Search device, search method, and search program

Abstract

To provide a retrieval device, retrieval method, and retrieval program, which enable more accurately retrieving a location from a retrieval query entered by a user.SOLUTION: In a retrieval device 100, a control unit 130 comprises: an acquisition unit 131 for acquiring a retrieval query entered by a user and a latitude and longitude indicative of a current location of the user; a conversion unit 132 for converting the latitude and longitude into an address; a coupling unit 133 for coupling the retrieval query and the address into one character string; and a retrieval unit 134 configured to input the character string to a machine learning model to retrieve a location and output the retrieved location, the machine learning model designed to output location information corresponding to a retrieval query.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a search device, a search method, and a search program.

Background Art

[0002] A user can access a search site or an application using a smartphone or the like and search for information on various bases such as landmarks, convenience stores, hotels, and the like. In such a base search, for example, a user inputs a search query such as the name of a base into an input form of a search site or the like via a smartphone or the like, and the base is searched using a DNN (Deep Neural Network) based on the search query.

[0003] Also, in base search, the base indicated by the search query input by the user may change depending on the user's current location. Therefore, for example, it is conceivable to acquire location information from a smartphone or the like used by the user and perform a search for a base that the user more desires using the search query and the location information.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, location information obtained from smartphones and other devices is latitude and longitude information indicating the current location, obtained using the GPS (Global Positioning System) function of the smartphone or other device. Furthermore, DNNs used for location searches are good at interpreting natural language but are not good at interpreting numerical sequences such as latitude and longitude. Therefore, it is difficult to correctly search for locations using latitude and longitude information (at least, correctly searching would require a very large machine learning model and training data, which is not easy).

[0006] This application was made in view of the above, and aims to enable more accurate searching of locations from search queries entered by users. [Means for solving the problem]

[0007] The search device according to the present invention is characterized by comprising: an acquisition unit that acquires a search query entered by the user and the latitude and longitude indicating the user's current location; a conversion unit that converts the latitude and longitude into an address; a concatenation unit that combines the search query and the address into a single string; and a search unit that inputs the string into a machine learning model that outputs location information from the search query, searches for locations, and outputs the results. [Effects of the Invention]

[0008] According to one aspect of this embodiment, the ability to more accurately search for locations based on search queries entered by the user is achieved. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 shows an example of a problem with location search. [Figure 2] Figure 2 shows an example of the configuration of the information processing system 1 according to this embodiment. [Figure 3] Figure 3 shows an example of a location search according to this embodiment. [Figure 4]Figure 4 shows an example of the configuration of the search device 100 according to this embodiment. [Figure 5] Figure 5 shows an example of information stored in the input data storage unit 121 according to this embodiment. [Figure 6] Figure 6 shows an example of training data according to this embodiment. [Figure 7] Figure 7 is a flowchart showing an example of the procedure for the location search process according to this embodiment. [Figure 8] Figure 8 is a flowchart showing an example of the training process procedure according to this embodiment. [Figure 9] Figure 9 is a hardware configuration diagram showing an example of a computer that implements the functions of the search device 100. [Modes for carrying out the invention]

[0010] The following describes in detail, with reference to the drawings, the forms (hereinafter referred to as "embodiments") for implementing the search device, search method, and search program according to the present application. Note that these embodiments do not limit the search device, search method, and search program according to the present application. Furthermore, the same parts are denoted by the same reference numerals in each of the following embodiments, and redundant descriptions are omitted.

[0011] [0. Problem] First, let's explain an example of location search using Figure 1. When a user searches for a location such as a convenience store, they access a search site using a smartphone or similar device and enter a search query into the input form, and the search returns information on the locations that match the search. Figure 1 is a diagram illustrating an example of a problem with location search. However, as shown in Figure 1, if the search query is only the name of the location, the user's current location may be required to find and return the location the user is looking for. For example, as shown in Figure 1, if the search query is "XXX Mart", a user currently at Shinjuku Station is likely to want to find the nearest "XXX Mart Shinjuku Station West Exit Store", and a user currently at Kokusai-dori Shopping Street is likely to want to find the nearest "XXX Mart Naha Kokusai-dori Store". For example, if a user currently at Shinjuku Station wants to get information on "XXX Mart Naha Kokusai-dori Store", they would likely enter "XXX Mart Naha Kokusai-dori Store" as the search query. Therefore, it is believed that a more accurate location search can be achieved by considering not only the search query but also the user's current location during the location search.

[0012] However, since the location information obtained from smartphones and other devices is a numerical sequence representing latitude and longitude, in this embodiment, this numerical sequence is converted into a string (address) suitable for the DNN used for location search, enabling searches in natural language. Also, in cases where the search query is "Tokyo Tower," for example, it is not necessary to consider the user's current location (it is clear that "Tokyo Tower" referred to in the search query is the only Tokyo Tower, regardless of where the user is). Therefore, in this embodiment, the machine learning model is trained so that the DNN can independently determine whether or not to use the current location when performing a location search.

[0013] [1. Configuration of the Information Processing System] The information processing realized by the search device 100 and other components of this embodiment will be explained using Figure 2. Figure 2 is a diagram showing an example configuration of the information processing system 1 according to this embodiment. In the example shown in Figure 2, the information processing system 1 has a user terminal 10 and a search device 100. The user terminal 10 and the search device 100 are connected to each other via a predetermined communication network (not shown) by wire or wireless connection. Network N is a communication network such as a LAN (Local Area Network), WAN (Wide Area Network), telephone network (mobile phone network, fixed telephone network, etc.), regional IP (Internet Protocol) network, or the Internet. Network N may include a wired network or a wireless network. Furthermore, the information processing system 1 shown in Figure 2 may include multiple user terminals 10 and multiple search devices 100.

[0014] The user terminal 10 shown in Figure 2 is an information processing terminal used by a user performing a location search. For example, the user terminal 10 could be a desktop PC (Personal Computer), a notebook PC, a tablet device, a mobile phone, or a PDA (Personal Digital Assistant). In the example shown in Figure 2, the user terminal 10 is shown as a smart device such as a smartphone or tablet used by the user.

[0015] The user terminal 10 shown in Figure 2 displays, for example, a search site or application for location searching provided by a search device 100, and transmits the search query entered by the user into the input form of the search site or application to the search device 100. The input of the search query into the input form may be done not only by text, but also by voice input. The user terminal 10 also transmits, for example, latitude and longitude information indicating the current location, obtained using GPS functionality, to the search device 100 along with the search query. Furthermore, the user terminal 10 receives and displays the search results for locations from the search device 100.

[0016] Further, the search device 100 shown in FIG. 2 is an information processing device managed by a service provider that provides a base search service or the like. For example, the search device 100 is a server computer or the like. Alternatively, the search device 100 may be a cloud computer device managed by a service provider that provides a cloud computing service.

[0017] The search device 100 shown in FIG. 2 receives and obtains, for example, a search query input by a user and the latitude and longitude indicating the user's current location from the user terminal 10. Further, the search device 100, for example, converts the obtained latitude and longitude into an address. Further, the search device 100, for example, combines the obtained search query and the converted address into one character string. Then, the search device 100, for example, inputs the combined character string into a machine learning model that outputs base information from the search query, searches for a base, and outputs a search result to the user terminal 10. The base search of the present embodiment will be described more specifically with reference to FIG. 3.

[0018] FIG. 3 is a diagram showing an example of the base search according to the present embodiment. As shown in FIG. 3, the search device 100 uses, for example, a query DNN, which is a machine learning model that outputs a query vector, which is vector data of a base, from a search query, and a base DNN, which is a machine learning model that outputs a base vector, which is vector data of a base, from base information, and performs a vector search using the vector data of the base output from the query DNN and the base DNN to search for a base. The data input to the query DNN is, for example, a character string obtained by combining a search query input by a user and an address. The data input to the base DNN is, for example, the name, address, telephone number, etc. of the base.

[0019] [2. Configuration of Search Device] Next, the configuration of the search device 100 will be described with reference to FIG. 4. FIG. 4 is a diagram showing a configuration example of the search device 100 according to the present embodiment. As shown in FIG. 4, the search device 100 includes a communication unit 110, a storage unit 120, and a control unit 130.

[0020] (Regarding Communications Unit 110) The communication unit 110 is implemented, for example, by a NIC (Network Interface Card). The communication unit 110 is connected to the network N by wire or wireless connection and transmits and receives information with other devices such as the user terminal 10.

[0021] (Regarding memory unit 120) The storage unit 120 is implemented by, for example, semiconductor memory elements such as RAM (Random Access Memory) and flash memory, or by storage devices such as hard disks and optical discs. As shown in Figure 4, the storage unit 120 includes an input data storage unit 121, an address data storage unit 122, and a model data storage unit 123.

[0022] (Regarding the input data storage unit 121) The input data storage unit 121 stores, for example, data to be input into a machine learning model for location search. Here, an example of input data stored in the input data storage unit 121 will be explained using Figure 5. Figure 5 is a diagram showing an example of information stored in the input data storage unit 121 according to this embodiment. In the example in Figure 5, the input data storage unit 121 stores data items such as "search query," "current location / latitude and longitude," "current location / address," and "input value" in association with each other.

[0023] "Search query" is, for example, a search query entered by the user via the user terminal 10. "Current location / latitude and longitude" is, for example, the latitude and longitude indicating the current location obtained using the GPS function of the user terminal 10. "Current location / address" is, for example, the address obtained by converting the latitude and longitude information indicating the current location. "Input value" is, for example, an input value (string) that is input to the machine learning model for location search. The "input value" may be, for example, a string formed by concatenating the search query and the address indicating the current location with a predetermined prefix, as shown in Figure 5. This allows the machine learning model for location search to be trained to identify, for example, the search query entered by the user and the attached current location, which are included in the input value, and to independently decide whether or not to use the current location when searching for a location.

[0024] (Regarding the address data storage unit 122) The address data storage unit 122 stores, for example, data for converting the latitude and longitude indicating the current location, obtained using the GPS function of the user terminal 10, into an address. The information stored by the address data storage unit 122 is, for example, data that associates latitude and longitude with an address or a predetermined area.

[0025] (Regarding the model data storage unit 123) The model data storage unit 123 stores, for example, information about a machine learning model for outputting location information from a search query, and model parameters for constructing the machine learning model. The machine learning model is generated by training it, for example, using a string formed by combining the search query and the address as input data, and a vector representing the location as the ground truth label. This machine learning model is, for example, the query DNN shown in Figure 3, and corresponds to the first machine learning model.

[0026] Furthermore, when training the machine learning model with the search device 100, the model data storage unit 123 may store, for example, the training data for the machine learning model. Figure 6 shows an example of training data according to this embodiment. In the example in Figure 6, the training data is stored by associating data items such as "input data" and "correct label / location". "Input data" is, for example, a string formed by concatenating a search query and an address, which can be input into the machine learning model for location searching, using a predetermined prefix. "Correct label / location" is, for example, vector data of a location that is the correct answer for the "input data".

[0027] Furthermore, the model data storage unit 123 stores, for example, information about a machine learning model for outputting vector data of locations from location information, and model parameters for constructing the machine learning model. The machine learning model is generated, for example, by training it with location information as input data and vectors representing locations as ground truth labels. This machine learning model is, for example, the location DNN shown in Figure 3, and corresponds to the second machine learning model. Also, when training this machine learning model with the search device 100, for example, the training data of this machine learning model may be stored in the model data storage unit 123.

[0028] (Regarding the control unit 130) The control unit 130 is a controller, and is realized by executing various programs stored in the memory device inside the search device 100 using RAM as a working area, for example, by a CPU (Central Processing Unit) or MPU (Micro Processing Unit). Alternatively, the control unit 130 is a controller and can be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), or GPGPU (General Purpose Graphic Processing Unit). As shown in Figure 4, the control unit 130 according to this embodiment has an acquisition unit 131, a conversion unit 132, a coupling unit 133, a search unit 134, and a training unit 135, and realizes or executes the information processing functions and operations described below.

[0029] (Regarding acquisition section 131) The acquisition unit 131 receives and acquires, for example, a search query entered by the user via the user terminal 10, and the latitude and longitude indicating the user's current location, obtained using the GPS function of the user terminal 10. The search query and latitude and longitude acquired from the user terminal 10 may be stored in the input data storage unit 121. The acquisition unit 131 also acquires user attributes, for example, the user's gender, age, and age group, including at least one of these. These user attributes may be entered by the user via the user terminal 10 together with the search query and stored in the input data storage unit 121 or the like, or they may be stored in advance in the storage unit 120 or the like as user information.

[0030] (Regarding the conversion unit 132) The conversion unit 132 converts, for example, the latitude and longitude obtained by the acquisition unit 131 into an address. For this conversion to an address, data stored in the address data storage unit 122 may be used, for example.

[0031] (Regarding the joint 133) The concatenation unit 133 combines, for example, the search query obtained by the acquisition unit 131 and the address converted by the conversion unit 132 into a single string. The process of combining into this string may include, for example, a process of concatenating the search query and the address with a predetermined prefix into a single string. Alternatively, the process of combining into this string may include, for example, a process of concatenating the search query, the address, and the user attributes obtained by the acquisition unit 131 with a predetermined prefix into a single string.

[0032] (Regarding search unit 134) The search unit 134 inputs, for example, the string concatenated by the join unit 133 into a machine learning model that outputs location information from a search query, and searches for and outputs locations. The process of searching for and outputting locations may include a process, as shown in Figure 3, in which a vector search is performed using the location vector data output from both machine learning models, with a machine learning model that outputs location vector data from a search query and another machine learning model that outputs location vector data from location information, to search for and output locations. More specifically, the process of searching for and outputting locations may include, for example, obtaining a third vector output by inputting the string combined by the joining unit 133 into a first machine learning model, which is generated by training a first string formed by combining the first search query and the first address as the first input data, and a first vector indicating the first location as the first ground truth label. Then, using the fourth vector output from a second machine learning model, which is generated by training a second machine learning model, which is linked to a second search query of the URL displayed in the search results of the location by the search unit 134, with the second location information as the second input data, and a second vector indicating the second location as the second ground truth label, along with the third vector, a vector search is performed to output locations. For example, the first machine learning model is the query DNN shown in Figure 3, and the second machine learning model is the location DNN shown in Figure 3. Furthermore, the strings input to the machine learning model may include, for example, search queries, addresses, and user attributes such as the user's gender, age, and age group, and the location search may be further based on these user attributes. For example, the machine learning model may be trained with the training data further including these user attributes so that locations more suitable for the user's gender and age are found.

[0033] (Regarding Training Unit 135) The training unit 135 trains and generates a first machine learning model using, for example, the first input data and the first ground truth label. For example, the first machine learning model is a query DNN as shown in Figure 3, the first input data is a first string formed by combining the first search query and the first address, and the first ground truth label is a first vector indicating the first location. The first input data and the first ground truth label are generated based on the click log of the URL (Uniform Resource Locator) displayed in the location search results by the search unit 134, in order to train the first machine learning model to determine whether or not to use the first address included in the first input data. For example, a URL clicked from the location search results by the search unit 134 is considered to be a URL relating to a location that is the correct answer for the location search, so the training unit 135 can use the location vector data corresponding to that URL as the ground truth label for the training data, and the string input to the machine learning model at that time as the input data for the training data. Furthermore, training a machine learning model to determine whether or not to use addresses included in the input data means, for example, when searching for Tokyo Tower, regardless of the current location address added, if the clicked URL points to Tokyo Tower, the machine learning model can be trained to understand that the current location address does not need to be used in the Tokyo Tower location search by generating training data based on the clicked URL (for example, it will be trained to have less influence from the address).

[0034] Furthermore, the training unit 135 trains and generates a second machine learning model using, for example, the second input data and the second ground truth label. For example, the second machine learning model is the location DNN shown in Figure 3, the second input data is information about the second location, and the second ground truth label is a second vector indicating the second location, linked to the second search query of the URL displayed in the search results of the location by the search unit 134. The second input data and the second ground truth label may also be generated, for example, based on the click log of the URL displayed in the search results of the location by the search unit 134, in order to train the second machine learning model to determine whether or not to use the location's address.

[0035] Furthermore, if the machine learning model is trained using an information processing device other than the search device 100, the search device 100 does not need to have the training unit 135.

[0036] [3. Flowchart of location search process] The procedure for the location search process of the information processing system 1 according to this embodiment will be explained using Figure 7. Figure 7 is a flowchart showing an example of the procedure for the location search process according to this embodiment.

[0037] As shown in Figure 7, the search device 100 obtains, for example, a search query entered by the user via the user terminal 10, and the latitude and longitude indicating the user's current location, obtained using the GPS function of the user terminal 10 (step S101).

[0038] Next, the search device 100 converts the latitude and longitude obtained in step S101 into an address (step S102).

[0039] Next, the search device 100 combines, for example, the search query obtained in step S101 and the address converted in step S102 into a single string (step S103).

[0040] Next, the search device 100 inputs the string concatenated in step S103, for example, into a machine learning model that outputs location information from a search query, and searches for locations (step S104).

[0041] Next, the search device 100 outputs the search results for the locations searched in step S104 (step S105). The output of the search results may be, for example, by sending the search results to the user terminal 10 and displaying them via a display device. After the execution of step S105, the location search process shown in Figure 7 is completed.

[0042] [4. Training Process Flow] The procedure for training using the search device 100 according to this embodiment will be explained with reference to Figure 8. Figure 8 is a flowchart showing an example of the training procedure according to this embodiment.

[0043] As shown in Figure 8, the search device 100 obtains click logs of URLs displayed in the search results for locations output to the user terminal 10, for example, by the location search process shown in Figure 7, from the user terminal 10 (step S201). These click logs may be collected from each of the user terminals 10 by the search device 100 and stored in the storage unit 120.

[0044] Next, the search device 100 generates input data and ground truth labels for training a machine learning model that outputs location information from search queries, based on the click logs obtained in step S201 (step S202). For example, based on the click logs, the vector data of the location corresponding to the URL clicked from the search results for the location is used as the ground truth label, and the string input to the machine learning model at that time is used as input data.

[0045] Next, the search device 100, for example, uses the input data and correct labels generated in step S202 to train and generate a machine learning model that outputs location information from a search query (step S203). After the execution of step S203, the training process shown in Figure 8 is completed.

[0046] [5. Effects] As described above, the search device 100 according to this embodiment includes an acquisition unit 131 that acquires a search query entered by the user and the latitude and longitude indicating the user's current location, a conversion unit 132 that converts the latitude and longitude into an address, a concatenation unit 133 that combines the search query and the address into a single string, and a search unit 134 that inputs the string into a machine learning model that outputs location information from the search query, searches for locations, and outputs the results.

[0047] As a result, the search device 100 according to this embodiment can more accurately search for locations based on search queries entered by the user.

[0048] Furthermore, the string concatenation process performed by the search device 100 according to this embodiment includes the process of concatenating the search query and the address with a predetermined prefix to form a single string.

[0049] As a result, the search device 100 according to this embodiment can identify the search query entered by the user and the address that has been added.

[0050] Furthermore, the process of searching for and outputting locations, performed by the search device 100 according to this embodiment, includes obtaining a third vector output by inputting a string into a first machine learning model, which is generated by training a first machine learning model using a first string formed by combining a first search query and a first address as the first input data, and a first vector indicating the first location as the first ground truth label; performing a vector search using the third vector and a fourth vector output from a second machine learning model, which is generated by training a second machine learning model using information about the second location as the second input data, and a second vector indicating the second location linked to the second search query of the URL displayed in the search results of the location by the search unit 134 as the second ground truth label; and outputting locations.

[0051] As a result, the search device 100 according to this embodiment can more accurately search for locations based on search queries entered by the user.

[0052] Furthermore, the first input data and the first correct label used by the search device 100 according to this embodiment are generated based on the click log of the URL displayed in the search results of the location by the search unit 134, in order to train the first machine learning model to determine whether or not to use the first address included in the first input data.

[0053] This allows the search device 100 according to this embodiment to train a machine learning model so that it can independently determine whether or not to use the current location when searching for a location.

[0054] Furthermore, the search device 100 according to this embodiment includes a training unit 135 that trains and generates a first machine learning model using the first input data and the first correct label.

[0055] As a result, the search device 100 according to this embodiment can more accurately search for locations based on search queries entered by the user.

[0056] Furthermore, the second input data and second correct label used by the search device 100 according to this embodiment are generated based on the click log of the URL displayed in the search results for locations by the search unit 134, in order to train the second machine learning model to determine whether or not to use the location's address.

[0057] This allows the search device 100 according to this embodiment to train a machine learning model so that it can independently determine whether or not to use the address of a location when searching for a location.

[0058] Furthermore, the search device 100 according to this embodiment includes a training unit 135 that trains and generates a second machine learning model using the second input data and the second correct label.

[0059] As a result, the search device 100 according to this embodiment can more accurately search for locations based on search queries entered by the user.

[0060] Furthermore, the search device 100 according to this embodiment has an acquisition unit 131 that acquires user attributes including at least one of the user's gender, age, and age group, and the string concatenation process performed by the search device 100 includes the process of concatenating the search query, address, and user attributes with a predetermined prefix to form a single string.

[0061] As a result, the search device 100 according to this embodiment can search for locations that are more suitable for the user based on the search query entered by the user.

[0062] [6. Hardware Configuration] Furthermore, each of the information processing devices according to the above embodiment is implemented by a computer 1000 having a configuration such as that shown in Figure 9. The following explanation will use the search device 100 as an example. Figure 9 is a hardware configuration diagram showing an example of a computer that implements the functions of the search device 100. The computer 1000 has a CPU 1100, ROM 1200, RAM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700.

[0063] The CPU 1100 operates based on programs stored in the ROM 1200 or HDD 1400, controlling various components. The ROM 1200 stores boot programs executed by the CPU 1100 when the computer 1000 starts up, as well as programs that depend on the computer 1000's hardware.

[0064] The HDD 1400 stores programs executed by the CPU 1100, as well as data used by such programs. The communication interface 1500 receives data from other devices via the communication network 500 (corresponding to network N in this embodiment) and sends it to the CPU 1100, and also transmits data generated by the CPU 1100 to other devices via the communication network 500.

[0065] The CPU 1100 controls output devices such as displays and printers, and input devices such as keyboards and mice, via the input / output interface 1600. The CPU 1100 acquires data from input devices via the input / output interface 1600. The CPU 1100 also outputs data it has generated to output devices via the input / output interface 1600.

[0066] The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1300. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1300 via the media interface 1700 and executes the loaded program. The recording medium 1800 can be, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase Change Rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), tape media, magnetic recording medium, or semiconductor memory.

[0067] For example, when computer 1000 functions as a search device 100, the CPU 1100 of computer 1000 implements the functions of the control unit 130 by executing a program loaded onto RAM 1300. The HDD 1400 stores the data from the storage device of the search device 100. The CPU 1100 of computer 1000 reads and executes these programs from the recording medium 1800, but as another example, these programs may be obtained from other devices via a predetermined communication network.

[0068] [7. Other] Although some embodiments of this invention have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be implemented in various other forms with modifications and improvements based on the knowledge of those skilled in the art, starting with the embodiments described in the disclosure section of the invention.

[0069] Furthermore, the configuration of the aforementioned search device 100 can be flexibly changed, for example, by calling external platforms via APIs (Application Programming Interfaces) or network computing depending on the function.

[0070] Furthermore, the term "part" in the claims can be replaced with "means," "circuit," etc. For example, "acquisition part" can be replaced with "acquisition means" or "acquisition circuit." [Explanation of symbols]

[0071] 10 User terminals 100 Search device 110 Communications Department 120 Storage section 121 Input data storage unit 122 Address data storage unit 123 Model Data Storage Unit 130 Control Unit 131 Acquisition Department 132 Conversion section 133 Joint 134 Search Section 135 Training Department

Claims

1. An acquisition unit that acquires a search query entered by the user, and the latitude and longitude indicating the user's current location, A conversion unit that converts the latitude and longitude into an address, A concatenation unit that combines the aforementioned search query and the aforementioned address into a single string, A search unit that uses a first machine learning model that outputs vector data of locations from a search query and a second machine learning model that outputs vector data of locations from location information to input the string into the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by inputting the string into the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by inputting the string into the first machine learning model, which is generated by training the second machine learning model to obtain a fourth vector output by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second vector output by inputting the information of a second location into the second input data and the second search query of a URL (Uniform Resource Locator) displayed in the search results for the location into the third vector to perform a vector search and search for and output the locations. A search device characterized by having the following features.

2. The process of concatenating the aforementioned string is as follows: The search query and the address are joined together with a predetermined prefix to form a single string. A search device according to claim 1, characterized by including processing.

3. The search device according to claim 1, characterized in that the first input data and the first correct label are generated based on the click log of URLs displayed in the search results of the location by the search unit, in order to train the first machine learning model to determine whether or not to use the first address included in the first input data.

4. The search device according to claim 1 or 3, further comprising a training unit that trains and generates a first machine learning model using the first input data and the first correct label.

5. The search device according to claim 1, characterized in that the second input data and the second correct label are generated based on the click log of the URL displayed in the search results for the location by the search unit, in order to train the second machine learning model to determine whether or not to use the address of the location.

6. The search device according to claim 1 or 5, further comprising a training unit that trains and generates a second machine learning model using the second input data and the second correct label.

7. The acquisition unit is, Obtain user attributes including at least one of the user's gender, age, and age group. The process of concatenating the aforementioned string is as follows: The search query, the address, and the user attributes are combined into a single string by concatenating them with a predetermined prefix. A search device according to claim 1, characterized by including processing.

8. A search method performed by a search device, A process of obtaining a search query entered by the user, and the latitude and longitude indicating the user's current location, A conversion step of converting the latitude and longitude into an address, A concatenation step of combining the aforementioned search query and the aforementioned address into a single string, A search process is performed using a first machine learning model that outputs vector data of locations from a search query and a second machine learning model that outputs vector data of locations from location information, by inputting the string into the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by inputting the string into the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by inputting the string into the first machine learning model, which is generated by training the second machine learning model to obtain a fourth vector output by training the second machine learning model to obtain a fourth vector output by inputting the information of a second location into the second input data and the second vector indicating the second location, which is linked to the second search query of the URL (Uniform Resource Locator) displayed in the search results for the location, and the third vector, and searching for and outputting the locations. A search method characterized by including [a specific feature].

9. A procedure for obtaining a search query entered by the user, and the latitude and longitude indicating the user's current location, A conversion procedure for converting the aforementioned latitude and longitude into an address, A concatenation procedure that combines the aforementioned search query and the aforementioned address into a single string, A search procedure that uses a first machine learning model that outputs vector data of locations from a search query and a second machine learning model that outputs vector data of locations from location information, inputs the string into the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by training the first machine learning model, which is generated by training the first machine learning model to obtain a third vector output by training the first machine learning model to obtain a second vector output by training the second machine learning model, which is generated by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second location, which is generated by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second location, which is generated by training the second machine learning model to obtain a second vector output by training the second machine learning model to obtain a second location, which is generated by training the second machine learning model to obtain a second vector output by training the first machine learning model to obtain a second location, which is generated by training the first machine learning model to obtain a second vector output first vector output by training the first machine learning model, which is generated by training the first machine learning model to obtain a first vector output by training the first machine learning model, which is generated by training the first machine learning model to obtain a first vector A search program that causes a computer to execute a command.

Images