JP7317408B1 - Judgment system, judgment method, and judgment program - Google Patents

Abstract

[Problem] It is possible to determine a type that is close to an objective point of view, rather than a type specified based on the subjective view of the publisher, and improve the accuracy of space searches by searchers who wish to rent a space. A judgment system judgment method and a judgment program are provided. Kind Code: A1 A judgment system that associates a type used for searching a space with a type specified based on the subjectivity of the publisher for the space, and includes image data of the space that the publisher wishes to lend and the space Receiving means for receiving designation of a type; learning means for performing machine learning model learning using image data of a space and the designated type as training data; received image data of the space and the type designated for the space; and calculating means for calculating the likelihood of the space with respect to the type based on the learned model, and determining means for determining the type associated with the space based on the likelihood. [Selection drawing] Fig. 2

Description

Application of Patent Act Article 30, Paragraph 2 February 2, 2022 Website <http://www. instabase. Posted on jp/>

The present invention relates to a judgment system, a judgment method, and a judgment program.

Conventionally, it is known that a large percentage of people who search for rental spaces are looking for "rental conference rooms," so many of the spaces listed on search sites are "rental conference rooms." A flag of "rental conference room" is given even though it is not "room". Therefore, when a searcher searches for "rental conference room", many spaces different from the "rental conference room" that the user is looking for appear in the search results, leading to giving up on the search. it is conceivable that.

In Patent Document 1, a robot device captures an image and stores it as a log of sensor data, uses the log of the sensor data as a training data set, identifies the type of room by image recognition, and more accurately detects a collision with an object. An invention for predicting is disclosed.

Patent No. 7025532

However, although the invention described in Patent Document 1 discloses determining what kind of room it is from a room image, it does not relate to determining a specific room type such as a "rental conference room." is not disclosed, nor is it disclosed about improving the search function based on the determination result of the room type.

In view of the above circumstances, an object of the present invention is to provide a novel technique related to a judgment system.

In order to solve the above-mentioned problems, the present invention provides a determination system that associates a type of space specified based on the subjectivity of the publisher with a type used for searching for the space, and a space that the publisher wishes to lend. Receiving means for receiving the image data of and designation of the type of the space, learning means for learning a machine learning model using the image data of the space and the designated type as teacher data, and the received image data of the space, calculation means for calculating the likelihood of the space with respect to the type based on the type designated for the space and a trained model; determination means for determining the type associated with the space based on the likelihood; characterized by having With such a configuration, the present invention can determine a type close to an objective point of view rather than a type specified based on the subjectivity of the publisher, and a searcher who wishes to rent a space can It can improve accuracy when searching for spaces.

In a preferred embodiment of the present invention, one or more types out of a plurality of types are specified for the space, and the determining means determines whether the space corresponds to each of the plurality of types. characterized by With such a configuration, when one space can be used for multiple purposes, the present invention makes it difficult for a searcher to miss a search when searching for a space that the searcher wishes to rent. .

In a preferred embodiment of the present invention, the determining means associates the type with the space when the likelihood of the space with respect to the type is equal to or greater than a threshold. With such a configuration, the present invention can present to the searcher only spaces in which the likelihood of a given type exceeds the threshold.

In a preferred embodiment of the present invention, the learning means performs learning using the image data and the type received via the receiving means as teaching data. With such a configuration, the present invention can generate a machine learning model capable of performing relearning and making a type of judgment closer to an objective point of view.

In a preferred embodiment of the present invention, the reception means receives image data transmitted as the main image of the space and image data transmitted as the sub-image, and the learning means receives the image data transmitted as the main image. It is characterized by performing learning as teacher data. With such a configuration, the present invention can remove image data that may become noise during learning when a plurality of image data is received.

In a preferred embodiment of the present invention, search processing means for receiving a space search request and transmitting a search processing result is provided, and the search processing means determines the display order of the spaces filtered based on the type according to the likelihood. It is characterized in that it is determined based on the amount. With such a configuration, the present invention can present a space closer to the searcher's request.

In order to solve the above-mentioned problems, the present invention provides a determination method for associating a type used for searching a space from a type specified based on the publisher's subjectivity for the space, wherein a computer determines whether the publisher lends A receiving step of receiving image data of a desired space and designation of a type for the space, a learning step of learning a machine learning model using the image data of the space and the designated type as training data, and a learning step of the received space. a calculation step of calculating a likelihood of the space with respect to the type based on image data, a type specified for the space, and a trained model; and a determination of determining a type associated with the space based on the likelihood. It is characterized by performing steps and

In order to solve the above-mentioned problems, the present invention provides a judgment program that associates a type of space specified based on the subjectivity of the publisher with a type used for space search, and a space that the publisher wishes to lend. Receiving means for receiving the image data of and designation of the type of the space, learning means for learning a machine learning model using the image data of the space and the designated type as teacher data, and the received image data of the space, calculation means for calculating the likelihood of the space with respect to the type based on the type designated for the space and a trained model; determination means for determining the type associated with the space based on the likelihood; characterized by having

The present invention can provide a novel technology related to the determination system.

1 shows a configuration diagram of a determination system in an embodiment of the present invention; FIG. 2 shows a hardware configuration diagram of a publisher terminal 2A and a searcher terminal 2B in the embodiment of the present invention; FIG. 1 shows a hardware configuration diagram of a management server 1 in an embodiment of the present invention; FIG. 4 shows an example venue label in an embodiment of the present invention. 4 is a flow chart showing the procedure for determining venue labels for search in the embodiment of the present invention.

This specification describes the configuration, effects, and the like according to an embodiment of the present invention with reference to the drawings. The present invention is not limited to the following embodiments, and can employ various configurations. In addition, the embodiments of the present invention may mutually employ a part of the configuration of each of the respective embodiments within a range that does not impede the realization of the intended effect of the present invention.

In addition, although the configuration, operation, etc. of the determination system will be described in this embodiment, similar effects can be achieved by methods, computer programs, and the like that are executed. The program in this embodiment may be provided as a computer-readable non-transitory recording medium, may be provided as a downloadable form from an external server, or may be provided as a downloadable program from an external server. The program may be started on an external computer immediately (so-called cloud computing).

Further, in the present embodiment, "means" may include, for example, a combination of hardware resources implemented by circuits in a broad sense and software information processing that can be specifically realized by these hardware resources. In this embodiment, "information" refers to, for example, the physical value of a signal value representing voltage or current, the height of a signal value as a binary bit aggregate composed of 0 or 1, or the quantum superposition. (so-called quantum bits), and communication and computation can be performed on a circuit in a broad sense.

A circuit in a broad sense is a circuit realized by appropriately combining circuits, circuits, processors, memories, and the like. That is, CPU (Central Processing Unit), GPU (Graphics Processing Unit), LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Ga Array) and the like.

In this embodiment, a case of searching for a space on a space rental platform for renting a space by the hour will be described. In the determination system, the publisher registers the space that the publisher wants to rent out, and the searcher searches and reserves the space that the searcher wants to use.
Note that the method of using the determination system is not limited to this. For example, it may be possible to use it for a real estate brokerage platform and a public facility reservation system.

<System configuration>
FIG. 1 is a schematic diagram of a determination system according to Embodiment 1. FIG. The determination system includes a management server 1, a publisher terminal 2A for registering information on a space that a publisher wishes to rent, and a searcher terminal for searching and reserving a space that a searcher wishes to rent. 2B. The management server 1, the publisher terminal 2A, and the searcher terminal 2B are configured to be able to communicate via a communication network NW.

The communication network NW is composed of an IP (Internet Protocol) network such as the Internet. In the following description, the intervention of the communication network NW will be omitted unless it is unclear.

In FIG. 1, one management server 1, one publisher terminal 2A, and one searcher terminal 2B are shown, but a plurality of each may exist.

<Management server 1>
As the management server 1, it is possible to use a general-purpose server-oriented computer, a personal computer, or the like. Moreover, in this embodiment, it is also possible to configure the management server 1 using a plurality of computers.

<Publisher Terminal 2A and Searcher Terminal 2B>
Each publisher terminal 2A and searcher terminal 2B may be a personal computer, a smart phone, a tablet terminal, or the like. The publisher terminal 2A and the searcher terminal 2B have an application (typically a web browser) for making requests to the management server 1 and receiving responses.

<Hardware Configuration of Management Server 1>
FIG. 2 is a diagram showing an example of the hardware configuration of the management server 1. As shown in FIG. The management server 1 includes, as a hardware configuration, an arithmetic unit (CPU (Central Processing Unit)) 11 and a main storage device (RAM (Random Access Memory)) 12 as a working memory.

The management server 1 includes an auxiliary storage device 13 such as an HDD, SSD, or flash memory that rewritably stores an OS (Operating System), application programs, and various information (including data), a communication control unit 16, and a NIC ( and a communication interface (IF) unit 17 such as a Network Interface Card).

<Hardware configuration of publisher terminal 2A and searcher terminal 2B>
FIG. 3 is a diagram showing an example of the hardware configuration of the publisher terminal 2A and the searcher terminal 2B. The publisher terminal 2A and the searcher terminal 2B include, as hardware configurations, an arithmetic unit (CPU (Central Processing Unit)) 21 and a main storage device (RAM (Random Access Memory)) 22 as a working memory. .

The publisher terminal 2A and the searcher terminal 2B have an OS (Operating System), application programs, and an auxiliary storage device 23 such as an HDD, SSD, or flash memory that rewritably store various information (including data), and communication control. It further includes a unit 26 and a communication interface (IF) unit 27 such as a NIC (Network Interface Card).

In addition, the publisher terminal 2A and the searcher terminal 2B include an input unit 24, which is an interface that enables operation input, such as a mouse, keyboard, and touch panel, and a monitor, display, etc., for the publisher and the searcher. It further includes an output unit 25 that is an interface for displaying a screen, which will be described later.

<Functional components>
As illustrated in FIG. 2, the management server 1 has reception means 101, learning means 102, calculation means 103, determination means 104, search processing means 105, output processing means 106, and database DB.

In order to logically implement the functional components described above in the management server 1, a judgment program is installed in the auxiliary storage device 13 as an application program. In the management server 1, when the power is turned on, the arithmetic device 11 expands the determination program stored in the auxiliary storage device 13 to the main storage device (RAM) 12 and executes it, thereby performing the functions described above. A configuration is realized.

<Type definition>
In this embodiment, a type is a subjectively assigned index for a space. Examples of subjectively specified indicators are space usage, space quality, taste, concept and atmosphere. Type can be a difficult indicator to set qualitatively. In this embodiment, as types (hereafter referred to as "venue types") representing the use of the rental space, there are "rental meeting rooms", "rental spaces", "seminars", "rental studios", "music studios", "rental salons", and "photography". Studio, party room, co-working space, chartered cafe/restaurant, exhibition hall/gallery, pop-up store, live house/theater, sports facility, event space, old folk house, house studio "Rooftop/outdoor", "workspace", "amusement facility", "hotel", etc. are set. When setting the taste of the room as the type, for example, "Brooklyn style", "hotel-like interior", "mid-century", "Japanese modern", and "shabby chic" are set.

<Enter information to be posted>
The accepting unit 101 accepts image data of a space that the publisher wishes to lend and designation of the type of the space. In this embodiment, the publisher operates a space information registration screen (not shown) displayed on the publisher terminal 2A, and registers the space name, nearest station, address, and hourly rate for the purpose of lending. It transmits to the management server 1 the rental fee, available date and time for rent, days of the week, size, capacity, description, image data of the space, and the venue type indicating the use of the space specified by the publisher. The reception unit 101 associates this information with the space ID, assigns a label (hereinafter referred to as "designated venue label") based on the designated venue type, and stores it in the database DB as space information.

In this embodiment, the image data received by the receiving means 101 may be one or a plurality of data. When receiving a plurality of image data, the image data transmitted as the main image and the image data transmitted as the sub-image may be received. . Also, the publisher can designate a plurality of venue labels within a predetermined number. In this embodiment, the venue labels include a designated venue label designated by the publisher as the venue label corresponding to the space shown in the image data, a search venue label associated with the space information by the determination means 104 described later, may contain

<Example of designated venue label>
FIG. 4A is a diagram showing an example of a designated venue label for each piece of space information received via the receiving means 101. FIG. In the illustrated example, the venue type designated by the publisher is stored as "1", which defines the designated venue label. On the other hand, venue types that are not specified are stored as "0".

<Generation of machine learning model>
The learning means 102 generates a machine learning model using the image data of the space and the venue type associated with the image data as teacher data. The training data is, for example, image data pre-stored in the database DB and venue types associated with the image data. Here, the image data previously held in the database DB may be image data prepared in advance by the owner of the management server 1 or the like and the venue type associated with the image data. It may be the image data received from the publisher via the receiving means 101 and the venue type associated with the image data.

Further, when space information including a plurality of image data is received via the receiving means 101, the learning means 102 uses the image data transmitted as the main image when receiving the space information as an input value, and the space information is A machine learning model is generated using teacher data whose output value is the venue type, including the venue type.

<Likelihood calculation>
The calculating means 103 calculates the likelihood of the space for the venue type based on the image data received via the receiving means 101, the venue type associated with the image data, and the learned model. Calculation means 103 calculates a likelihood, which is a numerical value indicating “likeness” for each venue type of space, by image recognition based on the trained model.

The calculating means 103 may calculate the likelihood for all venue types of one space, or may calculate the likelihood only for a predetermined venue type. For example, if there are many spaces with the specified venue label of "Rental conference room" even though they are not " Rental conference room", spaces other than "Rental conference room" will not be displayed in search results by searchers. Then, only the likelihood for the venue type of “rental meeting room” is calculated.

<Example of likelihood for venue type>
FIG. 4(b) shows the likelihood for each venue type of a certain space calculated by the calculating means 103. FIG. Here, in the present embodiment, the likelihood calculated by the calculating means 103 is a numerical value between 0 and 1, but the range of likelihood may be arbitrary.

<Determination of venue type>
The determination means 104 determines the venue type associated with the space based on the likelihood calculated by the calculation means 103 . In this embodiment, if the likelihood of a space for a venue type is greater than a threshold, then it is determined that the space corresponds to the venue type.

As an example, if the threshold is 0.7, the space with "space ID: 0002" in FIG. However, it is determined that the two types of "rental space" and "rental studio" do not apply. Note that the same threshold value may be used for some or all venue types, or may be different for each venue type.

<Attachment of venue label for search>
Further, the determining means 104 assigns a label (hereinafter referred to as a "search venue label") based on the determined venue type, and stores the label in the database DB in association with the space information.

Note that the number of search venue labels assigned to space information is equal to or less than a predetermined number. If the number of venue types with likelihoods greater than the threshold is greater than a predetermined number, venue labels for search within the same number as the above number may be associated with the space information in descending order of likelihood. Further, when all the likelihoods in a certain space calculated by the calculation means 103 are smaller than the threshold value, the determination means 104 assigns a search venue label based on the venue type with the highest calculated likelihood, and the space It may be stored in the database DB in association with the information.

Further, when the calculation means 103 calculates only the likelihood for a predetermined venue type, the determination means 104 determines whether the space corresponds to the venue type, and the venue label for search other than the venue type is posted. It may be the same as the designated venue label based on the designation of the party.

<Search for space information>
The search processing means 105 receives a search request via the input unit 24 and transmits the search processing result to the searcher terminal 2B. In this embodiment, the search processing means 105 receives search conditions and search requests input via the input unit 24 on a space search screen (not shown), which will be described later, and based on the search conditions, the search conditions are retrieved from the database DB. Acquire space information corresponding to . The search conditions are conditions related to the space that the searcher wishes to rent, and are information related to other conditions such as the search site label specified by the searcher, the nearest station, and the rent per hour.

<Display processing of space information registration screen and space search screen>
The output processing means 106 displays a space information registration screen (not shown) and a space search screen (not shown), and transmits the processing results to the publisher terminal 2A and the searcher terminal 2B.

On the space information registration screen, it is possible for the publisher to upload image data, select the venue type to which the image data corresponds based on the publisher's subjectivity, and input other information related to the image data. At this time, on the space information registration screen, frames for uploading the main image and the sub image may be provided, and the image data to be the main image may be determined from among the images to be uploaded. Also, the image that is uploaded first among the plurality of images may be the main image.

On the space search screen, it is possible to receive the input of search conditions relating to the space that the searcher wishes to rent, and search for the space. The space search screen has a label designation section that accepts designation of a designated venue label, a condition designation section that allows designation (input) of information related to other conditions, and a condition confirmation section that confirms search conditions and transmits a search request. . Further, the space search screen can display the space information acquired by the search processing means 105 according to the above search conditions. Here, the search processing means 105 determines the display order of the space information to which the same venue label is assigned and displayed on the space search screen, depending on the degree of likelihood for the venue type of the venue label. Also, when a plurality of specified venue labels are specified in the search condition, the space information to be displayed is determined in descending order of the total likelihood for the specified venue label or venue type with the highest priority.

<Flowchart of procedure for determining venue label for search>
An example of the overall procedure for determining venue labels for search using the determination system will be described with reference to FIG. In this embodiment, when searching for a space that a searcher wishes to rent, it is based on a more objective judgment based on a machine learning model rather than a designated venue label that is specified based on the subjectivity of the publisher. It is used for the purpose of enabling a search using the decided search venue label. Note that the order of each step shown in FIG. 5 is an example, and the order may be changed as appropriate unless otherwise specified.

The receiving means 101, through the input unit 24, specifies the image data of the space that the publisher wishes to rent, the designation of the venue label indicating the venue type that the publisher considers to be applicable to the space, and other information related to the space. The input of the space information including the information of is accepted (step S101).

The calculating means 103 calculates the likelihood for each venue type of the space based on the learned model and the image data and designated venue label of the space received in step S101 (step S102).

Based on the likelihood calculated in step S102, the determining means 104 determines the venue label for search to which the space corresponds (step S103). The determining means 104 associates the venue label for search determined in step S103 with the space information and stores them in the database DB (step S104).

In this embodiment, the venue label for search determined in step S103 is stored in the database DB in association with the space information in step S104. The search processing means 105 may acquire information on a space whose likelihood in the search venue label specified as the search condition is higher than a threshold value, and the determination means 104 may assign the search venue label to the space information. .

The determination system has reservation means (not shown), and after searching for space information via the space search screen, the searcher can reserve the searched space via the reservation means.

Further, the learning unit 102 receives the image data and the relevant image data received via the receiving unit 101 after a predetermined period of time has elapsed since the previous learning or when a predetermined number of pieces of space information have been received. The learned model may be retrained using the venue type associated with the image data as teacher data. Further, when the learned model is re-learned, the calculation means 103 calculates each venue type of the space for which the likelihood for the venue type was calculated before the re-learning using the learned model after the re-learning. may be recomputed.

According to the present invention, it is possible to provide a novel technology related to the determination system.

1: Management server 2A: Publisher terminal 2B: Searcher terminals 11, 21: Arithmetic devices 12, 22: Main storage devices 13, 23: Auxiliary storage device 24: Input unit 25: Output units 16, 26: Communication control unit 17 , 27: communication IF section 101: reception means 102: learning means 103: calculation means 104: determination means 105: search processing means 106: output processing means

Claims (6)

A determination system that associates a space with a type used to search for the space,
learning means for learning a machine learning model using image data of a space prepared in advance and a type associated with the image data of the space prepared in advance as teacher data;
Receiving means for receiving image data of a space that a publisher wishes to rent;
calculation means for calculating the likelihood of the space for the type used for the search based on the received image data of the space and the trained model;
determination means for determining a type to be associated with the space based on the likelihood;
search processing means for receiving space search requests and transmitting search processing results;
A determination system in which the search processing means determines a display order of the spaces filtered based on the type determined by the determination means based on the magnitude of the likelihood. 2. The determination system according to claim 1, wherein when the likelihood of said space with respect to said type is equal to or greater than a threshold, said determination means associates said type with said space. The receiving means further receives, from among a plurality of types, designation of one or more types for the space that the publisher wishes to lend,
2. The determination system according to claim 1, wherein said learning means performs learning using said image data and said type received through said receiving means as teaching data. The receiving means receives a plurality of the image data,
The reception means receives image data transmitted as a main image of the space and image data transmitted as a sub-image,
2. The determination system according to claim 1, wherein said learning means performs learning using image data transmitted as said main image as teacher data. A determination method for associating a space with a type used for space search,
the computer
a learning step of learning a machine learning model using image data of a space prepared in advance and a type associated with the image data of the space prepared in advance as training data;
a receiving step of receiving image data of a space that a publisher wishes to rent;
a calculation step of calculating the likelihood of the space for the type used for the search based on the received image data of the space and the trained model;
a determination step of determining a type associated with the space based on the likelihood;
executing a search processing step of receiving a space search request and transmitting search processing results;
The determination method, wherein the search processing step determines the display order of the spaces filtered based on the type determined by the determination step , based on the magnitude of the likelihood. A determination program that associates a space with a type used for space search,
the computer,
learning means for learning a machine learning model using image data of a space prepared in advance and a type associated with the image data of the space prepared in advance as teacher data;
Receiving means for receiving image data of a space that a publisher wishes to rent;
calculation means for calculating the likelihood of the space for the type used for the search based on the received image data of the space and the trained model;
determination means for determining a type to be associated with the space based on the likelihood;
functioning as a search processing means for receiving space search requests and transmitting search processing results;
A determination program in which the search processing means determines the display order of the spaces filtered based on the type determined by the determination means, based on the magnitude of the likelihood.