Information processing device, information processing method, and information processing program

Abstract

This enables verification of the reliability of the generated model. [Solution] An information processing device according to an embodiment of the present disclosure is characterized by having a first generation unit that generates first quantified information that quantifies the relationship between a search query entered by each seed user and a predetermined behavior based on the commonality of search queries entered by multiple seed users that are related to a predetermined event; a second generation unit that generates second quantified information that quantifies the relationship between the predetermined behavior and each user based on the search queries entered by each user and the first quantified information; an estimation unit that estimates the relevance between the predetermined event corresponding to the second quantified information and the search query that was the source of the second quantified information; and a calculation unit that calculates the percentage of search queries that were the source of the second quantified information that include, for each user, search queries whose estimated relevance satisfies a predetermined condition.

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] There are known methods for predicting user behavior on the Internet.

[0003] For example, Patent Document 1 proposes a technology that learns the search query trends of users on the Internet, calculates a relevance score between the learned search query trends and the user's subsequent behavior to generate a model, and uses this model to predict the behavior of other users, thereby making it possible to predict user behavior with higher accuracy than before. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-177377 Summary of the Invention [Problem to be solved by the invention]

[0005] However, this system had the problem that it was unclear whether the reliability of the generated model was guaranteed.

[0006] The present application has been made in view of the above, and aims to provide a method for verifying the reliability of a generated model. [Means for solving the problem]

[0007] An information processing device according to an embodiment of the present disclosure is characterized by having a first generation unit that generates first quantified information that quantifies the relationship between a search query entered by each seed user and a predetermined behavior based on the commonality of search queries entered by multiple seed users that are related to a predetermined event; a second generation unit that generates second quantified information that quantifies the relationship between the predetermined behavior and each user based on the search query entered by each user and the first quantified information; an estimation unit that estimates the relevance between the predetermined event corresponding to the second quantified information and the search query that was the source of the second quantified information; and a calculation unit that calculates the percentage of search queries that were the source of the second quantified information that include search queries whose estimated relevance satisfies a predetermined condition for each user. [Effects of the Invention]

[0008] According to one aspect of the embodiment, the first group of users that served as the basis for model generation is clustered using a search query, which enables the reliability of the generated model to be evaluated for each cluster, thereby enabling more accurate and reliable prediction of user behavior. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of information processing according to an embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of the information processing device according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of the first digitized information according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of the second digitized information according to the embodiment. [Figure 5] FIG. 5 is a flowchart illustrating an example of information processing according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating a variation of information processing according to the embodiment. [Figure 7] FIG. 7 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing device.

[0010] Hereinafter, a detailed description will be given of an information processing device, an information processing method, and an information processing program (hereinafter referred to as an "embodiment") according to the present application, with reference to the drawings. Note that the information processing device, the information processing method, and the information processing program according to the present application are not limited to the embodiment. Furthermore, the same components in the following embodiments are denoted by the same reference numerals, and redundant description will be omitted.

[0011] (Embodiment) [1. Information Processing] First, an example of information processing according to an embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of information processing according to an embodiment. In the example shown below, the target is a user, and all users are referred to as each user, and a portion of users who have taken a predetermined action among all users is referred to as a seed user. Note that the target is not limited to a user, and may be any target from which information can be collected, such as a city, a product, or a service.

[0012] First, the information processing device 10 stores a part of a group of users who have taken a predetermined action as a seed user U1 (step S1). Whether or not the predetermined action has been taken may be determined by a user questionnaire or the like, or may be determined based on a communication history between the terminal device 100 owned by the seed user U1 and the information processing device 10.

[0013] For example, the name of a certain predetermined store may be input as a search query, and a user who then actually visits the certain predetermined store may be set as seed user U1. As another example, the name of an event may be input as a search query, and a participant who then actually attends the event may be set as seed user U1. As yet another example, the name of a certain product may be input as a search query, and a person who then actually purchases a product with that name may be set as seed user U1. Furthermore, the information processing device 10 may set as seed user U1 a person who actually visits a certain predetermined store (for example, a store corresponding to a distributed advertisement). Furthermore, the information processing device 10 may set as seed user U1 a person who has viewed a certain advertisement, web page, movie, music, or other content.

[0014] That is, the information processing device 10 estimates a user who has performed any expected action as a seed user U1 based on various information such as various purchase histories, usage histories, location histories collected by GPS, etc. For such estimation, any of various known techniques can be adopted.

[0015] Next, the information processing device 10 quantifies the relationship between the search queries entered by each seed user U1 and a predetermined behavior based on the commonality of the search queries entered by the seed users U1 (step S2). This quantified information is defined as first quantified information.

[0016] The first quantified information can be obtained, for example, as follows: When the search query entered by the seed user U1 is taken as a positive example and the search query entered by a randomly selected user is taken as a negative example, the learning model is trained so that it outputs a higher value when a positive example is input and a lower value when a negative example is input. Any method can be used for the training. For example, if a neural network is used as the model, this can be achieved by backpropagation or the like.

[0017] The learning model may be trained to output a higher numerical value for search queries commonly input by the seed user U1. For example, the learning model may be trained to output a higher value the more seed users U1 input search queries that are positive examples. The learning model may also be trained to output a smaller value (negative value) the more users input search queries that are negative examples. A list of pairs of each search query and the score output when the search query is input into such a learning model is generated as a "model." This "model" is an example of first quantified information in this embodiment. That is, the information processing device 10 generates, as first quantified information, a pair of the search query input by the seed user U1 who performed a predetermined action and the score indicating the relationship between the predetermined action and the search query.

[0018] Furthermore, the information processing device 10 stores the search queries input by all users including the seed user U1 (step S3). Then, the information processing device 10 digitizes the relationship between the predetermined behavior and each user based on the search queries input by all users and the first digitized information (step S4). This digitized information is used as second digitized information.

[0019] For example, the information processing device 10 identifies scores associated with search queries entered by users other than the seed user U1 from a model, which is the first quantified information. Then, the information processing device 10 calculates a score indicating the relationship between the behavior performed by the seed user U1 and the user as the second quantified information from the sum of the identified scores. Here, for example, a user who has a high relationship with a predetermined behavior has a high second quantified information, and a user who has a low relationship with the predetermined behavior has a low second quantified information.

[0020] The second quantified information is calculated, for example, as follows: For each of the multiple search queries entered by the user, the associated score (first quantified information) is obtained, and the sum of the scores is calculated. This sum is used as the second quantified information. In this way, the relationship between the multiple search queries entered by the user and a specific behavior can be expressed numerically, thereby achieving more reliable quantification compared to the quantification of a single search query.

[0021] After calculating the second quantified information for each user in this way, the users are ranked in descending order of the value of the second quantified information. In addition, a list of search queries entered by each user is generated in order of rank. Such a list becomes a list (score list) that indicates which users entered search queries that are highly relevant to a specific behavior. Such a list is used for analyzing the relevance between search queries and a specific behavior, analyzing users, etc. Note that such a list is created for each behavior set when selecting seed user U1.

[0022] Here, when the second digitized information is used, it is believed that it is possible to provide information that is likely to interest the user. For example, a list corresponding to an action of visiting a certain store includes a ranking of users who are highly relevant to that action. Therefore, it is believed that the distribution server that distributes advertisements can improve the effectiveness of advertisements by providing advertisements related to that store to users with high rankings.

[0023] On the other hand, when users are ranked in order of relevance to the predetermined behavior based on the second quantified information, it becomes unclear how high a user's rank is actually related to the predetermined behavior. In other words, although it can be assumed that a high value of the second quantified information means a high possibility of performing the predetermined behavior, it is difficult to estimate how high the second quantified information needs to be to indicate a high relevance to the predetermined behavior or whether the user is actually capable of performing the predetermined behavior.

[0024] Furthermore, there was insufficient verification as to whether a high level of the second quantified information can be said to be highly correlated with a predetermined behavior.

[0025] Therefore, after step S4, the information processing device 10 causes an estimation unit to estimate the relevance between the predetermined behavior and the search query input by each user (step S5). For example, the information processing device 10 may estimate the relevance based on whether or not a category related to the predetermined behavior matches a category of the search query input by each user.

[0026] For example, the information processing device 10 estimates the relevance between a predetermined action corresponding to the second digitized information and the search query that is the source of the second digitized information. That is, the information processing device 10 estimates the relevance between a predetermined action (hereinafter, may be referred to as a "seed action") performed by a user who was set as a seed user when generating the first digitized information and a search query (hereinafter, may be referred to as an "input query") entered by a user with a high score among the information included in the second digitized information.

[0027] For example, if the category of the seed action matches that of the input query, the information processing device 10 may determine that the input query has a relationship with the seed action and assign a score of "1" to the search query, indicating the relationship. Furthermore, if the category of the seed action is similar to that of the input query, the information processing device 10 may assign a score of "0.5." On the other hand, if the category of the seed action is not similar to that of the input query, the information processing device 10 may assign a score of "0." It should be noted that the categories that are considered similar or dissimilar may be determined manually in advance, or may be determined based on the similarity of the hierarchical structures to which the categories belong (for example, commonality of upper categories, similarity of hierarchies, etc.). Furthermore, the information processing device 10 may assign a higher score to the higher similarity of the category of the seed action and the input query, and a lower score to the lower similarity.

[0028] Furthermore, the information processing device 10 may calculate the relationship score between the seed action and the input query based on a trained model that has learned the degree of relevance between words, such as w2v. For example, the information processing device 10 may vectorize the seed action and the input query using the trained model of w2v, and calculate the relationship score based on the similarity of each vector (for example, Hamming distance or cosine similarity). In other words, the information processing device 10 may calculate a higher score the higher the relationship between the seed action and the input query (the similarity in the meanings of the words).

[0029] Then, the information processing device 10 calculates the proportion of search queries that are determined by the trained model to have a predetermined relationship. For example, the information processing device 10 calculates, for each ranking in the second digitized information, the proportion of search queries that have an assigned score exceeding a predetermined threshold. That is, the information processing device 10 causes the calculation unit to calculate, for each user's search query, the proportion of search queries that satisfy a predetermined condition (for example, matching categories) in the estimated relevance (step S6).

[0030] In this way, if the proportion of search queries that match the category of a predetermined behavior and the category of the search query are high among the multiple search queries of the top users among the users sorted by the second quantified information, it can be inferred that the top users among the users sorted by the second quantified information have a high relevance to the predetermined behavior.On the other hand, if the proportion of search queries related to seed behavior among the search queries entered by the top users is low, it can be inferred, for example, that the reliability of the first quantified information and the second quantified information may be low.In this way, the strength of the relevance to the predetermined behavior, which cannot be determined simply by sorting by the second quantified information, can be verified using the parameter of the proportion calculated by the calculation unit.

[0031] The information processing device 10 may calculate the score based on a weighting set for each search query. For example, the information processing device 10 may set a higher weighting value the more similar the category to which the search query belongs and the category to which the seed action belongs, and may use the weighting multiplied by a numerical value indicating the similarity between the search query and the seed action as the score. The information processing device 10 may also set a weighting that is set in advance depending on the category to which the search query belongs.

[0032] Here, a technique for estimating the search intent of a search query using a DQNN (Deep Q Neural Network) is known. For example, a model is trained to convert multiple search queries entered consecutively within a predetermined time period by a user, i.e., search queries estimated to have similar search intents, into vectors with high similarity, and to convert multiple search queries entered at different times, i.e., search queries estimated to have distant search intents, into vectors with low similarity.

[0033] The information processing device 10 may calculate a score indicating the relevance to a seed behavior by taking into account the search intent estimated for each search query using such DQNN technology. For example, the information processing device 10 acquires data to which a vector generated from the search query is attached as metadata using a model trained using DQNN technology for the search query. That is, the information processing device 10 acquires search query data to which metadata indicating the search intent is attached. Note that such metadata may be various multidimensional quantities such as distributed representations.

[0034] Then, the information processing device 10 may calculate a score indicating a relationship with a seed behavior for each pair of a character string that is a search query and metadata. That is, even if different users input the same character string as a search query, if the search intentions of the users are different, the information processing device 10 may calculate the score and the relationship with a seed behavior for each of the users as different search queries.

[0035] Furthermore, the information processing device 10 may determine a title associated with a predetermined behavior, i.e., information indicating a seed behavior, based on a search query whose calculated score satisfies a predetermined condition. For example, the information processing device 10 may identify one or more search queries whose calculated score exceeds a predetermined threshold. Note that the information processing device 10 may identify a predetermined number of search queries, for example, in descending order of score. Then, the information processing device 10 generates, based on the identified search query, a seed behavior corresponding to the second quantified information or the first quantified information from which the second quantified information is derived, or a title for identifying each piece of information. For example, the information processing device 10 may use a list of the identified search queries as the title, or may use various sentence generation techniques to generate a sentence using the identified search query as the title.

[0036] 2. Configuration of Information Processing Device Next, the configuration of the information processing device 10 according to the embodiment will be described with reference to FIG. 2. FIG. 2 is a diagram illustrating an example of the configuration of the information processing device 10 according to the embodiment. The information processing device 10 generates first quantified information that quantifies the relationship between the search query entered by each seed user U1 and the predetermined behavior based on the commonality of the search queries entered by the seed users U1 who performed the predetermined behavior. The information processing device 10 also generates second quantified information that quantifies the relationship between the predetermined behavior and each user based on the generated first quantified information and the search query entered by each user. Furthermore, the information processing device 10 classifies each user into two or more clusters based on the commonality of the search queries entered by each user. As shown in FIG. 2, the information processing device 10 includes a communication unit 20, a storage unit 30, and a control unit 40. The information processing device 10 may also include a display unit that displays various information and an input unit that inputs various information.

[0037] The communication unit 110 is realized by, for example, a NIC etc. The communication unit 110 is connected to a predetermined network via wired or wireless connection, and transmits and receives information to and from external terminal devices and information processing devices.

[0038] (Storage unit 120) The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 120 according to the embodiment may store information about each user, first digitized information, and second digitized information.

[0039] (First digitized information) FIG. 3 is a diagram illustrating an example of first quantified information according to an embodiment. As illustrated in FIG. 3, the first quantified information quantifies the relationship between a search query entered by a seed user U1 and a predetermined behavior performed by the seed user U1. In the example illustrated in FIG. 3, for example, if there is an all-you-can-eat steak chain called "Steak Shochikubai," and the behavior of actually visiting a Steak Shochikubai restaurant is assumed to be the predetermined behavior of the seed user U1, it can be seen that there is a very high correlation between the behavior of entering "Steak Shochikubai" as a search word and the behavior of actually visiting "Steak Shochikubai." It can also be seen that relatively high scores are assigned to search words that represent attributes of "Steak Shochikubai," such as "all-you-can-eat," and search words for meat dishes similar to steak, such as "shabu-shabu." While the example in FIG. 3 deals with a visit to the "Steak Shochikubai" restaurant, the predetermined behavior is not limited to this. That is, the predetermined behavior may be a visit to a specific restaurant or the behavior of purchasing a specific product. In this case, individual first quantified information is generated for each behavior.

[0040] (Second digitized information) 4 is a diagram showing an example of second digitized information according to an embodiment. The second digitized information is a digitized representation of the relationship between a predetermined behavior and each user U based on the search query entered by each user U and the first digitized information. In the example shown in FIG. 4, the users are sorted in ascending order starting from the user with the highest second digitized value (hereinafter referred to as "score"). The score and search query of each user are also shown.

[0041] In Figure 4, the score is calculated by applying the first digitized information, which is information that quantifies the relevance between a predetermined behavior and the search word, to each search word in each user's search log and adding up the results. That is, for each search word in each user U's search log, a search is performed for words similar to that word in the first digitized information, and if the word is found as a result of the search, the score assigned to that word is added to the total score. This calculation is performed for each word in the search log. In this way, it is possible to numerically express the likelihood of a user taking a predetermined behavior from that user's search log, and it is expected that it will be possible to predict with high accuracy whether or not the user will take a predetermined behavior.

[0042] In the example of FIG. 3, a visit to a restaurant called "Steak Shochikubai" is handled, but as with the first digitized information, the predetermined behavior to be handled is not limited to this. That is, a visit to a specific restaurant may be the predetermined behavior, or the act of purchasing a specific product may be the predetermined behavior. In such cases, separate second digitized information is generated for each behavior.

[0043] (control unit 40) 2, the control unit 40 is realized by, for example, a CPU, an MPU, or the like executing various programs stored in a storage device inside the information processing device 10 using RAM as a work area. The control unit 40 is also realized by, for example, an integrated circuit such as an ASIC or an FPGA.

[0044] As shown in FIG. 2, the control unit 40 has a first generation unit 41, a second generation unit 42, an estimation unit 43, a calculation unit 444, and a title determination unit 45, and realizes or executes the functions and actions of the information processing described below. The internal configuration of the control unit 40 is not limited to the configuration shown in FIG. 2 and may be other configurations as long as they perform the information processing described below. The connection relationship between the processing units of the control unit 40 is not limited to the connection relationship shown in FIG. 2 and may be other connection relationships. The control unit 40 may also have a receiving unit when receiving various information such as first digitized information from an external information processing device. The control unit 40 may also have a transmitting unit when transmitting second digitized information to an external information processing device.

[0045] (1st generation part) The first generation unit 41 generates first quantified information based on various information through the above-described process. In the embodiment, the first generation unit 41 generates first quantified information that quantifies the relationship between the search query and the predetermined behavior based on the commonality of the search queries entered by multiple seed users who performed the predetermined behavior. Here, the first generation unit 41 may generate first quantified information for each of the multiple predetermined behaviors.

[0046] (2nd generation part) The second generation unit 42 generates second quantified information indicating a relationship between a search query input by each user and a predetermined behavior of each user U, based on the search query input by each user and the first quantified information. Here, the second generation unit 42 may generate second quantified information for each of a plurality of predetermined behaviors. Furthermore, the second generation unit may generate second quantified information including information on the search query input by each user. That is, the second quantified information may be information relating a user ID to a score, but as shown in FIG. 4, the search log may also be information stored for each user as the second quantified information.

[0047] The second generating unit 42 may further generate a list of second digitized information only for users extracted based on information of users in the same cluster as the user whose value of the second digitized information is equal to or greater than the threshold.

[0048] (Estimation part 43) The estimation unit 43 estimates the relevance between the predetermined behavior corresponding to the second quantified information and the search query that is the source of the second quantified information. That is, the estimation unit 43 estimates the degree of relevance of each of the multiple search queries entered by each user to the predetermined behavior. For example, the relevance may be estimated based on whether a category related to the predetermined behavior matches the category of each search query. Alternatively, the relevance may be estimated by acquiring a distributed representation of words using a learning model that has previously learned the degree of relevance between words, and calculating the relevance between the name assigned to the predetermined behavior and each search query as a numerical value.

[0049] (Calculation unit 44) The calculation unit 44 calculates, for each user, the proportion of search queries that are the basis of the second quantified information that include search queries whose estimated relevance satisfies a predetermined condition. That is, the calculation unit 44 calculates the proportion of search queries that the estimation unit 43 estimates to be relevant to a predetermined behavior that are included in the multiple search queries of each user. In this way, it is possible to obtain a method for numerically verifying the relevance between a search query and a predetermined behavior for each user.

[0050] The calculation unit 44 may also estimate the relevance between a predetermined behavior and a search query based on whether or not the category underlined for each search query matches the category of the predetermined behavior.

[0051] The calculation unit 44 may also represent the relevance between a predetermined behavior and each search query as a numerical value, and use the numerical value being equal to or greater than a predetermined value as a basis for estimating the relevance. An example of a method for calculating the numerical value is, but is not limited to, a method of calculating the degree of relevance between words using a trained model based on the trained model.

[0052] The calculation unit 44 may also calculate the ratio based on a weighting set for each search query. For example, if the relevance between a predetermined behavior and each search query can be expressed as a numerical value using a trained model, when the numerical value of the relevance of a certain search query is high, the calculation unit may assign a heavier weight when calculating the ratio. Conversely, when the numerical value of the relevance of a certain search query is low, the calculation unit may assign a lighter weight when calculating the ratio.

[0053] (Title Decision Section 45) The title determination unit 45 determines a title to be attached to a predetermined behavior based on a search query that satisfies the predetermined conditions calculated by the calculation unit 44. That is, when the operator OP determines a title to be attached to a predetermined behavior, the title does not necessarily appropriately describe the predetermined behavior (for example, if there is a title that simply states "sample," it is unclear from the title what the predetermined behavior is). Even in such a case, appropriate words can be extracted from the information of the search query entered by each user, and the title can be replaced with meaningful words. In this way, the operator OP can be spared the trouble of inputting an appropriate title. Furthermore, since there are no predetermined behaviors to which meaningless titles are assigned, the efficiency of the operation OP can be improved.

[0054] [3. Information processing flow] Next, the procedure of information processing by the information processing device 10 according to the embodiment will be described with reference to Fig. 5. Fig. 5 is a flowchart showing the procedure of information processing by the information processing device 10 according to the embodiment.

[0055] 5, the information processing device 10 identifies a group of users who have taken a predetermined action as a seed user group (step S101). Then, the first generation unit 41 generates digitized information, which is the first digitized information, representing the relationship between the search query input by each seed user and the predetermined action, based on the commonality of the search queries input by the seed user group (step S102). Note that when the first digitized information is acquired, the information processing device 10 does not need to perform the processes of steps S101 and S102.

[0056] Then, the second generation unit 42 of the information processing device 10 generates quantified information based on the search query entered by each user and the first quantified information, which represents the relevance between the specified behavior and each user as second quantified information (step S103).

[0057] Next, the estimation unit 45 of the information processing device 10 estimates the relevance between the predetermined behavior and the search query of each user (step S104). Then, as a result of estimating the relevance, the proportion of search queries that satisfy a predetermined condition is calculated for each user.

[0058] [4. Modifications] The information processing device 10 according to the embodiment described above may be implemented in various different forms other than the above embodiment, so other embodiments of the information processing device 10 will be described below.

[0059] [4-1. Prescribed Actions] In the above-described embodiment, the information processing device 10 may generate first quantified information and second quantified information for each of a plurality of predetermined actions. In the above-described embodiment, entering the "Steak Shochikubai" restaurant was defined as the predetermined action, but the information processing device 10 may also generate first quantified information and second quantified information for each of other actions, such as the user U purchasing a predetermined product or installing a predetermined application on a terminal device owned by the user U. Furthermore, the generated second quantified information may be evaluated for each action. Furthermore, the predetermined action may be inputting a predetermined reference search query.

[0060] [4-2. Matching of names in the second digitized information and search logs] In the above-described embodiment, the evaluation of the second quantified information was performed based on users divided into clusters, but the evaluation method is not limited to this. For example, if an operator has previously assigned a name to the second quantified information, the matching rate between the category related to that name and the category of each search word in the search log can be calculated, and if the matching rate is high, it can be evaluated that appropriate second quantified information has been obtained. In this case, a possible method for determining the category from the name or search word is to create a dictionary in which various words are previously assigned categories and search from that dictionary. Alternatively, the category may be determined based on a learning model that has previously learned the degree of association between words.

[0061] Alternatively, instead of calculating the matching rate between categories, the similarity between the name and each search word may be calculated. That is, Word2Vec or the like may be used to obtain a distributed representation of words using a learning model that has previously learned the degree of association between words, and this learning model may be used to calculate the similarity between the name in the second digitized information and each search word in the search log.

[0062] [4-3. Seeding Activities] In the above example, the information processing device 10 determines a predetermined action performed by a user designated as a seed user as a seed action. However, the embodiment is not limited to this. For example, the information providing device 10 may determine a user who has made a search query or posted on an SNS about "hamburger chain store A" as a "user who likes hamburger chain store A," and generate first quantified information and second quantified information about such a "user who likes hamburger chain store A." Furthermore, the information processing device 10 may generate first quantified information and second quantified information according to the attributes of the user, such as "a user in his 30s who likes hamburger chain store A."

[0063] Furthermore, the information processing device 10 may generate the first quantified information and the second quantified information by regarding users who have performed a plurality of predetermined actions having a predetermined relationship as seed users. For example, the information processing device 10 may regard not only users who purchase camping equipment but also users who purchase ski equipment as seed users who "like the outdoors" or "likely have children," and generate the first quantified information and the second quantified information.

[0064] That is, the information processing device 10 may generate the first quantified information and the second quantified information using any event other than the seed action corresponding to the first quantified information and the second quantified information as a seed event, and calculate the relationship between the search query in the second quantified information and the seed event. That is, the information processing device 10 sets a user who is associated with a predetermined event (for example, a user's behavior or actions, or various attributes that can be estimated from registered information) as a seed user. For example, the information processing device 10 sets a user who is associated with a predetermined event as a seed user based on various user behaviors or pre-registered information, and generates the first quantified information and the second quantified information. Then, the information processing device 10 estimates the relationship between the predetermined event corresponding to the second quantified information and the search query that is the source of the second quantified information. For example, the information processing device 10 estimates the relationship between the predetermined event and the search query entered by each user based on W2V or the like. To give a more specific example, the information processing device 10 may infer that there is a strong relationship when the meaning of a character string indicating a specified event is similar to the meaning of a search query or when they belong to the same category, and may infer that there is a weak relationship when the meaning of a character string indicating a specified event is not similar to the meaning of a search query or when they do not belong to the same category.

[0065] For example, Fig. 6 is a diagram illustrating a variation of information processing according to an embodiment. For example, in the example shown in Fig. 6, it is assumed that a user attribute such as "a man in his 30s who likes yakiniku" is specified as the seed event, rather than the search query itself. Note that the seed event can be any of a variety of events, such as an action taken by the user, an active action or location-based event such as "I was in Tokyo on March 2nd," or the name of an event attended.

[0066] When such a seed event is set, the information processing device 10 identifies a search query that is estimated to be entered by a seed user. For example, the information processing device 10 identifies a search query such as "yakiniku, all-you-can-eat, delicious, charcoal grilled...". Note that such a search query may be automatically set from among search queries using various user attribute estimation techniques, or may be set by a predetermined operator. The information processing device 10 then identifies a user who is associated with the predetermined event as a seed user. For example, the information processing device 10 identifies a user who actually entered a search query that is estimated to be entered by a seed user, and whose attribute information is registered as "male in his 30s" or is estimated to be "male in his 30s", as the seed user. The information processing device 10 then generates first quantified information using the search query entered by the seed user, and generates second quantified information based on the first quantified information.

[0067] Furthermore, the information processing device 10 identifies search queries entered by each user in the second digitized information and estimates the relationship between the identified search queries and the seed event. For example, the information processing device 10 estimates the relationship based on the similarity between the category to which the search query identified from the second digitized information belongs and the category to which the character string "a man in his 30s who likes yakiniku" representing the seed event belongs. Note that this processing may be realized, for example, using a model (e.g., a w2v model) that categorizes objects represented by various character strings based on relative similarity.

[0068] Note that if the seed event shown in FIG. 6 is defined by the search query itself, for example, the search query itself entered by a user, or the user who entered a specific search query, the relationship may be estimated based on whether the search query and the search query identified from the second quantified information belong to the same category. Furthermore, the information processing device 10 may estimate the relationship based on the similarity between the categories to which each query belongs. Furthermore, the seed event may be defined by multiple search queries. In such cases, the relationship is estimated by comprehensively determining the similarity between each search query and the search query identified from the second quantified information.

[0069] [5. Hardware Configuration] The information processing device 10 according to the embodiment described above is realized by, for example, a computer 1000 configured as shown in Fig. 7. Fig. 7 is a hardware configuration diagram showing an example of the computer 1000 that realizes the functions of the information processing device 10. The computer 1000 has a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0070] The CPU 1100 operates and controls each unit based on programs stored in the ROM 1300 or the HDD 1400. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 starts up, programs that depend on the hardware of the computer 1000, and the like.

[0071] The HDD 1400 stores programs executed by the CPU 1100, data used by these programs, etc. The communication interface 1500 receives data from other devices via a predetermined network and sends it to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined network.

[0072] The CPU 1100 controls output devices such as a display and a printer, and input devices such as a keyboard and a mouse, via the input / output interface 1600. The CPU 1100 acquires data from the input devices via the input / output interface 1600. The CPU 1100 also outputs generated data to the output devices via the input / output interface 1600.

[0073] Media interface 1700 reads a program or data stored in recording medium 1800 and provides it to CPU 1100 via RAM 1200. CPU 1100 loads the program or data from recording medium 1800 onto RAM 1200 via media interface 1700 and executes the loaded program. Recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0074] For example, when the computer 1000 functions as the information processing device 10 according to the embodiment, the CPU 1100 of the computer 1000 executes programs loaded onto the RAM 1200 to realize the functions of the control unit 130. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, the CPU 1100 may obtain these programs from another device via a predetermined network.

[0075] Although some of the embodiments of the present application have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be implemented in other forms that incorporate various modifications and improvements based on the knowledge of those skilled in the art, including the aspects described in the Disclosure of the Invention.

[0076] [6. Other] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods. Furthermore, the information, including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings, can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0077] Furthermore, the components of each device shown in the figure are conceptual functional components and do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0078] Furthermore, the above-described embodiments can be combined as appropriate within the scope of not causing any contradiction in the processing content.

[0079] Furthermore, the above-mentioned "section, module, unit" can be read as "means" or "circuit," etc. For example, the first generation unit can be read as first generation unit means or first generation unit circuit. [Explanation of symbols]

[0080] 10. Information processing equipment 20 Communications Department 30 Storage section 40 Control Unit 41 1st generation part 42 Second generation part 43 Display control unit 44 Evaluation Department 100 Terminal Device

Claims

1. a generation unit that generates information indicating a relationship between a predetermined event and each user based on a relationship between a search query input by a seed user and a predetermined event related to the seed user, and the search query input by each user; an estimation unit that estimates a relevance between the predetermined event and a search query input by each user, the search query being the basis of the relationship, based on the information generated by the generation unit; a calculation unit that calculates, for each user, a ratio of search queries entered by the user that include search queries whose estimated relevance satisfies a predetermined condition; An information processing device comprising:

2. The calculation unit sets the predetermined condition that a category set for each search query matches a category set for the predetermined event. The information processing device according to claim 1 .

3. the calculation unit calculates a rate at which search queries whose relevance exceeds a predetermined threshold are included. The information processing device according to claim 1 .

4. The calculation unit calculates, based on a trained model that has learned the degree of relevance between words, a proportion of search queries that are determined by the trained model to have a predetermined relationship with the predetermined event. The information processing device according to claim 1 .

5. The calculation unit is further characterized in that the calculation is performed based on a weighting set for each of the search queries.

5. The information processing device according to claim 1.

6. The information processing device further comprises: a title determination unit that determines a title associated with the predetermined event based on the search query that satisfies the predetermined condition calculated by the calculation unit; 5. The information processing device according to claim 1, wherein:

7. An information processing method executed by an information processing device, a generating step of generating information indicating a relationship between a predetermined event and each user based on a relationship between a search query input by a seed user and a predetermined event related to the seed user, and the search query input by each user; an estimation step of estimating a relevance between the predetermined event and a search query input by each user, which is a source of the relationship, based on the information generated by the generation step; a calculation step of calculating, for each user, a proportion of search queries entered by the user that include search queries whose estimated relevance satisfies a predetermined condition; An information processing method comprising:

8. a generation step of generating information indicating a relationship between a predetermined event and each user based on a relationship between a search query input by a seed user and a predetermined event related to the seed user, and the search query input by each user; an estimation step of estimating a relevance between the predetermined event and a search query input by each user, the search query being the basis of the relationship, based on the information generated by the generation step; a calculation step of calculating, for each user, a percentage of search queries entered by the user that include search queries whose estimated relevance satisfies a predetermined condition; An information processing program that causes a computer to execute the above.

Images