processing apparatus 100 performs the search query Q11 ("Roppongi pasta") The action corresponding to the correct category is determined. Then, since the information processing apparatus 100 determines that the action corresponding to the correct answer category is an action for searching for a restaurant, the correct answer category to which the search query Q11 ("Roppongi pasta") belongs is CAT11 ("search for a restaurant") Decide on.

  For example, when the search query Q11 ("Roppongi pasta") is input to the second model M2 before learning, the information processing apparatus 100 classifies the vector BQV11, which is a distributed expression, into CAT11 ("search for a restaurant"). The probability of being classified is 80%, the probability of being classified as CAT 12 ("Find a product") is 0%, the probability of being classified as CAT 13 ("Reserve a restaurant") is 20%, CAT 14 ("Buy a product") ) Is output as 0%. In this case, the information processing apparatus 100 learns the second model M2 such that the probability that the vector BQV11, which is a distributed expression, is classified into the CAT 11 ("search for a restaurant") exceeds a predetermined threshold (for example, 90%) Let In addition, the information processing apparatus 100 learns that the probability that the vector BQV11, which is a distributed expression, is classified into the CAT 11 (“search for a restaurant”) exceeds a predetermined threshold (for example, 90%), The second model M2 is trained so as to reduce the probability that the vector BQV11, which is a distributed expression, is classified into the other category CAT13 ("Reserve restaurant") to 10%.

  As described above, when the predetermined search query is input as the input information, the information processing apparatus 100 causes the probability that the distributed expression of the predetermined search query as the output information is classified into the correct answer category exceeds the predetermined threshold. 2 Train the model. Then, when a predetermined search query is input as input information, the information processing apparatus 100 selects a category in which the probability that the distributed expression of the predetermined search query belongs to the category exceeds a predetermined threshold, Output as For example, when the search query Q11 ("Roppongi pasta") is input as input information to the learned second model M2, the information processing apparatus 100 generates a vector BQV11 which is a distributed expression of the search query Q11 ("Roppongi pasta"). Since the probability of belonging to the category CAT11 ("search for a restaurant") exceeds 90%, the category to which the search query belongs is output as CAT11 ("search for a restaurant") as output information (step S16). Thus, the information processing apparatus 100 generates a second model that predicts a category of a predetermined search query from a predetermined search query by learning a set of a search query and a correct category of the search query (step S17). ).

  In general, since a user is considered to search several times with a certain intention, it is assumed that search queries sequentially input within a predetermined time have a similar search intention. Therefore, the information processing apparatus 100 according to the present invention is similar to each other in that a plurality of search queries input continuously within a predetermined time are search queries searched under a predetermined search intention. The first model M1 is learned by assuming that the search query has the characteristic of Thereby, the information processing apparatus 100 can cause the first model M1 to learn the characteristics of the search query considering the search intention. Then, the information processing apparatus 100 uses the first model M1 that has learned the characteristics of the search query in consideration of the search intention, and efficiently uses the second model that predicts the category of the predetermined search query from the predetermined search query. Can be generated. As a result, the information processing apparatus 100 can classify the search query into a category in consideration of the search intention of the user who has input the search query. 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 various and have long tail properties, it is very time consuming and difficult to label correct answer categories corresponding to a large number of search queries. Here, the information processing apparatus 100 learns the second model from the first model that has learned the characteristics of the search query in consideration of the search intention, so that the user's search can be performed instead of labeling the correct category. The second model which predicts the category of the search query can be learned with the intention (the context of the user who has input the search query) as a sort of correct answer. Thereby, the information processing apparatus 100 can learn the second model without manually labeling the correct answer category of the search query. That is, the information processing apparatus 100 can obtain sufficient classification accuracy even when there are few correct answer data. Further, the information processing apparatus 100 can obtain higher classification accuracy when there is a lot of correct answer data. Therefore, the information processing apparatus 100 can improve the classification accuracy of the search query.

[2. Configuration of information processing apparatus]
Next, the configuration of the information processing apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram illustrating an exemplary configuration of the information processing apparatus 100 according to the embodiment. As shown in FIG. 3, the information processing apparatus 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 types of information. You may have.

(Communication unit 110)
The communication unit 110 is realized by, for example, a network interface card (NIC). The communication unit 110 is connected to the network by wire or wirelessly, and transmits and receives information between the user terminal 10 and the search server 50, for example.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. As illustrated in FIG. 3, the storage unit 120 includes a query information storage unit 121, a vector information storage unit 122, a classification definition storage unit 123, a category information storage unit 124, and a model information storage unit 125.

(Query information storage unit 121)
The query information storage unit 121 stores various types of information related to the search query input by the user. FIG. 4 shows an example of the query information storage unit according to the embodiment. In the example illustrated in FIG. 4, the query information storage unit 121 has items such as “user ID”, “date and time”, “search query”, and “search query ID”.

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

  In the example shown in the first record of FIG. 4, 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 has input the search query Q11 is the user (user U1) identified by the user ID “U1”. The date and time “2018/9/1 PM 17:00” indicates that the date and time when the search server accepted the search query Q11 from the user U1 is 17:00 pm on September 1, 2018. The search query “Roppongi pasta” indicates a search query Q11 input by the user U1. Specifically, the search query “Roppongi pasta” indicates that the character string “Roppongi” indicating the place name and the character “pasta” indicating the type of food are separated by a space as a delimiter.

(Vector information storage unit 122)
The vector information storage unit 122 stores various types of information regarding vectors that are distributed representations of search queries. FIG. 5 shows an example of the vector information storage unit according to the embodiment. In the example illustrated in FIG. 5, the vector information storage unit 122 includes 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. “Search query ID” indicates identification information for identifying a search query corresponding to a vector. “Vector information” indicates an N-dimensional vector which is a distributed expression of the search query. The vector that is a 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 vector (vector QV11) identified by the vector ID “QV11” corresponds to the vector QV11 that is a distributed representation 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. Also, vector information “QVDT11” indicates an N-dimensional vector that is a distributed expression of the search query Q11.

(Classification definition storage unit 123)
The classification definition storage unit 123 stores various types of information related to the definition of the category into which the search query is classified. FIG. 6 shows an example of the classification definition storage unit according to the embodiment. In the example shown in FIG. 6, the classification definition storage unit 123 has items such as “large classification ID”, “large classification”, “small classification ID”, and “small classification”.

  “Major classification” indicates the major classification of the category 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. 6, the large classification “purchasing behavior system” corresponds to the large classification described in the example illustrated in the lower part of FIG. 1. The large classification “purchase behavior system” indicates a large classification of categories for classifying the search query based on the purchase behavior of the user. In the example illustrated in FIG. 6, the major classification “purchasing behavior system” further includes four minor classifications. The large category ID “CAT1” indicates identification information for identifying the large category “purchasing behavior system”.

  “Small category” indicates a minor category of a category into which the search query is classified. “Small category ID” indicates identification information for identifying a minor category. In the example shown in FIG. 6, the minor category "find restaurant" is a category belonging to the major category "purchasing behavior system", and a search query classified as minor category is input with the intention of the user to search for a restaurant Indicates that the search query was made. The small category ID "CAT11" indicates identification information for identifying the small category "finding restaurant".

  The subcategory “Find Product” is a category belonging to the major category “Purchasing Behavior”, and indicates that the search query that is classified into the subcategory is a search query that is entered by the user with the intention of searching for a product. . The small classification ID “CAT12” indicates identification information for identifying the small classification “search for a product”.

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

  The subcategory “Purchase product” is a category belonging to the main category “Purchasing behavior”, and the search query classified into the subcategory is a search query entered 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 124)
The category information storage unit 124 stores various types of information related to the category to which the search query belongs. Specifically, the category information storage unit 124 stores various types of information related to the category output by the second learning model when the search query is input to the learned second learning model. FIG. 7 shows an example of the category information storage unit according to the embodiment. In the example illustrated in FIG. 7, the category information storage unit 124 has items such as “search query ID”, “large category ID”, “small category ID”, and “probability (%)”.

  “Search query ID” indicates identification information for identifying a search query input by a user. In the example shown in FIG. 7, 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. “Small category ID” indicates identification information for identifying a minor category. “Probability (%)” indicates the probability for each minor category output by the second learning model when the search query is input to the learned second learning model. In the example shown in FIG. 7, the probability (%) "90" indicates that the probability that the search query Q11 is classified into the category CAT11 is 90%.

(Model information storage unit 125)
The model information storage unit 125 stores various types of information regarding the learning model generated by the information processing apparatus 100. FIG. 8 shows an example of a model information storage unit according to the embodiment. In the example illustrated in FIG. 8, the model information storage unit 125 includes items such as “model ID” and “model data”.

  “Model ID” indicates identification information for identifying a learning model generated by the information processing apparatus 100. “Model data” indicates model data of a learning model generated by the information processing apparatus 100. For example, “model data” stores data for converting a search query into a distributed expression.

  In the example shown in the first record in FIG. 8, the learning model identified by the model ID “M1” corresponds to the first model M1 shown in FIG. The model data “MDT1” indicates model data (model data MDT1) of the first model M1 generated by the information processing apparatus 100.

  The model data MDT1 is an input layer to which a search query is input, an output layer, and any layer from the input layer to the output layer, and is a first element belonging to a layer other than the output layer, a first element and a first A second element whose value is calculated based on the weight of one element, and outputs a distributed representation of the search query input to the input layer from the output layer in response to the search query input to the input layer As such, the information processing apparatus 100 may function.

  Here, it is assumed that the model data MDT1 is realized by a 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 input data (xi) such as x1 or 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 correspond to any node of the input layer, and the second element can be regarded as a node of 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 of the input layer or the middle layer. Also, the second element corresponds to the next-stage node that is a node to which a value is transmitted from the node corresponding to the first element. Also, 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 information processing apparatus 100 calculates a distributed expression using a model having an arbitrary structure such as the above-described regression model or neural network. Specifically, in the model data MDT1, a coefficient is set so that a distributed expression is output when a search query is input. The information processing apparatus 100 calculates the distributed representation using such model data MDT1.

  In the above example, the model data MDT1 is an example in which the distributed expression of the search query is output (hereinafter referred to as a model X1) when the search query is input. However, model data MDT1 which concerns on embodiment may be a model produced | generated based on the result obtained by repeating input and output of data to model X1. For example, the model data MDT1 may be a model (hereinafter, referred to as a model Y1) that has been learned so that a search query is input and a distributed representation output by the model X1 is output. Alternatively, the model data MDT1 may be a model learned by using a search query as an input and outputting the output value of the model Y1.

  When the information processing apparatus 100 performs an estimation process using GAN (Generative Adversarial Networks), the model data MDT1 may be a model that constitutes a part of the GAN.

  In the example shown in the second record of FIG. 8, the learning model identified by the model ID “M2” corresponds to the second model M2 shown in FIG. 1. The model data “MDT2” indicates model data (model data MDT2) of the second model M2 generated by the information processing apparatus 100.

  The model data MDT2 is an input layer to which a search query is input, an output layer, and any layer from the input layer to the output layer, which is a first element belonging to a layer other than the output layer, a first element and a first A second element whose value is calculated based on the weight of one element, and in accordance with the search query input to the input layer, the probability that the search query input to the input layer belongs to each category is output to the output layer The information processing apparatus 100 may be caused to function so as to be output from.

  Here, it is assumed that the model data MDT2 is realized by a 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 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 correspond to any node of the input layer, and the second element can be regarded as a node of the output layer.

  Further, it is 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 that the input layer or the middle layer has. Also, the second element corresponds to the next-stage node that is a node to which a value is transmitted from the node corresponding to the first element. Also, 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 information processing apparatus 100 calculates the probability that the search query belongs to each category, using a model having an arbitrary structure, such as the above-described regression model or neural network. Specifically, in the model data MDT2, coefficients are set so that the probability that the search query belongs to each category is output when the search query is input. The information processing apparatus 100 calculates the probability that the search query belongs to each category, using such model data MDT2.

  In the above example, the model data MDT2 is an example of a model (hereinafter referred to as model X2) that outputs the probability that the search query belongs to each category when the search query is input. However, model data MDT2 which concerns on embodiment may be a model produced | generated based on the result obtained by repeating input and output of data to model X2. For example, the model data MDT2 may be a model (hereinafter referred to as a model Y2) that has been learned so as to have a search query as an input and a probability that the model X2 outputs as an output. Alternatively, the model data MDT2 may be a model learned by using a search query as an input and outputting the output value of the model Y2.

  When the information processing apparatus 100 performs an estimation process using GAN (Generative Adversarial Networks), the model data MDT2 may be a model that constitutes a part of the GAN.

(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 apparatus 100 by, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). Various programs (corresponding to an example of a generation program) are realized by executing the RAM as a work area. The control unit 130 is a controller, and is realized by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  In addition, the control unit 130 performs a search query input to the input layer on each layer other than the output layer by information processing according to the first model M1 (model data MDT1) stored in the model information storage unit 125. The computer is functioned to output the distributed representation from the output layer by performing an operation based on the first element and the weight of the first element with each element to which it belongs as the first element.

  In addition, the control unit 130 performs a search query input to the input layer on each layer other than the output layer by information processing according to the second model M2 (model data MDT2) stored in the model information storage unit 125. The computer is functioned to output the probability that the search query belongs to each category from the output layer by performing an operation based on the first element and the weight of the first element with each element to which it belongs as the first element.

  As shown in FIG. 3, the control unit 130 includes an acquisition unit 131, an extraction unit 132, and a generation unit 133, and implements or executes the action of the information processing described below. Note that the internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 3, and may be another configuration as long as the information processing described below is performed.

(Acquisition part 131)
The acquisition unit 131 acquires various information. Specifically, the acquisition unit 131 acquires a search query input by the user from the search server 50. When acquiring the search query input by the user, the acquiring unit 131 stores the acquired search query in the query information storage unit 121. In addition, the acquisition unit 131 acquires vector information related to a vector that is a distributed expression of a search query. When acquiring the vector information, the acquiring unit 131 stores the acquired vector information in the vector information storage unit 122. In addition, the acquisition unit 131 acquires information that defines the search query and the category of the category to which the search query belongs. When acquiring the classification definition information defining the search query and the classification of the category to which the search query belongs, the acquisition unit 131 stores the acquired classification definition information in the classification definition storage unit 123. Further, the acquisition unit 131 acquires category information related to the category to which the search query belongs. When acquiring the category information, the acquiring unit 131 stores the acquired category information in the category information storage unit 124.

  In addition, the acquisition unit 131 acquires a first learning model in which a plurality of search queries input within a predetermined time by the same user have similar characteristics and learns the characteristics of the plurality of search queries. Specifically, the acquisition unit 131 learns that a plurality of search queries input by the same user within a predetermined time by the generation unit 133 have similar characteristics, and thus acquires the predetermined search query from the predetermined search query. The first learning model generated as a learning model for predicting the feature information of the search query is acquired. In addition, when a predetermined search query is input as input information, the acquisition unit 131 acquires a first learning model that outputs a distributed expression of the predetermined search query as output information. In addition, the acquisition unit 131 acquires a first learning model in which features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries input in succession within a predetermined time are similar. Do. In addition, the acquiring unit 131 learns that a plurality of search queries including character strings separated by predetermined delimiters have similar characteristics as a plurality of search queries input within a predetermined time by the same user. By doing this, the first learning model is obtained which has learned the features of the plurality of search queries. Further, the acquiring unit 131 acquires the first learning model in which the features included in the plurality of search queries are learned, by learning that the plurality of randomly extracted search queries have different features. In addition, the acquiring unit 131 acquires a first learning model in which features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries extracted at random are different. When acquiring the first learning model, the acquiring unit 131 stores model data MDT1 of the first learning model in the model information storage unit 125 in association with identification information that identifies the acquired first learning model.

  In addition, the acquisition unit 131 acquires a second learning model generated by the generation unit 133 as a learning model that predicts a category to which a predetermined search query belongs from a predetermined search query using the first learning model. When acquiring the second learning model, the acquiring unit 131 stores model data MDT2 of the second learning model in the model information storage unit 125 in association with identification information identifying the acquired second learning model.

(Extractor 132)
The extraction unit 132 extracts various information. Specifically, the extraction unit 132 extracts, from the search queries acquired by the acquisition unit 131, a plurality of search queries input by the same user within a predetermined time. For example, the extraction unit 132 extracts a plurality of search queries in which an interval of time in which each search query is input by the same user is within a predetermined time. Subsequently, the extraction unit 132 extracts a pair of search queries continuously input within a predetermined time by the same user among a plurality of search queries input within the predetermined time by the same user. For example, the extraction unit 132 extracts a plurality of search queries in which an interval of time in which each search query pair is input by the same user is within a predetermined time. For example, the extraction unit 132 includes, among the search queries acquired by the acquisition unit 131, a search query Q11 (“Roppongi Pasta”) that is four search queries that are continuously input by the same user U1 within a predetermined time. ), Search query Q12 ("Roppongi Italian"), search query Q13 ("Akasaka pasta"), search query Q14 ("Azabu pasta") are extracted. When arranged in the order in which the search queries are input, the extraction unit 132 extracts 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 extraction unit 132 extracts four search queries, it is a pair of three pairs of search queries with two search queries adjacent in time series as a pair of search queries (search query Q11, search query Q12), (search query Q12, search query Q13), and (search query Q13, search query Q14) are extracted. Note that the extraction unit 132 may extract a plurality of search queries in which all search queries are input by the same user within a predetermined time. Then, the extraction unit 132 selects two search queries from the plurality of extracted search queries regardless of whether they are temporally adjacent to each other, and sets the selected two search queries as a pair of search queries. You may extract.

  The extraction unit 132 also extracts a predetermined search query and other search queries unrelated to the predetermined search query from the search queries acquired by the acquisition unit 131. For example, the extraction unit 132 extracts a predetermined search query from among the search queries acquired by the acquisition unit 131. Subsequently, the extraction unit 132 randomly extracts another search query from the search queries acquired by the acquisition unit 131 regardless of a predetermined search query.

(Generator 133)
The generation unit 133 generates various information. The generation unit 133 generates a first learning model. Specifically, the generation unit 133 assumes that the plurality of search queries input by the same user within a predetermined time have similar characteristics, and the plurality of searches input by the same user within the predetermined time. A first learning model is generated in which the features of the query are learned. More specifically, the generation unit 133 learns that among the search queries acquired by the acquisition unit 131, a plurality of search queries input by the same user within a predetermined time have similar characteristics. The first learning model that predicts the feature information of the predetermined search query from the predetermined search query is generated.

  The generation unit 133 learns the learning model so that distributed expressions of a plurality of search queries input within a predetermined time by the same user are similar, thereby obtaining feature information of the predetermined search query from the predetermined search query. Generate a first model to predict. Specifically, the generation unit 133 generates a first learning model by learning so that the distributed expressions of a pair of search queries input in succession within a predetermined time are similar. More specifically, the generation unit 133 generates a first learning model that outputs a distributed representation (vector) from the search query, using a DSSM technique using LSTM, which is a type of RNN, for distributed representation generation. For example, the generation unit 133 assumes that the pair of search queries input by the same user within a predetermined time as the correct answer data of the first learning model has similar characteristics, and a distributed expression (vector) of the predetermined search query ) And distributed expressions (vectors) of other search queries that are paired with a predetermined search query are learned so that they exist close together in the distributed expression space.

[Example of first learning model]
Here, an example of the first learning model generated by the information processing apparatus 100 will be described with reference to FIG. FIG. 9 is a diagram illustrating an example of the first learning model according to the embodiment. In the example illustrated in FIG. 9, the first learning model M1 generated by the information processing apparatus 100 is configured by three layers of LSTM RNN. In the example shown in FIG. 9, the extraction unit 132 is a pair of a search query Q11 "Roppongi Pasta" and a search query Q12 "Roppongi Italian", which are continuously input within a predetermined time by the same user U1. Extract the search query. The generation unit 133 inputs the search query Q11 extracted by the extraction unit 132 into the input layer of the first learning model M1 (step S31).

  Subsequently, the generation unit 133 outputs a 256-dimensional vector BQV11, which is a distributed representation of the search query Q11, from the output layer of the first learning model M1. Further, the generation unit 133 inputs the search query Q12 extracted by the extraction unit 132 to the input layer of the first learning model M1. Subsequently, the generation unit 133 outputs a 256-dimensional vector BQV12 that is a distributed representation of the search query Q12 from the output layer of the first learning model M1 (step S32).

  Subsequently, the generation unit 133 learns so that the distributed expressions (vectors) of two search queries input in succession are similar to each other, whereby the first learning model M1 that outputs the distributed expressions (vectors) from the search query. Are generated (step S33). For example, let 大 き be a size of an angle formed by the vector BQV11 which is a dispersed expression of the search query Q11 before feedback (before learning) to the first learning model M1 and the vector BQV12 which is a distributed expression of the search query Q12. Also, let Φ be the magnitude of the angle between vector QV11, which is a distributed representation of search query Q11 after feedback to first learning model M1 (after learning), and vector QV12, which is a distributed representation of search query Q12. . At this time, the generation unit 133 causes the first learning model M1 to be learned so that Φ is smaller than Θ. That is, the generation unit 133 sets the first learning model such that the value of cosine similarity between vector QV11 and vector QV12 is larger than the value of cosine similarity between vector BQV11 and vector BQV12 (so that the value approaches 1). Make M1 learn. As described above, the generation unit 133 distributes from the search query by causing the first learning model M1 to learn so that two vectors that are a pair of distributed expressions corresponding to the pair of search queries are similar in the distributed expression space. A first learning model M1 that outputs an expression (vector) is generated.

  In addition, the generation unit 133 learns that a plurality of search queries including character strings separated by predetermined delimiters have similar characteristics as a plurality of search queries input within a predetermined time by the same user. By doing so, a first learning model is generated. For example, the generation unit 133 may use a search query "Roppongi pasta" divided by a space where the characters "Roppongi" indicating a place name and "pasta" indicating a food type are delimiters, and "Roppongi" indicating a place name. The first learning model is generated by learning that the search query "Roppongi Italian" separated by a space, which is the character of "Italian" indicating the type of food, is a similar character.

  Moreover, the production | generation part 133 produces | generates a 1st learning model by learning as what has the characteristic from which the some search query extracted at random among the search queries acquired by the acquisition part 131 differs. Specifically, the generation unit 133 generates the first learning model by learning that the distributed expressions of the pair of search queries extracted at random among the search queries acquired by the acquisition unit 131 are different. Do. For example, the generating unit 133 may be configured such that the distributed expression of the predetermined search query extracted by the extracting unit 132 and the distributed expression of the search query extracted at random regardless of the predetermined search query are far apart in the distributed expression space The first learning model M1 is trained so as to be mapped.

  In addition, the generation unit 133 generates a second learning model. The generation unit 133 uses the first learning model acquired by the acquisition unit 131 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 unit 133 acquires the generated first learning model (the model data MDT1 of the first learning model M1). When the generation unit 133 acquires the first model M1, the generation unit 133 generates a second learning model M2 using the acquired first model M1. The generation unit 133 re-learns the first model M1, thereby generating a 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 133 performs learning so that the classification result of the distributed expression output from the learning model corresponds to the category to which the search query belongs. A second model M2 is generated which predicts a category to which a predetermined search query belongs from the search query of.

  Specifically, when the search query is input to the learning model, the generation unit 133 performs learning so that the classification result of the distributed expression output from the learning model corresponds to the category to which the search query belongs. Generating a second learning model that predicts a category to which a predetermined search query belongs from the search query of When the search query is input to the learning model as input information, the generation unit 133 generates a second learning model that outputs a probability for each category to which the search query belongs as output information. For example, the generation unit 133 uses the first model M1 to calculate the probability that the distributed expression of the search query is classified into the category as output information when a predetermined search query is input as input information to the learning model. A second model M2 that is output every time is generated. When a predetermined search query is input as input information, the generation unit 133 causes the second model to be learned such that the probability that the distributed expression of the predetermined search query is classified into the correct category as output information exceeds a predetermined threshold. . Then, when a predetermined search query is input as input information, the generation unit 133 sets, as a category of the predetermined search query, a category in which the probability that the distributed expression of the predetermined search query belongs to the category exceeds a predetermined threshold. A second model M2 to be output is generated.

  For example, the generation unit 133 refers to the model information storage unit 125 illustrated in FIG. 8 to acquire a first model M1 (model data MDT1 of the first model M1). Subsequently, the generation unit 133 refers to the classification definition storage unit 123 illustrated in FIG. 6 and selects a large classification of categories for classifying the search query. Subsequently, when the large classification is selected, the generation unit 133 learns a set of a search query and a small classification to which the search query belongs as learning data of the second model M2.

  For example, it is assumed that the correct answer category to which the search query Q11 ("Roppongi pasta") belongs is CAT 11 ("search for a restaurant"). When the search query Q11 ("Roppongi pasta") is input to the second model M2 as input information, the generation unit 133 uses the distributed representation of the search query Q11 ("Roppongi pasta") from the output layer of the second model M2 A certain vector BQV11 is output. Here, the vector BQV11 is a distributed representation of the search query Q11 that has just been output from the output layer of the second model M2, and indicates the distributed representation before applying feedback to the second model M2 (before learning). In this case, the generation unit 133 determines that the probability that the vector BQV11, which is a distributed expression of the output search query Q11 ("Roppongi pasta"), is classified into the correct category CAT11 ("search for a restaurant") exceeds a predetermined threshold. Thus, the second model M2 is learned.

  For example, when the search query Q11 ("Roppongi pasta") is input to the second model M2 before learning, the generation unit 133 classifies the vector BQV11, which is a distributed expression, into CAT 11 ("search for a restaurant"). The probability of being classified as 80%, the probability of being classified as CAT 12 ("Find a product") as 0%, the probability of being classified as CAT 13 ("Reserve a Restaurant") as 20%, CAT 14 ("Buy a Product") It is assumed that the probability of being classified as 0 is output as 0%. In this case, the generation unit 133 causes the second model M2 to be learned such that the probability that the vector BQV11, which is a distributed expression, is classified into the CAT 11 (“search for a restaurant”) exceeds a predetermined threshold (for example, 90%). . In addition, the generation unit 133 performs dispersion in accordance with learning so that the probability that the vector BQV11, which is a distributed expression, is classified into the CAT 11 (“search for a restaurant”) exceeds a predetermined threshold (for example, 90%). The second model M2 is trained so that the probability that the vector BQV11 as an expression is classified into another category CAT13 (“Reserve restaurant”) is reduced to 10%. Subsequently, when search query Q11 ("Roppongi pasta") is input as input information to second model M2 that has been learned, generation unit 133 generates vector BQV11 that is a distributed expression of search query Q11 ("Roppongi pasta"). Since the probability of belonging to the category CAT11 ("search for a restaurant") exceeds 90%, the category to which the search query belongs is output as CAT11 ("search for a restaurant") as output information.

  The generation unit 133 may select an arbitrary number of large classifications as the large classifications. Then, when a search query is input to the second model M2 as input information, the generation unit 133 sets the probability of belonging to each small category belonging to an arbitrary number of large categories selected by the search query as output information for each small category. The second model M2 to be output to may be generated. In addition, the generation unit 133 may select all of the large classifications as the large classifications. And the production | generation part 133 may produce | generate the 2nd model M2 which outputs the probability which belongs to each small classification for every small classification, when a search query is input into the 2nd model M2.

[Example of second learning model]
Here, an example of the second learning model generated by the information processing apparatus 100 will be described with reference to FIG. FIG. 10 is a diagram illustrating an example of the second learning model according to the embodiment. In the example illustrated in FIG. 10, the second learning model M2 generated by the information processing apparatus 100 is generated using the first learning model M1. That is, the information processing apparatus 100 generates the second learning model M2 having a different connection coefficient, which is a weight of the learning model, from the first learning model M1 by causing the first learning model M1 to re-learn.

  More specifically, the second learning model M2 generated by the information processing apparatus 100 is composed of three layers of LSTM RNNs, like the first learning model M1. In the example illustrated in FIG. 10, the extraction unit 132 inputs the search query Q11 “Roppongi pasta” input by the user U1 to the input layer of the second learning model M2 (step S41).

  Subsequently, the generation unit 133 outputs a 256-dimensional vector BQV11 that is a distributed representation of the search query Q11 from the output layer of the second learning model M2 (step S42).

  Subsequently, the generation unit 133 outputs the probability that the vector BQV11, which is a distributed expression of the search query Q11, is classified into each category (step S43).

  Subsequently, the generation unit 133 learns the second learning model M2 so as to increase the probability that the vector BQV11, which is a distributed expression of the search query Q11, is classified into the correct answer category, and thereby the search query category from the search query. A second model to be predicted is generated (step S44).

[3. Flow of Generation Process of First Learning Model]
Next, the procedure of the first learning model generation process according to the embodiment will be described with reference to FIG. FIG. 11 is a flowchart illustrating a first learning model generation processing procedure according to the embodiment. In the example illustrated in FIG. 11, the information processing apparatus 100 acquires a search query input by a user (Step S <b> 101).

  Subsequently, the information processing apparatus 100 extracts a plurality of search queries input by the same user within a predetermined time (step S102).

  Subsequently, the information processing apparatus 100 generates a first learning model that predicts feature information of a predetermined search query from the predetermined search query by learning that the extracted search queries have similar characteristics. (Step S103).

[4. Flow of information processing]
Next, the procedure of the information processing according to the embodiment will be described with reference to FIG. FIG. 12 is a flowchart showing an information processing procedure according to the embodiment. In the example shown in FIG. 12, the information processing apparatus 100 acquires a first learning model (model data MDT1 of the first learning model M1) (step S201).

  Subsequently, the information processing apparatus 100 uses the first learning model to generate a second learning model that predicts a category of the predetermined search query from the predetermined search query (step S202).

[5. effect〕
As described above, the information processing apparatus 100 according to the embodiment includes the acquisition unit 131 and the generation unit 133. The acquiring unit 131 acquires a first learning model in which the features included in 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. The generation unit 133 uses the first learning model acquired by the acquisition unit 131 to generate a second learning model for predicting the category to which the predetermined search query belongs from the predetermined search query.

  In general, since a user is considered to search several times with a certain intention, it is assumed that search queries sequentially input within a predetermined time have a similar search intention. Therefore, the information processing apparatus 100 according to the present invention is similar to each other in that a plurality of search queries continuously input within a predetermined time are search queries searched under a predetermined search intention. The first model M1 is trained by regarding it as a search query having the following features. Thereby, the information processing apparatus 100 can cause the first model M1 to learn the characteristics of the search query considering the search intention. Then, the information processing apparatus 100 efficiently uses the first model M1 learned the features of the search query in consideration of the search intention to efficiently estimate the second model for predicting the category of the predetermined search query from the predetermined search query. Can be generated. As a result, the information processing apparatus 100 can classify the search query into a category in consideration of the search intention of the user who has input the search query. 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 various and have long-tailed nature, it is very time-consuming and difficult to label the correct answer categories corresponding to a large number of search queries. Here, the information processing apparatus 100 searches for the user instead of labeling the correct answer category by learning the second model by using the first model learned the features of the search query considering the search intention as the departure point. The second model for predicting the category of the search query can be learned using the intention (the context of the user who has input the search query) as a kind of correct answer. Thereby, the information processing apparatus 100 can learn the second model without manually labeling the correct answer category of the search query. That is, even when the correct data is small, the information processing apparatus 100 can obtain sufficient classification accuracy. In addition, the information processing apparatus 100 can obtain higher classification accuracy if the number of correct data is large. Therefore, the information processing apparatus 100 can improve the classification accuracy of the search query.

  In addition, when a predetermined search query is input as input information, the acquisition unit 131 acquires a first learning model that outputs a distributed expression of the predetermined search query as output information.

  As a result, the information processing apparatus 100 can measure the abstract concept of the characteristics of a predetermined search query using specific numerical values called distributed expressions.

  In addition, the acquisition unit 131 acquires a first learning model in which features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries input in succession within a predetermined time are similar. Do.

  In general, two search queries input by the same user in a short time are considered to have the same search intention or close search intention even if they are not the same. That is, it is considered that the search intentions of a pair of search queries that are continuously input within a predetermined time period have the same search intention or the search intentions that are not the same. That is, the information processing apparatus 100 can improve the learning accuracy of the first model by learning so that the distributed expressions of a pair of search queries subsequently inputted within a predetermined time are similar.

  In addition, the acquiring unit 131 learns that a plurality of search queries including character strings separated by predetermined delimiters have similar characteristics as a plurality of search queries input within a predetermined time by the same user. By doing this, the first learning model is obtained which has learned the features of the plurality of search queries.

  In general, a search query including a plurality of strings is considered to have a clearer search intention than a search query consisting of a single string. That is, the information processing apparatus 100 can improve the learning accuracy of the first model by learning using a search query including a character string divided by a predetermined delimiter.

  Further, the acquiring unit 131 acquires the first learning model in which the features included in the plurality of search queries are learned, by learning that the plurality of randomly extracted search queries have different features. In addition, the acquiring unit 131 acquires a first learning model in which features of a plurality of search queries are learned by learning so that the distributed expressions of a pair of search queries extracted at random are different.

  Generally, a plurality of search queries extracted at random are search queries that are input independently of each other, and therefore, the search intention is different or the search intention is considered to be far away. Therefore, the information processing apparatus 100 according to the present invention is a search query having characteristics that are different from each other in that the plurality of search queries extracted at random are search queries that are searched under different search intentions. The learning model M1 is learned by assuming that there is. Thus, the learning model can learn incorrect answer data, which is a pair of search queries far from the search intention, in addition to correct data that is a pair of search queries close to the search intention. That is, the information processing apparatus 100 can improve the learning accuracy of the first model.

  In addition, when the search query is input to the learning model, the generation unit 133 learns the classification result of the distributed expression output from the learning model so as to correspond to the category to which the search query belongs, so that the predetermined search query A second learning model for predicting a category to which a predetermined search query belongs is generated. In addition, when a search query is input to the learning model as input information, the generation unit 133 generates a second learning model that outputs a probability for each category to which the search query belongs as output information.

  Thereby, the information processing apparatus 100 utilizes the distributed expression including the characteristics of the search query considering the search intention, and uses the second learning model to classify the search query into a category considering the search intention of the user who has input the search query. Can be generated efficiently. In other words, the information processing apparatus 100 can classify the search query into a category in consideration of the search intention of the user who has input the search query. Therefore, the information processing apparatus 100 can improve the classification accuracy of the search query.

[6. Hardware configuration)
Further, the information processing apparatus 100 according to the embodiment described above is realized by a computer 1000 having a configuration as shown in FIG. 13, for example. FIG. 13 is a hardware configuration diagram illustrating an example of a computer that implements the functions of the information processing apparatus 100. 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 depending on the hardware of the computer 1000, and the like.

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

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

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

  For example, when the computer 1000 functions as the information processing apparatus 100 according to the embodiment, the CPU 1100 of the computer 1000 loads a program or data loaded on the RAM 1200 (for example, model data MDT1 of the first model M1 and second model M2). The function of the control unit 130 is realized by executing the model data MDT2). The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them. As another example, these programs or data (for example, the model of the first model M1) from other devices via a predetermined communication network. Data MDT1 and model data MDT2 of the second model M2 may be acquired.

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

[7. Others]
In addition, among the processes described in the above-described embodiments and modifications, all or a part of the processes described as being automatically performed can be manually performed, or are described as being performed manually. All or part of the treatment may be performed automatically by known methods. In addition, information including processing procedures, specific names, 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 figure is not limited to the illustrated information.

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

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

  In addition, the "section (module, unit)" described above can be read as "means" or "circuit". For example, the generation unit can be read as a generation unit or a generation circuit.

DESCRIPTION OF SYMBOLS 1 information processing system 10 user terminal 50 search server 100 information processing apparatus 121 query information storage unit 122 vector information storage unit 123 classification definition storage unit 124 category information storage unit 125 model information storage unit 131 acquisition unit 132 extraction unit 133 generation unit

Claims (17)

An acquisition unit that acquires a first learning model in which features of a plurality of search queries are learned as having distributed features of a pair of search queries input within a predetermined time by the same user.
A generation unit configured to generate a second learning model that predicts a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit;
An information processing apparatus comprising: A plurality of search queries including a character string delimited by a predetermined delimiter as a plurality of search queries input within a predetermined time by the same user are learned as having similar characteristics. An acquisition unit that acquires a first learning model that has learned features of the plurality of search queries, as search queries having similar features;
A generation unit configured to generate a second learning model that predicts a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit;
An information processing apparatus comprising: An acquisition unit configured to acquire a first learning model obtained by learning the features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar features;
A generation unit configured to generate a second learning model that predicts a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit;
With
The acquisition unit
An information processing apparatus that acquires a first learning model that learns features of a plurality of search queries by learning that the plurality of randomly extracted search queries have different features . An acquisition unit configured to acquire a first learning model obtained by learning the features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar features;
A generation unit configured to generate a second learning model that predicts a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit;
With
The acquisition unit
An information processing apparatus that acquires a first learning model in which features of the plurality of search queries are learned by learning such that distributed expressions of a pair of randomly extracted search queries are different . An acquisition unit configured to acquire a first learning model obtained by learning the features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar features;
A generation unit configured to generate a second learning model that predicts a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition unit;
With
The generation unit is
When the search query is input to the learning model, the distributed search classification result output from the learning model is learned so as to correspond to the category to which the search query belongs. An information processing apparatus that generates a second learning model that predicts a category to which the user belongs . The acquisition unit
Claim 1-5, characterized in that the predetermined search query as input information when entered, obtains a first learning model for outputting a distributed representation of the predetermined search query as output information The information processing apparatus according to claim 1. The generator is
When a search query as input information is input to the training model, it claims 1-6, wherein the search query and generates a second learning model that outputs the probability belonging to the category in each category as output information The information processing apparatus according to any one of the above. An information processing method executed by a computer,
Obtaining a first learning model in which features of a plurality of search queries are learned, assuming that the distributed expressions of a pair of search queries input within a predetermined time by the same user have similar features;
A generation step of generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition step;
An information processing method comprising: An acquisition procedure for acquiring a first learning model in which features of a plurality of search queries are learned as having distributed features of a pair of search queries input within a predetermined time by the same user,
A generation procedure for generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition procedure;
An information processing program characterized by causing a computer to execute. An information processing method executed by a computer,
As the plurality of search queries input within a predetermined time by the same user, the plurality of search queries including a character string separated by a predetermined delimiter are learned as having similar characteristics. An acquiring step of acquiring a first learning model obtained by learning the features of the plurality of search queries, assuming that the search queries have similar features;
A generation step of generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition step;
An information processing method comprising: A plurality of search queries including a character string delimited by a predetermined delimiter as a plurality of search queries input within a predetermined time by the same user are learned as having similar characteristics. An acquisition procedure for acquiring a first learning model that has learned features of the plurality of search queries, as search queries having similar features;
A generation procedure for generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition procedure;
An information processing program characterized by causing a computer to execute. An information processing method executed by a computer,
Obtaining a first learning model in which 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;
A generation step of generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition step;
Including
The acquisition step is
An information processing method comprising: acquiring a first learning model in which features of the plurality of search queries are learned by learning that the plurality of randomly extracted search queries have different features . An acquisition procedure for acquiring a first learning model that learns features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar characteristics;
A generation procedure for generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition procedure;
To the computer,
The acquisition procedure is
An information processing program characterized by acquiring a first learning model in which features of the plurality of search queries are learned by learning that the plurality of randomly extracted search queries have different features . An information processing method executed by a computer,
Obtaining a first learning model in which 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;
A generation step of generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition step;
Including
The acquisition step is
An information processing method comprising: acquiring a first learning model obtained by learning features of the plurality of search queries by learning so that distributed expressions of a pair of randomly extracted search queries are different . An acquisition procedure for acquiring a first learning model that learns features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar characteristics;
A generation procedure for generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition procedure;
To the computer,
The acquisition procedure is
An information processing program that acquires a first learning model in which features of the plurality of search queries are learned by learning so that distributed expressions of a pair of randomly extracted search queries are different . An information processing method executed by a computer,
Obtaining a first learning model in which 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;
A generation step of generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition step;
Including
The generation step is
When the search query is input to the learning model, the classification result of the distributed expression output by the learning model is learned so as to correspond to the category to which the search query belongs, whereby the predetermined search query can be obtained from the predetermined search query An information processing method comprising: generating a second learning model that predicts a category to which the user belongs . An acquisition procedure for acquiring a first learning model that learns features of the plurality of search queries, assuming that the plurality of search queries input by the same user within a predetermined time have similar characteristics;
A generation procedure for generating a second learning model for predicting a category to which the predetermined search query belongs from a predetermined search query using the first learning model acquired by the acquisition procedure;
To the computer,
The generation procedure is
When the search query is input to the learning model, the distributed search classification result output from the learning model is learned so as to correspond to the category to which the search query belongs. An information processing program for generating a second learning model for predicting a category to which a user belongs .

Images