Interest / preference inference device and interest / preference inference method

Abstract

Provided are an interest / preference inference device and an interest / preference inference method that can more accurately infer the interests / preferences of users by considering that the meaning of the attributes of POI can change depending on the locations of the users.　An interest / preference inference device 10 comprises an input unit 11 that inputs information about the attributes of points of interest (POI) and information about a plurality of set areas that have different characteristics, an attribute specifying unit 14 that specifies attributes of the POI that correspond to information about the location of a user on the basis of the information about the attributes of the POI, and a determination unit 15 that determines first information for inferring the interests / preferences of the user on the basis of the attributes of the POI specified by the attribute specifying unit 14 and stay area information inferred on the basis of the information about the location of the user using the information about the set areas.

Description

Hobby and taste estimation device and hobby and taste estimation method

[0001] One aspect of the present invention relates to a hobby and preference estimation device and a hobby and preference estimation method.

[0002] A technique is known in which a user estimates a point of interest (POI) that the user will visit based on the relationship between the location of the POI and the user's stay location, and the user's interests and tastes are estimated based on the POI. Patent Literature 1 describes a technique in which a user's interests and tastes are estimated based on category information of the POI in accordance with visit date and time information of the POI that the user will visit.

[0003] International Publication No. 2019 / 202783

[0004] In Patent Document 1, POI categories are pre-associated with types indicating hobbies and interests at a given date and time, and a user's hobbies and interests are estimated based on the date and time of visit to the POI and the POI category. However, if the location of a user whose hobbies and interests are to be estimated changes, the meaning of the POI category may change. For example, in Japan, the POI category "SPA" refers to places related to hot springs and public baths, but in the United States, it refers to places related to beauty salons for women. In such cases, it may be difficult to accurately estimate a user's hobbies and interests based on the POI category.

[0005] One aspect of the present invention has been made in consideration of the above-mentioned situation, and aims to provide a hobby / interest estimation device and a hobby / interest estimation method that are capable of more accurately estimating a user's hobby / interests, taking into account that the meaning of a POI attribute may differ depending on the user's location.

[0006] In order to achieve the above object, the system includes an input unit that inputs information on the attributes of a POI (Point Of Interest) and information on a plurality of set areas having mutually different characteristics, an attribute identification unit that identifies the attributes of the POI according to information on the user's location based on the information on the attributes of the POI, and a determination unit that determines first information, which is information for estimating the user's hobbies and preferences, based on stay area information estimated using information on the set area based on the information on the user's location and the attributes of the POI identified by the attribute identification unit.

[0007] In the hobby and preference estimation device according to the present invention, the characteristics of multiple set areas are preset to be different from one another. Information for estimating a user's hobby and preference is then determined based on the set area corresponding to information related to the user's location and the attributes of POIs corresponding to the location of the information related to the user. With this configuration, the characteristics (e.g., cultural background) of the set area corresponding to the information related to the user's location are taken into consideration, and the information for estimating a user's hobby and preference is determined based on the attributes of POIs corresponding to the information related to the user's location. For example, the information for estimating a user's hobby and preference is determined based on the attribute of a POI corresponding to the user's location being "hot spring" considering that the user is located in Japan. Furthermore, for example, the information for estimating a user's hobby and preference is determined based on the attribute of a POI corresponding to the user's location being "spa" considering that the user is located in the United States. This enables a more accurate estimation of a user's hobby and preference, taking into consideration that the meaning of a POI attribute may differ depending on the user's location.

[0008] According to one aspect of the present invention, it is possible to estimate a user's interests and tastes with higher accuracy by taking into consideration that the meaning of a POI attribute may differ depending on the user's location.

[0009] FIG. 1 is a diagram for explaining an overview of processing in the hobby and preference estimation device according to the first embodiment. FIG. 2 is a block diagram illustrating the hobby and preference estimation device according to the first and second embodiments. FIG. 3 is a diagram illustrating an example of data indicating information about a user's location and data indicating information about a set area corresponding to the information about the user's location. Part (a) of FIG. 4 is a diagram illustrating an example of data indicating information about attributes of POIs in a predetermined set area, and part (b) of FIG. 4 is a diagram illustrating an example of data indicating information about attributes of POIs in another set area. FIG. 5 is a diagram illustrating an example of data indicating information about POI attributes corresponding to information about the user's location. FIG. 6 is a block diagram illustrating the functional configuration of a determination unit of the hobby and preference estimation device according to the first embodiment. FIG. 7 is an example of data illustrating a state in which multiple combinations of information including stay area information and POI attributes identified by the attribute identification unit are classified into multiple groups. FIG. 8 is a diagram illustrating an example of information indicating a user's degree of interest in a group. Part (a) of FIG. 9 is a diagram illustrating an example of information about hobby and preference attributes, and part (b) of FIG. 9 is a diagram illustrating an example of attribute correct answer information. FIG. 10 is a flowchart showing an example of processing for estimating a user's hobbies and preferences. FIG. 11 is a diagram for explaining an overview of processing in a hobby and preference estimation device according to a second embodiment. FIG. 12 is a block diagram showing the functional configuration of a determination unit of the hobby and preference estimation device according to the second embodiment. FIG. 13 is a diagram showing an example of data indicating information about a character string. FIG. 14 is a diagram showing an example of a vector based on information about the character string. Part (a) of FIG. 15 is a diagram showing an example of attribute correct answer information, and part (b) of FIG. 15 is a diagram showing an example of information about hobby and preference attributes based on a vector. FIG. 16 is a diagram showing the hardware configuration of the hobby and preference estimation device according to the first embodiment and the second embodiment.

[0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0011] First Embodiment FIG. 1 is a diagram illustrating an overview of processing in a hobby and taste estimation device according to a first embodiment. The hobby and taste estimation device 10 acquires information indicating a set area according to a user's location and information indicating attributes of a point of interest (POI) according to the user's location. The hobby and taste estimation device 10 generates preparatory information for estimating the user's hobby and taste based on the acquired information. The hobby and taste estimation device 10 estimates the user's hobby and taste based on the generated preparatory information. For example, the hobby and taste estimation device 10 calculates a first score for each hobby and taste attribute for each user.

[0012] A plurality of set areas are set in advance. The plurality of set areas have different characteristics from each other. Specifically, each set area is a predetermined geographical area and is set based on the characteristics of that area. The characteristics of the set area are the characteristics or properties of the predetermined geographical area. As an example, the plurality of set areas have different cultural backgrounds from each other. The cultural background means the formation of the culture in the predetermined area. For example, the plurality of set areas may correspond to multiple countries, or may correspond to multiple areas where different languages ​​are used, or may correspond to multiple regions within a single country, or may correspond to multiple areas set based on criteria other than those described above. The plurality of set areas may overlap with each other.

[0013] A POI is a place that may be visited by a person. A POI represents a specific store or facility, for example, convenience store A and stadium B. A POI attribute (POI category) is an attribute associated with a set of POIs. Examples of POI attributes are convenience stores and stadiums. Hobby and preference attributes are attributes that indicate a user's hobbies, preferences, and interests, such as a love of sports, travel, hot springs, and beauty.

[0014] 1 , the hobby and taste estimation device 10 acquires information indicating that the user stayed in the United States and information indicating that the attributes of the POI visited by the user are "Park" and "Spa." The hobby and taste estimation device 10 acquires location information of the user, information indicating that the user stayed in Japan, and information indicating that the attributes of the POI visited by the user are "Hot Spring / Spa" and "Park."

[0015] The hobby and preference estimation device 10 classifies the POI attribute "Spa" in the United States and the POI attribute "Hot Spring / Spa" in Japan into different groups because the meaning of the POI attribute "Spa" in the United States is different from the meaning of the POI attribute "Hot Spring / Spa" in Japan. The hobby and preference estimation device 10 classifies the POI attribute "Park" in the United States and the POI attribute "Park" in Japan into the same group because the meaning of the POI attribute "Park" in the United States is the same as the meaning of the POI attribute "Park" in Japan. The hobby and preference estimation device 10 estimates the hobby and preference of a user by inputting multiple first scores calculated for each of multiple groups into an estimation model.

[0016] Next, a functional configuration of the hobby and taste estimation device 10 will be described. Fig. 2 is a block diagram showing the hobby and taste estimation device 10 according to the first embodiment. As shown in Fig. 2, the hobby and taste estimation device 10 includes an input unit 11, a setting unit 12, a storage unit 13, an attribute identification unit 14, a determination unit 15, an estimation unit 16, and a construction unit 17.

[0017] The input unit 11 acquires user location information. The input unit 11 outputs the acquired location information to the attribute identification unit 14. In the first embodiment, the user location information is information indicating the user's location, such as GPS information. The input unit 11 acquires the user location information from, for example, a mobile terminal held by the user. Note that the user location information may be information that has been preprocessed, such as by removing noise and correcting numerical fluctuations.

[0018] The user's location information is associated with time information indicating the stay start time, which is the time when the user started staying, and the stay end time, which is the time when the user ended staying, as well as identification information, which is information that identifies the user. Figure 3 is a diagram showing an example of data indicating the user's location information and an example of data indicating a set area according to the user's location. In the example shown in Figure 3, the location information is information that indicates the latitude and longitude of the user's location. The time information is information that indicates the stay start time, which is the time when the user started staying at a specified location, and information that indicates the stay end time, which is the time when the user ended staying. The user identification information is a user ID.

[0019] The input unit 11 inputs information about the attributes of the POI to the storage unit 13. The information about the attributes of the POI is, for example, POI attribute information that indicates the attributes of the POI according to its location.

[0020] The input unit 11 inputs information about a plurality of set areas having different characteristics to the storage unit 13. Specifically, the input unit 11 inputs information about a plurality of set areas having different cultural backgrounds. As an example, the input unit 11 inputs information about a plurality of set areas corresponding to a plurality of countries, respectively. In addition, the input unit 11 inputs attribute correct answer information to the construction unit 17.

[0021] The setting unit 12 sets a plurality of setting areas. The setting unit 12 outputs setting area information indicating the positions of the plurality of setting areas to the input unit 11. The setting area information is, for example, information indicating the latitude and longitude of a position included in each setting area.

[0022] The storage unit 13 stores POI attribute information and set area information. The POI attribute information indicates the attributes of a POI according to its location. In the first embodiment, the POI master DB 13a of the storage unit 13 stores POI attribute information for each set area. Sections (a) and (b) of FIG. 4 are diagrams illustrating an example of data indicating POI attribute information for each set area. In the example shown in section (a) of FIG. 4, POI attribute information including information indicating the location of the POI, information indicating the attributes of the POI, information indicating the name of the POI, and information identifying the POI is shown. In section (a) of FIG. 4, multiple pieces of POI attribute information are associated with a set area called "Japan." In the example shown in section (b) of FIG. 4, the POI attribute information includes the same information as above. In section (b) of FIG. 4, multiple pieces of POI attribute information are associated with a set area called "United States."

[0023] The attribute identification unit 14 identifies attributes of the POI corresponding to information about the user's location based on information about the POI's attributes. The information about the user's location includes, for example, the user's location information and buildings visited by the user. The POI's attributes corresponding to the information about the user's location include, for example, attributes of the POI corresponding to the user's location. Specifically, the attribute identification unit 14 identifies attributes of the POI corresponding to the user's location based on the user's location information acquired by the input unit 11 and the POI attribute information stored in the storage unit 13. In the first embodiment, the attribute identification unit 14 includes a stay area estimation unit 14a and a linking processing unit 14b. The stay area estimation unit 14a estimates stay area information indicating a set area corresponding to the user's location from the user's location information and the set area information. Specifically, the stay area estimation unit 14a estimates stay area information indicating a set area corresponding to the user's location by comparing the user's location with the location of the set area. That is, the stay area estimation unit 14a determines the set area where the user is staying based on the latitude and longitude of the user's location. The stay area estimation unit 14a outputs the stay area information to the linking processing unit 14b. The process of estimating the stay area information by comparing the user's position with the position of the set area is realized by a known technique.

[0024] In the example shown in FIG. 3 , when the latitude and longitude of the location of user "user_1" are 35.1 degrees and 130.1 degrees, the stay area estimation unit 14a determines that the set area corresponding to the location of user "user_1" is "Japan" because these latitude and longitude are included in the set area "Japan." The stay area estimation unit 14a associates the determined set area "Japan" with the user's location information. The stay area estimation unit 14a performs the same process as above for users "user_2" to "user_4."

[0025] The linking processing unit 14b identifies attributes of POIs corresponding to the user's location based on the user's location information, stay area information, and POI attribute information. Specifically, the linking processing unit 14b acquires the user's location information acquired by the input unit 11, the stay area information estimated by the stay area estimation unit 14a, and the POI attribute information stored for each set area by the storage unit 13. The linking processing unit 14b extracts multiple pieces of POI attribute information corresponding to the set area indicated by the stay area information. The linking processing unit 14b compares the positions of the multiple POIs included in the extracted POI attribute information with the user's location information to extract the position of the POI corresponding to the user's location. The linking processing unit 14b extracts attributes of POIs corresponding to the extracted positions of POIs by referring to the POI attribute information. The linking processing unit 14b sets the extracted POI attributes as the attributes of the POI corresponding to the user's location. In this way, the attribute identification unit 14 identifies the attributes of the POI corresponding to the user's location. The process of extracting the position of the POI corresponding to the user's position is realized by a known technique.

[0026] The attribute identification unit 14 may generate, as combination information, information that associates the attribute of the POI according to the user's location with the set area according to the user's location. Specifically, the attribute identification unit 14 associates information associated with the user's location information with information associated with the attribute of the POI, based on the attribute of the POI according to the user's location.

[0027] FIG. 5 is a diagram illustrating an example of data indicating attributes of POIs corresponding to a user's location. In the examples illustrated in FIGS. 3 to 5, the attribute identification unit 14 extracts multiple pieces of POI attribute information (e.g., multiple pieces of POI attribute information shown in part (a) of FIG. 4) corresponding to a set area (e.g., "Japan") associated with the user's location. The attribute identification unit 14 extracts the location of the POI corresponding to the user's location by comparing the latitude and longitude of the multiple POI locations in the extracted POI attribute information with the latitude and longitude of the user's location shown in FIG. 3 (e.g., latitude 35.1 degrees and longitude 130.1 degrees). The attribute identification unit 14 sets the POI attribute corresponding to the extracted POI location (e.g., "SPA / Hot Spring") as the attribute of the POI corresponding to the user's location. Then, as illustrated in FIG. 5, the attribute identification unit 14 generates visited POI information that associates information associated with the user's location information with information associated with the POI attribute based on the POI attribute corresponding to the user's location. For example, the attribute identification unit 14 generates visited POI information in which the POI ID, POI attribute, and POI name are associated with the user's ID, stay start time, stay end time, and set area. Therefore, the second information described below is associated with third information, which is information regarding the time the user stayed at a place related to the location. The third information is, for example, information indicating the time the user stayed at a place related to the location. The time the user stayed at a place related to the location is the time the user stayed at a place related to the user's location.

[0028] The determination unit 15 determines first information, which is information for estimating the user's hobbies and preferences, based on stay area information estimated using information about the set area based on information about the user's location and the attributes of the POIs identified by the attribute identification unit 14. Specifically, the determination unit 15 calculates, as the first information, information indicating the user's level of interest in the attributes of the POIs based on at least one or more pieces of second information, which are information including the stay area information and the attributes of the POIs identified by the attribute identification unit 14. Each piece of second information includes the stay area information and the attributes of the POIs identified by the attribute identification unit 14. The second information is, for example, combination information indicating a combination of the stay area information estimated by the stay area estimation unit 14a and specific POI information indicating the attributes of the POIs identified by the attribute identification unit 14. The stay area information estimated by the stay area estimation unit 14a and the specific POI information indicating the attributes of the POIs identified by the attribute identification unit 14 is, for example, a first score indicating the user's level of interest in the attributes of multiple POIs.

[0029] In the present embodiment, the determination unit 15 generates preparatory information for estimating a user's interests and tastes based on the stay area information estimated by the stay area estimation unit 14a and specific POI information indicating the attributes of the POIs identified by the attribute identification unit 14. Specifically, the determination unit 15 calculates, as the preparatory information, a first score indicating the user's degree of interest in the attributes of multiple POIs, based on combination information indicating a combination of the stay area information and the specific POI information. The preparatory information (first score) is, for example, the user's record of visits to a specific facility (e.g., the number of visits and the duration of stay).

[0030] 6 is a block diagram showing the functional configuration of the determination unit 15 of the hobby and taste estimation device 10 according to the first embodiment. In the first embodiment, the determination unit 15 includes a combination unit 15 a, a first vector conversion unit 15 b, a classification unit 15 c, and a calculation unit 15 d.

[0031] First, the combination unit 15a generates one or more pieces of combination information indicating combinations of stay area information and specific POI information. Fig. 7 is an example of data indicating combination information classified into multiple groups. In the example shown in Fig. 7, the combination information in the first and second rows indicates information combining "Japan" and "Spa / Hot Spring". The combination information in the third row indicates information combining "Japan" and "Park". The combination information in the fourth row indicates information combining "United States" and "Spa / Hot Spring". The combination information in the fifth and sixth rows indicates information combining "United States" and "Spa / Hot Spring".

[0032] The first vector conversion unit 15b converts the plurality of pieces of second information into a plurality of semantic vectors using a semantic estimation model that receives as input the stay area information and the second information including the attributes of the POI identified by the attribute identification unit 14 and outputs a semantic vector related to the meaning of the attributes of the POI. Specifically, the first vector conversion unit 15b converts the plurality of pieces of combination information into a plurality of semantic vectors using a semantic estimation model that receives as input the combination information and outputs a semantic vector indicating the meaning of the attributes of the POI.

[0033] The classification unit 15c classifies the plurality of pieces of second information into a plurality of groups. Specifically, the classification unit 15c classifies the plurality of pieces of combination information into a plurality of groups based on the distance between the plurality of semantic vectors. The semantic estimation model is a trained model constructed by the construction unit 17, which will be described later. The semantic vector indicates the meaning of the attribute of the POI for each set area. If the meanings of the attributes of the POIs in each piece of combination information are similar (for example, if the POIs are used in the same way), the distance between the semantic vectors generated from each piece of combination information will be short. If the meanings of the attributes of the POIs in each piece of combination information are distant (for example, if the POIs are used in different ways), the distance between the semantic vectors generated from each piece of combination information will be long.

[0034] The classification unit 15c determines whether or not to group multiple pieces of vector information together (determines whether to aggregate POI categories) based on the distance between multiple semantic vectors. Specifically, the classification unit 15c assigns an aggregation code to each of multiple pieces of combination information, and classifies one or more pieces of combination information assigned a predetermined aggregation code into the same group. In the example shown in FIG. 7 , the first vector conversion unit 15b converts information combining "Japan" and "Spa / Hot Springs," which are the combination information in the first and second rows, into a semantic vector. The first vector conversion unit 15b also converts information combining "United States" and "Spa / Hot Springs," which are the combination information in the fifth and sixth rows, into a semantic vector. If, as a result of comparing the semantic vectors, the distance between the semantic vectors is found to be large (for example, the distance between the semantic vectors exceeds a threshold), the classification unit 15c assigns different aggregation codes "1" and "3" to the attribute of the POI "Spa / Hot Spring" in Japan and the attribute of the POI "Spa / Hot Spring" in the United States, thereby classifying the information combining "United States" and "Spa / Hot Spring" and the information combining "Japan" and "Spa / Hot Spring" into different groups.

[0035] The first vector conversion unit 15b converts the information combining "Japan" and "park", which is the combination information in the third row, into a semantic vector. The first vector conversion unit 15b converts the information combining "United States" and "park", which is the combination information in the fourth row, into a semantic vector. If, as a result of comparing the semantic vectors, the distance between the semantic vectors is close (for example, if the distance between the semantic vectors is below a threshold), the classification unit 15c assigns the same aggregation code "2" to the attribute of the POI "park" in Japan and the attribute of the POI "park" in the United States, thereby classifying the information combining "United States" and "park" and the information combining "Japan" and "park" into the same group.

[0036] The calculation unit 15d calculates, for each group included in the plurality of groups, information indicating the user's degree of interest in the group based on at least one or more pieces of second information included in the group as first information that is information for estimating the user's hobbies and tastes. Specifically, for each group included in the plurality of groups, the calculation unit 15d calculates, as the first information, information regarding the time the user stayed at a place related to the position based on at least one or more pieces of third information corresponding to at least one or more pieces of second information included in the group. The third information is, for example, stay time information indicating the time the user stayed at a place related to the position.

[0037] In the present embodiment, the calculation unit 15 d calculates, as preparation information, a first score indicating the degree of interest of the user in the attribute of the POI based on the combination information. Specifically, for each group included in the multiple groups, the calculation unit 15 d calculates a first score indicating the degree of interest of the user in the group based on one or more pieces of combination information included in the group.

[0038] For example, the calculation unit 15d calculates the difference between the stay start time and the stay end time for each of one or more pieces of combination information included in a predetermined group, thereby calculating the stay time, which is the time during which the user stayed. In this case, the calculation unit 15d calculates the multiple first scores by aggregating the stay times corresponding to one or more pieces of combination information included in each group.

[0039] FIG. 8 illustrates an example of multiple first scores calculated for each group. In the example illustrated in FIGS. 7 and 8 , the calculation unit 15d calculates the difference between the stay start time and the stay end time as the stay time for each of two sets of combination information corresponding to the aggregation code “1” and the user “user_1.” The calculation unit 15d aggregates the stay times corresponding to the aggregation code “1” and the user “user_1.” The calculation unit 15d defines the aggregated stay times as “aggregated feature 1.” The calculation unit 15d performs similar aggregation for each user and each aggregation code, and calculates “aggregated feature 2” indicating the time spent in the park and “aggregated feature 3” indicating the time spent at a spa in the United States. In this way, the calculation unit 15d shapes the feature by aggregating the stay times for each user based on the aggregation code. Note that the aggregation method may be changed depending on the stay start time. The first score may be any numerical value indicating the degree of stay. The first score may be the number of stays or may be any other numerical value other than those mentioned above.

[0040] The estimation unit 16 estimates the user's hobbies and preferences based on first information, which is information for estimating the user's hobbies and preferences. Specifically, the estimation unit 16 estimates information about the attributes of the hobbies and preferences based on the information for estimating the user's hobbies and preferences, using an attribute estimation model that receives the information for estimating the user's hobbies and preferences as input and outputs information about the attributes of the hobbies and preferences.

[0041] In the present embodiment, the estimation unit 16 estimates the user's hobbies and preferences based on the preparation information generated by the determination unit 15. Specifically, the estimation unit 16 estimates the user's hobbies and preferences based on the first scores. More specifically, the estimation unit 16 calculates a second score for each attribute of the hobbies and preferences based on the preparation information using an attribute estimation model that receives the preparation information as input and outputs second scores related to the attributes of the hobbies and preferences. For example, the estimation unit 16 outputs the attribute of the hobbies and preferences with the highest second score as the attribute of the user's hobbies and preferences. The attribute estimation model is a learning model constructed by the construction unit 17 described below. The second score is a numerical value indicating the probability that the user has a predetermined attribute of the hobbies and preferences. In the first embodiment, the estimation unit 16 estimates the user's hobbies and preferences based on a plurality of first scores.

[0042] Part (a) of FIG. 9 is a diagram illustrating an example of calculation results of multiple second scores calculated for each hobby / interest attribute. In the example shown in part (a) of FIG. 9, the estimation unit 16 acquires an attribute estimation model corresponding to the hobby / interest attribute "like Japanese hot springs" from the construction unit 17. The estimation unit 16 inputs aggregate features 1 to 3 corresponding to user "user_1" into the attribute estimation model and obtains an "attribute score," which is a second score for the hobby / interest attribute "like Japanese hot springs," as an output. In the table shown in part (a) of FIG. 9, the attribute score is expressed as a real value between 0 and 1. The estimation unit 16 calculates second scores for other hobby / interest attributes for user "user_1" in the same manner as described above. If the second score for the hobby / interest attribute "like Japanese hot springs" is higher than the second scores for the other hobby / interest attributes, the estimation unit 16 outputs the hobby / interest attribute "like Japanese hot springs" as the hobby / interest attribute of user "user_1." The estimation unit 16 similarly estimates the hobbies and interests attributes of another user "user_2."

[0043] In the first embodiment, the construction unit 17 constructs a semantic estimation model and an attribute estimation model. Specifically, the construction unit 17 constructs a semantic estimation model that receives combination information of a set area and a POI attribute as input and outputs vector information indicating the meaning of the POI attribute. The semantic estimation model is, for example, a language model. The semantic estimation model understands the meaning of POI attributes in multiple regions with different cultural backgrounds. Examples of semantic estimation models include BERT (Bidirectional Encoder Representations from Transformers), GPT (chat-GPT), PaLM (Pathway Language Model), and LLaMa (large language model meta AI). The construction unit 17 may acquire the semantic estimation model from an external source, or may construct the semantic estimation model by fine-tuning a learning model acquired from an external source. Therefore, the semantic estimation model learns or additionally learns languages ​​or sentences from multiple regions.

[0044] The construction unit 17 constructs an attribute estimation model that receives the preparatory information as input and outputs a second score related to the hobby / interest attribute. Specifically, the construction unit 17 constructs the attribute estimation model using training data that receives the preparatory information as input and outputs the attribute correct answer information input from the input unit 11. The attribute correct answer information is information indicating whether or not the user has a predetermined hobby / interest attribute. For example, it is information indicating the user's hobby / interest attribute (e.g., information indicating the result of a questionnaire asking the user whether or not they like hot springs). Whether or not the user has a predetermined hobby / interest attribute may be indicated by a binary value (e.g., "1" if the user has the predetermined hobby / interest attribute, and "0" if the user does not have the predetermined hobby / interest attribute). Whether or not the user corresponds to one of multiple hobby / interest attributes may be indicated by a numerical value (e.g., if the user is a predetermined age, a numerical value indicating the age is indicated). Part (b) of FIG. 9 illustrates an example of attribute correct answer information. Examples of attribute estimation models include logistic regression, gradient boosting tree (LightGBM), and deep learning models. The construction unit 17 may acquire an attribute estimation model from outside, or may construct an attribute estimation model by fine-tuning a learning model acquired from outside.

[0045] In the example shown in part (b) of FIG. 9 , attribute correct answer information indicating whether users "user_3" and "user_4" have predetermined hobby / interest attributes is shown. The construction unit 17 constructs an attribute estimation model using the attribute correct answer information as training data. Then, the estimation unit 16 estimates the first scores of users "user_1" and "user_2" using the attribute estimation model constructed based on the attribute correct answer information of users "user_3" and "user_4" (see part (a) of FIG. 9 ).

[0046] The determining unit 15, the estimating unit 16, and the constructing unit 17 may be a single functional unit. In this case, the single functional unit executes learning of the semantic estimation model and the attribute estimation model, and estimation using the semantic estimation model and the attribute estimation model.

[0047] Next, a process for estimating a user's interests and preferences in the first embodiment will be described with reference to Fig. 10. Fig. 10 is a flowchart showing the process for estimating a user's interests and preferences. Before the processes of steps S01 to S04 are executed, the interest and preference estimation device 10 acquires user location information and sets POI attribute information and set area information in advance. First, information on the attributes of a POI (Point Of Interest) and information on a plurality of set areas are input into the interest and preference estimation device 10 (step S01: input step).

[0048] Next, the hobby and preference estimation device 10 specifies information on the attributes of the POI corresponding to the information on the user's location based on the information on the attributes of the POI (step S02). Specifically, the hobby and preference estimation device 10 specifies the attributes of the POI corresponding to the user's location based on the user's location information and POI attribute information. In the first embodiment, the hobby and preference estimation device 10 estimates stay area information indicating a set area corresponding to the user's location from the user's location information and set area information. The hobby and preference estimation device 10 specifies the attributes of the POI corresponding to the user's location based on the user's location information, stay area information, and POI attribute information.

[0049] Next, the hobby and taste estimation device 10 determines first information, which is information for estimating the user's hobby and taste, based on stay area information estimated using information about the set area based on information about the user's location and the attributes of the POIs identified in the attribute identification step (step S03). Specifically, the preparatory information is determined based on the set area and the attributes of the POIs. In the first embodiment, the hobby and taste estimation device 10 calculates, as the preparatory information, multiple first scores indicating the user's degree of interest in the attributes of multiple POIs, based on combination information indicating a combination of the stay area information and the identified POI information.

[0050] Finally, the hobby and taste estimation device 10 estimates the user's hobby and taste based on the first information (step S04). In the first embodiment, the hobby and taste estimation device 10 estimates the user's hobby and taste based on the plurality of first scores.

[0051] Next, the effects of the hobby and preference estimation device 10 according to the first embodiment will be described. First, in the past, when the location of a user whose hobby and preference is to be estimated changes, it was sometimes impossible to accurately estimate the user's hobby and preference based on the attributes of a POI. For example, in Japan, the POI attribute "SPA" refers to hot springs, public baths, and other places used by users of various ages and genders. However, in the United States, where there is no hot spring culture, it refers to beauty spots primarily used by women. Despite these differences in meaning, treating a visit to a "SPA" in Japan and a visit to a "SPA" in the United States as the same information may result in a deterioration in the accuracy of the hobby and preference estimation.

[0052] In addition, there are cases where the meanings of the attributes of POIs are the same even though the attributes of the POIs are different. In such cases, even though the meanings of the attributes of the two POIs are essentially the same, if the user's interests and tastes are estimated based on the attributes of each POI, there is a risk that different estimation results will be obtained.

[0053] In contrast, the hobby and taste estimation device 10 includes an input unit 11 that inputs information about the attributes of a POI and information about a plurality of set areas having different characteristics from each other, an attribute identification unit 14 that identifies the attributes of the POI according to information about the user's location based on the information about the attributes of the POI, and a determination unit 15 that determines first information for estimating the user's hobby and taste based on stay area information estimated using information about the set area based on the information about the user's location and the attributes of the POI identified by the attribute identification unit 14.

[0054] In the hobby and preference estimation device 10 and hobby and preference estimation method according to the present embodiment, the characteristics of multiple set areas are preset to be different from one another. Information for estimating a user's hobby and preference is then determined based on the set area corresponding to information related to the user's location and the attributes of POIs corresponding to the location of the information related to the user. With this configuration, information for estimating a user's hobby and preference is determined based on the attributes of POIs corresponding to the information related to the user's location, taking into account the characteristics (e.g., cultural background) of the set area corresponding to the information related to the user's location. For example, information for estimating a user's hobby and preference is determined based on the attribute of a POI corresponding to the user's location being "hot spring" considering that the user is located in Japan. Furthermore, information for estimating a user's hobby and preference is determined based on the attribute of a POI corresponding to the user's location being "spa" considering that the user is located in the United States. This enables a user's hobby and preference to be more accurately estimated, taking into account that the meaning of a POI attribute may differ depending on the user's location.

[0055] The hobby and taste estimation device 10 may further include an estimation unit 16 that estimates information about the user's hobby and taste based on first information for estimating the user's hobby and taste. With this configuration, the hobby and taste estimation device 10 can estimate the user's hobby and taste, taking into account that the meaning of the attribute of the POI may differ depending on the user's location.

[0056] The determination unit 15 may calculate, as information for estimating the user's interests and preferences, information indicating the user's level of interest in the attributes of the POIs, based on at least one or more pieces of second information that include the stay area information and the attributes of the POIs identified by the attribute identification unit 14. For example, the combination unit 15a may generate combination information indicating a combination of the stay area information and the identified POI information, and the calculation unit 15d may calculate, as preparation information, a first score indicating the user's level of interest in the attributes of the POIs, based on the combination information, and the estimation unit 16 may estimate the user's interests and preferences based on the first score. This makes it possible to more accurately estimate the user's interests and preferences, taking into account that the meaning of the attributes of the POIs may differ depending on the user's location.

[0057] The first vector conversion unit 15b may convert multiple pieces of second information into multiple semantic vectors using a semantic estimation model that inputs second information, which is information including stay area information and attributes of the POI identified by the attribute identification unit 14, and outputs semantic vectors related to the meaning of the attributes of the POI.The classification unit 15c may classify the multiple pieces of second information into multiple groups based on the distance between the multiple semantic vectors.The calculation unit 15d may calculate, for each group included in the multiple groups, information indicating the user's degree of interest in the group based on at least one or more pieces of second information included in the group as first information for estimating the user's hobbies and tastes. For example, the combination unit 15a may generate a plurality of pieces of combination information, the first vector conversion unit 15b may convert each of the plurality of pieces of combination information into a plurality of semantic vectors using a semantic estimation model that takes the combination information as input and outputs a semantic vector indicating the meaning of the attribute of the POI, the classification unit 15c may classify the plurality of pieces of combination information into a plurality of groups based on the distance between the plurality of semantic vectors, and the calculation unit 15d may calculate, for each group included in the plurality of groups, a first score that indicates the degree of interest of the user in the group based on one or more pieces of combination information included in the group.

[0058] In this way, multiple pieces of combination information (multiple pieces of second information) are classified into multiple groups based on the distances between multiple semantic vectors. For each group included in the multiple groups, a first score (information indicating the user's level of interest in the group) is calculated based on one or more pieces of combination information (at least one or more pieces of second information included in the group) included in the group. This classifies multiple pieces of combination information based on the meanings of the POI attributes for each set area. For example, if the meanings of the POI attributes in multiple pieces of combination information are similar to each other, the multiple pieces of combination information are classified into the same group. If the meanings of the POI attributes in multiple pieces of combination information are different from each other, the multiple pieces of combination information are classified into different groups. Then, a first score is calculated for each group. Therefore, the first score is calculated taking into account the fact that the meanings of the POI attributes may differ depending on the user's location. As a result, the user's hobbies and preferences can be more accurately estimated by taking into account the fact that the meanings of the POI attributes may differ depending on the user's location.

[0059] In the hobby and taste estimation device 10, the second information may be associated with third information, which is information regarding the time the user stayed at a location related to the location. The calculation unit 15d may calculate, for each group included in the multiple groups, information regarding the time the user stayed at a location related to the location as the first information, based on at least one or more pieces of third information associated with at least one or more pieces of second information included in the group. For example, the user's location information may be associated with time information indicating a stay start time, which is the time the user started staying, and a stay end time, which is the time the user ended staying. The calculation unit 15d may calculate the stay time, which is the time the user stayed, by calculating the difference between the stay start time and the stay end time for each piece of combination information. The calculation unit 15d may calculate, for each group included in the multiple groups, multiple first scores by aggregating the stay times corresponding to one or more pieces of combination information included in the group. The first score for each group is calculated by aggregating, for each group, the time the user stayed at a location corresponding to each piece of combination information. According to this configuration, the degree of interest of the user in each group is evaluated as the time spent at a position corresponding to each group, and the first score corresponding to each group is calculated more accurately, thereby making it possible to more accurately estimate the user's hobbies and preferences based on the multiple first scores.

[0060] The estimation unit 16 receives first information for estimating the user's hobbies and preferences as input, and estimates information related to the hobbies and preferences attributes based on the first information using an attribute estimation model that outputs information related to the hobbies and preferences attributes. For example, the estimation unit 16 may receive preparatory information as input, and use an attribute estimation model that outputs second scores related to the hobbies and preferences attributes to calculate second scores for each of the hobbies and preferences attributes based on the preparatory information. In this manner, information related to the hobbies and preferences attributes is calculated. This makes it possible to evaluate the possibility that the user has a specific hobbies and preferences attribute for each hobbies and preferences attribute. For example, it is possible to calculate a second score as a numerical value indicating the probability that the user has a specific hobbies and preferences attribute. In such a case, it is possible to estimate the hobbies and preferences attribute with the highest second score as the user's hobbies and preferences. From the above, the user's hobbies and preferences can be estimated more accurately.

[0061] The input unit 11 may input information about a plurality of set areas having different cultural backgrounds. In such a case, the user's tastes and preferences can be estimated more accurately by taking into consideration that the meaning of the attributes of the POI differs depending on the cultural background of each set area.

[0062] The input unit 11 may input information about a plurality of set areas corresponding to a plurality of countries, respectively. In such a case, the user's interests and tastes can be more accurately estimated by taking into consideration that the meaning of the attributes of the POI may differ from country to country.

[0063] Furthermore, the hobby and preference estimation device 10 can estimate a user's hobby and preference based at least on the user's location in each of multiple set areas. This allows the construction unit 17 to construct an attribute estimation model that uses preparatory information generated using at least location information included in one of multiple countries or regions as input and outputs second scores related to hobby and preference attributes. Because such an attribute estimation model is trained using preparatory information related to multiple countries or regions and second scores related to hobby and preference attributes, it is possible to calculate second scores with higher accuracy for more countries or regions. For example, if a user is located in a region where correct attribute information (correct attribute information for hobby, preference, and gender / age attributes) is not available, the user's hobby and preference can be accurately estimated by using an attribute estimation model trained using correct attribute information for other regions. As an example, a case will be described in which POI attribute information, set area information, and correct attribute information are acquired in Japan, and POI attribute information and set area information are acquired in the United States. In this case, when it is desired to infer attributes for data from the United States, the above-described attribute estimation model can be constructed using only correct attribute information for Japan, and the constructed attribute estimation model can be applied to estimating hobbies and preferences in the United States. This makes it possible to estimate information about the attributes of a user's hobbies and preferences when the user is located in the United States, a region where correct attribute information is not available. As a result, by using this attribute estimation model, it becomes possible to provide more accurate marketing tailored to the needs of users in more countries or regions.

[0064] Second Embodiment Next, a hobby and taste estimation device 10 according to a second embodiment will be described with reference to Fig. 2 and Fig. 11 to Fig. 15. In the second embodiment, the description common to the first embodiment will be omitted, and differences from the first embodiment will be mainly described.

[0065] FIG. 11 is a diagram illustrating an overview of processing in the hobby and preference estimation device 10 according to the second embodiment. In the example illustrated in FIG. 11 , the hobby and preference estimation device 10 acquires location information of a user, information indicating that the user stayed in the United States, and information indicating that the attributes of the POIs visited by the user are "park" and "spa." The hobby and preference estimation device 10 generates a sentence, using English, a language corresponding to the United States, indicating that the user visited POIs whose attributes are "park" and "spa." The hobby and preference estimation device 10 acquires location information of the user, information indicating that the user stayed in Japan, and information indicating that the attributes of the POIs visited by the user are "hot spring / spa" and "park." The hobby and preference estimation device 10 generates a sentence, using Japanese, a language corresponding to Japan, indicating that the user visited POIs whose attributes are "hot spring" and "park." The hobby and preference estimation device 10 inputs preparation information based on each generated sentence into an estimation model, thereby estimating the user's hobby and preference.

[0066] 2 is also a block diagram showing a hobby and taste estimation device according to a second embodiment. Similar to the first embodiment, the hobby and taste estimation device 10 of the second embodiment includes an input unit 11, a setting unit 12, a storage unit 13, an attribute identification unit 14, a determination unit 15, an estimation unit 16, and a construction unit 17. The input unit 11, the setting unit 12, and the attribute identification unit 14 have the same functions as those of the first embodiment.

[0067] The storage unit 13 stores POI attribute information and set area information. In the second embodiment, the set area information is associated with language information indicating a language used in the set area. For example, if the set area corresponds to Japan, the language information includes information indicating that Japanese is associated with the set area.

[0068] The determination unit 15 generates preparatory information for estimating a user's interests and tastes based on the stay area information estimated by the stay area estimation unit 14a and specific POI information indicating the attributes of the POI identified by the attribute identification unit 14. The stay area information is associated with language information used in the area indicated by the stay area information. For example, the stay area information is associated with language information indicating the language used in the set area indicated by the stay area information. FIG. 12 is a block diagram showing the functional configuration of the determination unit 15 of the interest and taste estimation device 10 according to the second embodiment. In the second embodiment, the determination unit 15 includes a combination unit 15a, a character string formation unit 18b, and a second vector conversion unit 18c. As in the first embodiment, the combination unit 15a generates one or more combination information indicating a combination of stay area information and specific POI information.

[0069] The character string forming unit 18b converts the second information, which is information including the stay area information and the attributes of the POIs identified by the attribute identifying unit 14, into information regarding a character string corresponding to the language information corresponding to the stay area information included in the second information. Specifically, the character string forming unit 18b generates character string information indicating a character string related to the combination information using a language corresponding to the stay area information of the combination information. For example, the determining unit 15 may generate character string information by inserting names indicating the attributes of the POIs in the combination information into a pre-prepared template. When time information is associated with the combination information, the determining unit 15 may generate, as character string information, a sentence in which the names indicating the attributes of the POIs are arranged in chronological order of the user's visits. Such character string information may include a timestamp indicating the time the user visited.

[0070] Sections (a) to (d) of FIG. 13 are diagrams illustrating an example of data of character string information indicating a character string related to combination information. In the example shown in FIGS. 3 and section (a) of FIG. 13, the character string forming unit 18b generates sentences describing time information, a setting area corresponding to the user's location, POI attributes, and the name of the POI using a language (e.g., Japanese) corresponding to the setting area based on data corresponding to the user "user_1" shown in FIG. 3. The character string forming unit 18b generates the sentence "2023 / 10 / 01 12:00-13:00 Went to XX Hot Spring at SPA / Hot Spring in Tokyo, Japan" based on the data in the first line shown in FIG. 3. The character string forming unit 18b generates the sentence "2023 / 10 / 01 14:00-18:00 Went to XX Hot Spring at SPA / Hot Spring in Tokyo, Japan" based on the data in the second line shown in FIG. 3. The character string forming unit 18b regards the two generated sentences as character string information corresponding to the user "user_1." In the examples shown in Figures 3 and 13 (b) to (d), the string formation unit 18b generates sentences based on data corresponding to users "user_2" to "user_4", as described above, and generates string information corresponding to users "user_2" to "user_4".

[0071] The second vector conversion unit 18c receives information about a character string as input and converts the information about the character string into a vector related to the character string using a vector estimation model that outputs a vector related to the character string, and uses the vector related to the character string as first information for estimating the user's interests and preferences. Specifically, the second vector conversion unit 18c receives string information as input and converts the character string into a vector using a vector estimation model that outputs vector information indicating the character string. For example, the vector is an embedding vector of the character string indicated by the string information (embedding vector of a sentence). The vector estimation model is a trained model constructed by the construction unit 17. FIG. 14 is a diagram illustrating an example of a vector based on character string information. In the example illustrated in FIG. 14, the determination unit 15 inputs the four pieces of character string information shown in parts (a) to (d) of FIG. 13 and other character string information into the vector estimation model, thereby obtaining multiple vector information each having n components as output. The number of dimensions n of the vector depends on the vector estimation model and is, for example, 782.

[0072] The estimation unit 16 estimates the user's hobbies and preferences based on the preparation information generated by the determination unit 15. In the second embodiment, the estimation unit 16 estimates the user's hobbies and preferences according to the vectors generated by the determination unit 15. Specifically, the estimation unit 16 uses an attribute estimation model that receives as input the vectors generated by the second vector conversion unit 18c of the determination unit 15 and outputs second scores related to the hobbies and preferences attributes, and calculates second scores for each of the hobbies and preferences attributes based on the vectors.

[0073] In the second embodiment, the construction unit 17 constructs a vector estimation model and an attribute estimation model. Specifically, the construction unit 17 constructs a vector estimation model that receives character string information as input and outputs a vector. The vector estimation model is, for example, a language model. As an example of a semantic estimation model, there are BERT (Bidirectional Encoder Representations from Transformers), GPT (chat-GPT), PaLM (Pathway Language Model), LLaMa (large language model meta AI), etc. The construction unit 17 may acquire the vector estimation model from an external source, or may construct the vector estimation model by fine-tuning a learning model acquired from an external source.

[0074] Section (a) of FIG. 15 is a diagram illustrating an example of attribute correct answer information. Section (b) of FIG. 15 is a diagram illustrating an example of a second score calculated according to a vector. The construction unit 17 constructs an attribute estimation model using training data in which the vector information of users "user_2" to "user_100" shown in FIG. 14 is input and the attribute correct answer information of users "user_2" to "user_100" shown in section (a) of FIG. 15 is output. In the example shown in section (b) of FIG. 15, the estimation unit 16 acquires an attribute estimation model corresponding to the hobby / interest attribute of "I like Japanese hot springs" from the construction unit 17. The estimation unit 16 inputs vector information corresponding to user "user_1" shown in FIG. 14 into the attribute estimation model, thereby obtaining a second score, which is the "attribute score" of user "user_1," as an output. In this way, the attributes of user "user_1", whose attributes are unknown, are estimated using an attribute estimation model constructed using the vector information and attribute correct answer information of users "user_2" to "user_100", whose attributes are known.

[0075] Next, a process for estimating interests and preferences in the second embodiment will be described with reference to Fig. 10. Information about the attributes of POIs (Points of Interest) and information about multiple set areas are input to the interest and preference estimation device 10 (Step S01: input step). Subsequently, information about the attributes of the POIs corresponding to information about the user's location is identified based on the information about the POI attributes (Step S02). In the second embodiment, the set area information is associated with language information indicating a language corresponding to the set area.

[0076] Next, first information for estimating the user's interests and tastes is determined based on stay area information estimated using information about the set area based on information about the user's location and the attributes of the POI identified in the attribute identification step (step S03). In the second embodiment, the interest and taste estimation device 10 generates combination information indicating a combination of stay area information and specific POI information, generates string information indicating a string related to the combination information using a language corresponding to the stay area information of the combination information, and converts the string into a vector using a vector estimation model that receives the string information as input and outputs vector information indicating the string. The stay area information is associated with language information indicating a language corresponding to the set area indicated by the stay area information.

[0077] Finally, the hobby and taste estimation device 10 estimates the user's hobby and taste based on the preparation information (step S04). In the second embodiment, the hobby and taste estimation device 10 estimates the user's hobby and taste based on the vector generated in step S03.

[0078] Next, the effects of the hobby and taste estimation device 10 according to the second embodiment will be described. First, in the second embodiment, as in the first embodiment, it is possible to estimate the hobby and taste of a user by taking into consideration that the meaning of the attributes of a POI may differ depending on the location of the user.

[0079] In the hobby and taste estimation device 10, the stay area information is associated with language information used in the area indicated by the stay area information, and the string formation unit 18b may convert second information including the stay area information and the attributes of the POI identified by the attribute identification unit 14 into information about a string according to the language information corresponding to the stay area information included in the second information, and the second vector conversion unit 18c may convert information about the string into a vector about the string using a vector estimation model that takes information about the string as input and outputs a vector about the string, and may use the vector about the string as first information for estimating the user's hobby and taste. For example, the stay area information may be associated with language information indicating a language used in the set area indicated by the stay area information. The combination unit 15a may generate combination information indicating a combination of the stay area information and the specific POI information. The character string formation unit 18b may generate character string information indicating a character string related to the combination information using a language corresponding to the stay area information of the combination information. The second vector conversion unit 18c may convert the character string into a vector using a vector estimation model that receives the character string information as input and outputs vector information indicating the character string. The estimation unit 16 may estimate the user's interests and preferences based on the vector. In this case, the combination information of the stay area information and the specific POI information is generated, character string information indicating a character string related to the combination information is generated using a language corresponding to the set area, the character string information is converted into a vector, and the user's interests and preferences are estimated based on the vector. With this configuration, the character string information is generated taking into account the cultural background of each set area, so that the meaning or context of the character string indicated by the character string information reflects the meaning of the POI attributes for each set area. This allows the user's interests and preferences to be more accurately estimated based on the vector converted from the character string information.

[0080] The hobby and taste estimation device 10 may generate character string information from a combination of a set area corresponding to the user's location, attributes of POIs corresponding to the user's location, and time information. By inputting text into the vector estimation model, a semantic vector is generated that takes into account differences in set areas and the time series of the user's visits to each POI. This allows the semantic vector to be used as input into the attribute estimation model, thereby enabling the user's hobby and taste to be estimated more accurately.

[0081] Next, the hardware configuration of the hobby and taste estimation device 10 described above will be described with reference to Fig. 16. The hobby and taste estimation device 10 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

[0082] In the following description, the term "device" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the hobby and taste estimation device 10 may be configured to include one or more of the devices shown in the drawings, or may be configured to exclude some of the devices.

[0083] Each function of the hobby and taste estimation device 10 is realized by loading specific software (programs) onto hardware such as a processor 1001 and a memory 1002, causing the processor 1001 to perform calculations and control communication via a communication device 1004 and the reading and / or writing of data in the memory 1002 and storage 1003.

[0084] The processor 1001 controls the entire computer by running, for example, an operating system. The processor 1001 may be configured as a central processing unit (CPU) including an interface with peripheral devices, a control unit, an arithmetic unit, a register, etc. For example, the control function of the input unit 11, etc. may be realized by the processor 1001.

[0085] The processor 1001 also reads programs (program codes), software modules, and data from the storage 1003 and / or the communication device 1004 into the memory 1002, and executes various processes in accordance with these. The programs used are those that cause a computer to execute at least some of the operations described in the above embodiments.

[0086] For example, the control functions of the determination unit 15 and the like may be realized by a control program stored in the memory 1002 and running on the processor 1001, and similar functions may be realized for other functional blocks. Although the above-described various processes have been described as being executed by one processor 1001, they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented on one or more chips. The program may be transmitted from a network via a telecommunications line.

[0087] The memory 1002 is a computer-readable recording medium and may be composed of at least one of, for example, a read-only memory (ROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a random access memory (RAM), etc. The memory 1002 may also be called a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store executable programs (program codes), software modules, etc. for implementing a wireless communication method according to one embodiment of the present invention.

[0088] Storage 1003 is a computer-readable recording medium, and may be, for example, at least one of an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray disc), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy disk, a magnetic strip, etc. Storage 1003 may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other appropriate medium including memory 1002 and / or storage 1003.

[0089] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via a wired and / or wireless network, and is also called, for example, a network device, a network controller, a network card, or a communication module.

[0090] The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that outputs to the outside. The input device 1005 and the output device 1006 may be integrated into one device (e.g., a touch panel).

[0091] Furthermore, each device such as the processor 1001 and the memory 1002 is connected to a bus 1007 for communicating information. The bus 1007 may be configured as a single bus, or may be configured as different buses between the devices.

[0092] The hobby and taste estimation device 10 may also be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 1001 may be implemented by at least one of these pieces of hardware.

[0093] Although the present embodiment has been described in detail above, it is clear to those skilled in the art that the present embodiment is not limited to the embodiment described in this specification. The present embodiment can be implemented in modified and altered forms without departing from the spirit and scope of the present invention as defined by the claims. Therefore, the description in this specification is intended to be illustrative and does not have any limiting meaning on the present embodiment.

[0094] Each aspect / embodiment described herein may be applied to systems utilizing LTE (Long Term Evolution), LTE-Advanced (LTE-A), SUPER 3G, IMT-Advanced, 4G, 5G, FRA (Future Radio Access), W-CDMA, GSM, CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB (Ultra-Wide Band), Bluetooth, or other suitable systems and / or next generation systems enhanced thereon.

[0095] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described herein may be rearranged unless it is consistent. For example, the methods described herein present elements of various steps in an example order, and are not limited to the particular order presented.

[0096] Input and output information may be stored in a specific location (for example, memory) or managed in a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be sent to another device.

[0097] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).

[0098] The aspects / embodiments described herein may be used alone or in combination, or may be switched depending on the implementation. Notification of predetermined information (e.g., notification that "X is true") is not limited to explicit notification, but may be implicit (e.g., not notifying the predetermined information).

[0099] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0100] Software, instructions, etc. may also be transmitted or received over a transmission medium. For example, if the software is transmitted from a website, server, or other remote source using wired technologies such as coaxial cable, fiber optic cable, twisted pair, and Digital Subscriber Line (DSL), and / or wireless technologies such as infrared, radio, and microwave, these wired and / or wireless technologies are included within the definition of transmission media.

[0101] The information, signals, etc. described herein may be represented using any one of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0102] It should be noted that terms explained in this specification and / or terms necessary for understanding this specification may be replaced with terms having the same or similar meanings.

[0103] Furthermore, the information, parameters, etc. described in this specification may be expressed as absolute values, as relative values ​​from a predetermined value, or as corresponding other information.

[0104] A communications terminal may also be referred to by those skilled in the art as a mobile communications terminal, subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communications device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or some other suitable terminology.

[0105] As used herein, the phrase "based on" does not mean "based only on," unless expressly specified otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0106] When designations such as "first," "second," etc. are used herein, any reference to such elements does not generally limit the quantity or order of those elements. These designations may be used herein as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed therein or that the first element must precede the second element in some way.

[0107] To the extent that the terms "include," "including," and variations thereof are used herein or in the claims, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, the term "or," as used herein or in the claims, is not intended to be an exclusive or.

[0108] In this specification, a plurality of devices is also included unless the context or the technology clearly indicates that only one device exists.

[0109] Throughout this disclosure, the plural is intended to be included unless the singular is clearly indicated by the context.

[0110] Finally, various exemplary embodiments included in the present disclosure are described below in [E1] to [E10].

[0111] [E1] A hobby and taste estimation device comprising: an input unit that inputs information on attributes of a POI (Point Of Interest) and information on a plurality of set areas having mutually different characteristics; an attribute identification unit that identifies an attribute of a POI according to information on a user's location based on the information on the attribute of the POI; and a determination unit that determines first information, which is information for estimating the hobby and taste of the user, based on stay area information estimated using information on the set area based on the information on the user's location and the attribute of the POI identified by the attribute identification unit.

[0112] [E2] The hobby and taste estimation device according to [E1], further comprising an estimation unit that estimates information related to the hobby and taste of the user based on the first information.

[0113] [E3] The hobby and taste estimation device described in [E2], wherein the determination unit calculates, as the first information, information indicating the degree of interest of the user in the attributes of the POI based on at least one or more pieces of second information that include the stay area information and the attributes of the POI identified by the attribute identification unit.

[0114] [E4] The hobby and taste estimation device described in [E3], wherein the determination unit includes a first vector conversion unit, a classification unit, and a calculation unit, wherein the first vector conversion unit converts a plurality of pieces of second information into a plurality of semantic vectors using a semantic estimation model that receives the second information as input and outputs semantic vectors related to the meaning of an attribute of the POI, the classification unit classifies the plurality of pieces of second information into a plurality of groups based on distances between the plurality of semantic vectors, and the calculation unit calculates, for each group included in the plurality of groups, information indicating the degree of interest of the user in the group as the first information based on at least one or more pieces of second information included in the group.

[0115] [E5] The hobby and taste estimation device described in [E4], wherein the second information is associated with third information which is information relating to a time when the user stayed at a place related to the position, and the calculation unit calculates, for each group included in the plurality of groups, information relating to a time when the user stayed at a place related to the position as the first information based on at least one or more pieces of third information corresponding to at least one or more pieces of second information included in the group.

[0116] [E6] The hobby and taste estimation device described in [E2], wherein the stay area information is associated with language information used in the area indicated by the stay area information, the determination unit has a character string formation unit and a second vector conversion unit, the character string formation unit converts second information, which is information including the stay area information and the attribute of the POI identified by the attribute identification unit, into information about a character string according to the language information corresponding to the stay area information included in the second information, and the second vector conversion unit converts information about the character string into a vector about the character string using a vector estimation model that receives information about the character string as an input and outputs a vector about the character string, and the vector about the character string is used as the first information.

[0117] [E7] The hobby and taste estimation device described in any one of [E2] to [E6], wherein the estimation unit estimates information regarding the attributes of the hobby and taste based on the information for estimating the hobby and taste of the user, using an attribute estimation model that receives information for estimating the hobby and taste of the user as input and outputs information regarding the attributes of the hobby and taste.

[0118] [E8] The hobby and taste estimation device according to any one of [E1] to [E7], wherein the input unit inputs information about a plurality of set areas having different cultural backgrounds.

[0119] [E9] The hobby and taste estimation device according to any one of [E1] to [E7], wherein the input unit inputs information about the set areas, each of the set areas corresponding to a plurality of countries.

[0120] [E10] A hobby and taste estimation method comprising: an input step of inputting information on attributes of a POI (Point Of Interest) and information on a plurality of set areas having mutually different characteristics; an attribute identification step of identifying attributes of the POI according to information on the user's location based on the information on the attributes of the POI; and a determination step of determining first information, which is information for estimating the hobby and taste of the user, based on stay area information estimated using information on the set area based on the information on the user's location and the attributes of the POI identified in the attribute identification step.

[0121] 11...input unit, 14...attribute identification unit, 15...determination unit, 15a...combination unit, 15b...first vector conversion unit, 15c...classification unit, 15d...calculation unit, 16...estimation unit, 18b...character string formation unit, 18c...second vector conversion unit, 1001...processor, 1002...memory, 1003...storage, 1004...communication device, 1005...input device, 1006...output device.

Claims

1. An input unit that inputs information regarding the attributes of a POI (Point Of Interest) and information regarding a plurality of setting areas having different characteristics from each other; an attribute specifying unit that specifies the attributes of the POI according to the information regarding the user's position based on the information regarding the attributes of the POI; and a determination unit that determines first information, which is information for estimating the user's hobbies and interests, based on the residence area information estimated using the information regarding the setting areas based on the information regarding the user's position and the attributes of the POI specified by the attribute specifying unit. A hobby and interest estimation device.

2. The hobby and interest estimation device according to claim 1, further comprising an estimation unit that estimates information regarding the user's hobbies and interests based on the first information.

3. The determination unit calculates, as the first information, information indicating the degree of the user's interest in the attributes of the POI based on at least one or more pieces of second information, which is information including the residence area information and the attributes of the POI specified by the attribute specifying unit. The hobby and interest estimation device according to claim 2.

4. The determination unit includes a first vector conversion unit, a classification unit, and a calculation unit. The first vector conversion unit converts a plurality of pieces of the second information into a plurality of semantic vectors using a semantic estimation model that outputs a semantic vector regarding the meaning of the attributes of the POI with the second information as an input. The classification unit classifies the plurality of pieces of the second information into a plurality of groups based on the distances between the plurality of semantic vectors. The calculation unit calculates, as the first information, information indicating the degree of the user's interest in each group based on at least one or more pieces of the second information included in each group for each group included in the plurality of groups. The hobby and interest estimation device according to claim 3.

5. The second information is associated with third information, which is information regarding the time when the user stayed at the location related to the position. The calculation unit calculates, as the first information, information regarding the time when the user stayed at the location related to the position based on at least one or more pieces of the third information corresponding to at least one or more pieces of the second information included in each group for each group included in the plurality of groups. The hobby and interest estimation device according to claim 4.

6. The residence area information is associated with language information used in the area indicated by the residence area information. The determination unit includes a character string formation unit and a second vector conversion unit. The character string formation unit converts second information, which is information including the residence area information and the attribute of the POI specified by the attribute specifying unit, into information regarding a character string corresponding to the language information corresponding to the residence area information included in the second information. The second vector conversion unit uses a vector estimation model that takes the information regarding the character string as an input and outputs a vector regarding the character string to convert the information regarding the character string into a vector regarding the character string, and sets the vector regarding the character string as the first information. The hobby and interest estimation device according to claim 2.

7. The estimation unit uses an attribute estimation model that takes the first information as an input and outputs information regarding the attribute of the hobby and interest to estimate information regarding the attribute of the hobby and interest based on the first information. The hobby and interest estimation device according to claim 2.

8. The input unit inputs information regarding a plurality of setting areas having different cultural backgrounds. The hobby and interest estimation device according to claim 1.

9. The input unit inputs information regarding the setting areas where the plurality of setting areas respectively correspond to a plurality of countries. The hobby and interest estimation device according to claim 1.

10. An input step of inputting information regarding the attribute of a POI (Point Of Interest) and information regarding a plurality of setting areas having different characteristics; an attribute specifying step of specifying the attribute of the POI corresponding to the information regarding the user's position based on the information regarding the attribute of the POI; and a determination step of determining first information, which is information for estimating the hobby and interest of the user, based on the residence area information estimated using the information regarding the setting area based on the information regarding the user's position and the attribute of the POI specified in the attribute specifying step. A hobby and interest estimation method comprising the steps.

Images