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

Abstract

To handle various feature quantities associated with a building frame and its members mathematically and uniformly by applying natural language processing technique to the frame.SOLUTION: An information processing system comprises: an acquisition part which acquires building quantities; a building quantity documentation part which creates a document including the building quantities; a building quantity dispersion expression part which dispersedly expresses the documented building quantities; and a frame member cross section estimation part which estimates a frame member cross section of a member of interest on the basis of the dispersedly expressed building quantities.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] In general, the cross section of structural members is estimated by extracting the characteristics of the structure and members and then having a computer learn the relationship between these and the cross section of the members.

[0003] For example, Patent Document 1 discloses a similar building extraction device that includes an architectural frame information database unit that stores architectural frame information related to the frame of each of a plurality of buildings, a target architectural frame information acquisition unit that acquires target architectural frame information, which is architectural frame information related to the frame of the target building, a similar building extraction unit that extracts buildings from the plurality of buildings that have frames similar to the target building, and an output unit that outputs information indicating the extracted buildings, where the similar building extraction unit selects for each of the plurality of buildings and extracts similar buildings based at least on the architectural frame information of the selected building and the TanimotoScore calculated from the target architectural frame information, and the ratio of the maximum height of the target building to the maximum height of the selected building. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2023-108331 Summary of the Invention [Problem to be solved by the invention]

[0005] However, expressing structural features such as structural member configuration and structural shape is extremely difficult. While it is possible to express building structures using features such as building height, main spans, and aspect ratios, these features also have limitations, making it impossible to mathematically unify the overall structure of individual building structures, which vary in scale and structural configuration. Furthermore, even if more features were added to represent the structural details, extracting them from individual buildings would require significant time and effort, making it inefficient. Furthermore, even when combining these structural features with structural member features to estimate structural member cross sections, many of these features have unique meanings, making it difficult to treat them in a unified manner. This characteristic of building structures, which is extremely difficult to treat in a unified mathematical manner, poses a major obstacle to using AI technology to estimate structural member cross sections and is a major factor hindering the improvement of estimation accuracy.

[0006] The present invention has been made in consideration of the above points, and aims to make it possible to mathematically and unifiedly handle various feature quantities related to building frames and their components by applying natural language processing technology to frames. [Means for solving the problem]

[0007] The present invention has been made to solve the above-mentioned problems, and one aspect of the present invention is an information processing system comprising an acquisition unit that acquires building quantities, a building quantity documentation unit that documents the building quantities into a sentence including the building quantities, a building quantity distributed representation unit that distributes the documented building quantities, and a structural member cross section estimation unit that estimates the structural member cross section of a target member based on the distributed representation of the building quantities.

[0008] Another aspect of the present invention is an information processing device that includes an acquisition unit that acquires building quantities, a building quantity documentation unit that documents the building quantities into a sentence including the building quantities, a building quantity distributed representation unit that distributes the documented building quantities, and a structural member cross section estimation unit that estimates the structural member cross section of a target member based on the distributed representation of the building quantities.

[0009] Another aspect of the present invention is an information processing method executed by a computer of an information processing device, the information processing method including an acquisition step of acquiring building quantities, a building quantity textualization step of textualizing the building quantities into a sentence including the building quantities, a building quantity distributed representation step of distributing the textualized building quantities, and a structural member cross section estimation step of estimating the structural member cross section of a target member based on the distributed representation of the building quantities.

[0010] Another aspect of the present invention is a program that causes a computer of an information processing device to execute an acquisition step of acquiring building quantities, a building quantity textualization step of textualizing the building quantities into a sentence including the building quantities, a building quantity distributed representation step of distributing the textualized building quantities, and a structural member cross section estimation step of estimating the structural member cross section of a target member based on the distributed representation of the building quantities. [Effects of the Invention]

[0011] According to an embodiment of the present invention, by applying natural language processing technology to building frames, it is possible to mathematically and unifiedly handle various feature quantities related to building frames and their components. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a block diagram showing an example of a configuration of an information processing device 100 according to a first embodiment of the present invention. [Figure 2] FIG. 2 is an explanatory diagram illustrating an example of processing in the information processing device 100 according to the present embodiment. [Figure 3]FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 4] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 5] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 6] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 7] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 8] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to the present embodiment. [Figure 9] 10 is a flowchart showing an example of processing in the information processing device 100 according to the present embodiment. [Figure 10] FIG. 10 is an explanatory diagram illustrating an example of processing in the information processing device 100 according to a modified example of the present embodiment. [Figure 11] FIG. 10 is an explanatory diagram illustrating another example of the processing in the information processing device 100 according to a modified example of the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] First Embodiment Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0014] [Configuration of information processing system] First, an information processing system SYS according to the first embodiment will be described. Although an example of the information processing system SYS according to this embodiment is described in which the information processing device 100 is realized as one device, the functions of the information processing device 100 may be realized as multiple devices in the information processing system SYS.

[0015] The information processing system SYS (information processing device 100) is a system that acquires building quantities, converts them into text containing the building quantities, distributes the textual building quantities, and estimates the cross section of the structural member of a target member based on the distributed representation of the building quantities.

[0016] Specifically, the information processing system SYS treats the length of the building frame members, the attachment state of surrounding members, and the lengths of surrounding members as each word, and considers the building frame to be a sentence or paragraph made up of these, applying distributed representations in natural language processing to each feature to estimate the cross section of the frame members. In other words, the information processing system SYS represents each feature as a fixed-length vector of the same dimension, and combines this with various AI to estimate the building frame members.

[0017] By applying natural language processing technology to building structures, the information processing system SYS can mathematically unify the features related to building structures and their components. Specifically, the information processing system SYS can represent all building-related features and building component configurations as fixed-length vectors of the same dimension. This allows combinations of features to be expressed as their vector sums, and similarities in structures, etc., can be expressed as vector operations, such as the dot product of fixed-length vectors representing individual buildings. Furthermore, converting various features and building structures into fixed-length vectors makes them easier to incorporate into matrix and vector operations frequently used in AI technologies such as neural networks, potentially enhancing their compatibility with existing and future AI technologies. Furthermore, this fixed-length vectorization (distributed representation) of features and building component configurations, etc., is expected to be useful for future applications of natural language processing-related technologies developed in other fields, such as attention, to building structures.

[0018] Next, the information processing device 100 will be described with reference to the drawings. FIG. 1 is a block diagram showing an example of the configuration of an information processing device 100 according to the first embodiment of the present invention. The information processing device 100 includes a communication unit 11, a control unit 12, a storage unit 13, an input unit 14, and an output unit 15. The communication unit 11, the control unit 12, the storage unit 13, the input unit 14, and the output unit 15 are connected to one another via a bus.

[0019] The communication unit 11 has a function of communicating with other devices in the information processing system SYS via a network. The communication unit 11 is controlled by the control unit 12, and transmits various pieces of information output from the control unit 12 to the other devices. The communication unit 11 receives various pieces of information transmitted from the other devices, and outputs the received various pieces of information to the control unit 12.

[0020] The storage unit 13 stores the learning model. The input unit 14 is an input device such as a mouse, a keyboard, a touch panel, or a microphone. The output unit 15 is an output device such as a display unit and a speaker.

[0021] The control unit 12 controls the processing of each unit of the information processing device 100 and performs the functions of the information processing device 100. The control unit 12 may be configured with dedicated hardware, or may be configured with a general-purpose computer system. The computer system includes at least an arithmetic processing unit and a storage medium. For example, a processor such as a CPU (Central Processing Unit) can be used as the arithmetic processing unit. The storage medium is a storage unit 13 and a flash memory (not shown). The arithmetic processing unit reads a program stored in advance in the flash memory, for example, and loads the read program into the storage unit 13. The arithmetic processing unit executes the program loaded into the storage unit 13 to perform the functions of the control unit 12.

[0022] The term "CPU" as used herein refers to a processor in general, and includes not only a device known as a CPU in the narrow sense, but also, for example, a GPU, a DSP, etc. The CPU as used herein is not limited to being realized by a single processor, but may be realized by combining multiple processors of the same or different types.

[0023] The control unit 12 includes a building quantity acquisition unit 121 , a building quantity documentation unit 122 , a building quantity distributed expression unit 123 , a structural member cross section estimation unit 124 , and a learning unit 125 .

[0024] The building quantity acquisition unit 121 acquires building quantities stored in advance in the storage unit 13 or building quantities input via the communication unit 11 or the input unit 14. The building quantity acquisition unit 121 outputs the acquired building quantities to the building quantity documentation unit 122.

[0025] Here, the building quantities are quantities that characterize a building, such as the building height, the installation level (height) of the target component, the length of the beam, the type of end joint, the length of each joined component, etc. The target component is the component that is the target of cross-section estimation.

[0026] When the building quantities are input from the building quantity acquisition unit 121, the building quantity documentation unit 122 performs natural language processing to translate the building quantities into documentation. Specifically, when the building quantities are a building height a [m], a member (target member) installation level b [m], a beam (column) member length [m], an end joint type d [type], and lengths e [m] and f [m] of the members whose ends are joined, the building quantity documentation unit 122 performs pre-linguistic processing on the building quantities to document them as follows: "The building height is a meters, its installation level is b meters, the beam (column) member length is c meters, the end joint type is d, and the lengths e and f meters of the members whose ends are joined." The building quantity documentation unit 122 outputs the documented building quantities to the building quantity distributed expression unit 123.

[0027] Here, if sentences (sets of words) other than the building quantities a to f are treated as fixed sentences, the building quantities a to f can be described as sentences. In other words, if a distributed representation of each of the building quantities a to f is obtained when they are treated as words, the sentences can be converted into distributed representations (fixed-length vectors) in subsequent processing. The cross-sections of the frame members can be estimated using a learning model that has learned the correspondence between the distributed representations (vectorized) of the building quantities and the cross-sections of the frame members, and the distributed representations of the building quantities. Furthermore, if the cross-section size is x, then x can be sufficiently handled by assigning an index number to the cross-section, so there is no need to specifically convert it into a distributed representation. Therefore, it is sufficient to obtain the distributed representations of the building quantities a to f.

[0028] When the documented building quantities are input from the building quantity documenting unit 122, the building quantity distributed expression unit 123 converts the documented building quantities into a distributed expression. Although various techniques can be used to create a distributed representation of building quantities, this embodiment describes a case in which building quantities are created based on the maximum height of a frame including a target component, the position of the target component, and the relationship between the target component and surrounding components connected to it, as shown below. Here, this technique is referred to as Frame2Vec. Specifically, the building quantity distributed representation unit 123 identifies the types of building height, the component installation level of the target component, the end joint type, and the component lengths of the components to which the end joints are connected. The building quantity distributed representation unit 123 assigns index numbers to the identified types and creates a vector by assigning 1 to the elements of the corresponding row for each target component length, thereby creating a distributed representation (vectorization) of the building quantities. The building quantity distributed representation unit 123 also calculates the occurrence probability of each of these in a series of frame groups. The building quantity distributed representation unit 123 ultimately creates a distributed representation of the building quantities using a one-layer neural network (NN) with the above vectors composed of 1s and 0s as the input layer and the occurrence probability as the output layer. The building quantity distributed expression unit 123 outputs the building quantities that have been distributed to the frame member cross section estimation unit 124.

[0029] This will be described in detail with reference to FIGS. Fig. 2 is an explanatory diagram illustrating an example of processing in the information processing device 100 according to the present embodiment. Fig. 3 is an explanatory diagram illustrating another example of processing in the information processing device 100 according to the present embodiment. Fig. 4 is an explanatory diagram illustrating another example of processing in the information processing device 100 according to the present embodiment. Fig. 5 is an explanatory diagram illustrating another example of processing in the information processing device 100 according to the present embodiment.

[0030] 2, the building A100 is defined as the height from the ground of the building A100 as height A11, the installation level (height) from the ground of the target component A1 as height A12, and the components whose ends are joined as components A2 and A3. Here, the maximum height of the frame that includes the target component and the height of the position of the target component (area A10) are defined as height A12.

[0031] The example shown in Figure 3 is a detailed example of region A10 in Figure 2. Also, a component A2 (component A21 and component A22) is defined as having an end joined to the component A1 of interest by a joint A20, and a component A3 (component A31 and component A32) is defined as having an end joined to the component A1 of interest by a joint A30.

[0032] The building quantity distributed representation unit 123 generates a distributed representation of building quantities based on the maximum height A12 of the frame including the target component, the position of the target component A1, and the relationship between the length of the target component A1 and the lengths of the surrounding components A2 and A3 connected to it. Specifically, the building quantity distributed representation unit 123 identifies the building height A11, the component installation level A12 of the target component, the end joint A20 and the joint type of A20, and the lengths of the components A2 and A3 to which the end joints are connected, and determines the types of these for each item. The building quantity distributed representation unit 123 assigns index numbers to the identified types and creates a vector that assigns 1 to the elements of the corresponding row for each length of the target component, thereby generating a distributed representation (vectorization) of the building quantities as shown in Figure 4. The building quantity distributed representation unit 123 also calculates the occurrence probability of these types in a series of frames. The building quantity distributed expression unit 123 finally generates a distributed expression of the building quantities using a one-layer NN with the vector composed of 1s and 0s as the input layer and the occurrence probability as the output layer.

[0033] In the example shown in FIG. 4, the beam member length of the target member A1 is "22.0". The occurrence probability of the target member is 1.0. In the example shown in FIG. 4, the occurrence probability of each label is multiplied by 1 / 5, because the sum of the values ​​of each element is normalized to 1.0. This is because the occurrence rate is calculated for each item, and the sum of the values ​​of each element would be 5.0 if left as is.

[0034] An example of this is shown in Figure 5. The finally obtained column vector of Win is a distributed representation vector of the values ​​of each term. The building quantity distributed representation unit 123 concatenates the OneHot vectors of the corresponding item labels to generate a distributed representation vector.

[0035] Returning to FIG. 1, when the distributed representation of building quantities is input from the distributed representation of building quantities unit 123, the frame member cross section estimation unit 124 estimates the cross sections of the frame members using the following method (herein referred to as Vec2Sec). Specifically, the frame member cross section estimation unit 124 inputs the distributed representation of building quantities (target member V vector) into a learning model that has learned the correspondence between the distributed representation of building quantities and the cross sections of the frame members (OneHot vectors of cross section labels), thereby obtaining the cross sections of the frame members as output. Here, the target member V vector includes the building height, member installation level, joint type, and the sum of Σ member length vectors, which is the sum of the lengths of each member whose end is joined. The length of each member whose end is joined includes the length of the target member. An example of these is shown in FIG. 8.

[0036] Furthermore, the frame member cross section estimation unit 124 estimates the estimated member cross section of the planned building as the cross section corresponding to the largest value among the vector elements output when the V vector of each target member of the planned building is input to the NN. Here, if the row giving the maximum value of the vector element is index i, the cross section of index i is treated as the estimated cross section. These are shown in the examples in Figures 7 and 8.

[0037] Note that the building height, component installation level, component length sum, and component length of the planned building do not necessarily match the training data. Therefore, the distributed representations (vectors) of the building quantities of the planned building are interpolated and extrapolated based on the vectors obtained during training. For example, if the building height of the planned building is hd, and the upper and lower limits obtained during training are hu and hl, respectively, and the distributed representation vector corresponding to the upper limit hu is Vhu and the distributed representation vector corresponding to the lower limit hl is Vhl, then the distributed representation vector Vhd corresponding to the building height hd of the planned building is given by the internal division of the vector below. Vhd={Vhu(hd-hl)+Vl(hu-hd)} / (hu-hl) =Vhuα+Vl(1-α) α=(hd-hl) / (hu-hl)

[0038] In this way, by combining Frame2Vec and Vec2Sec, the information processing device 100 can construct a system (Frame2Sec) that estimates the cross section of a component member from the geometry information (feature amount) of a frame via its distributed representation.

[0039] Returning to FIG. 1 , the learning unit 125 generates a learning model that learns the correspondence between the distributed representation of building quantities and the cross sections of structural members (OneHot vectors of cross section labels). The learning model receives the distributed representation of building quantities (target member V vectors) as input and obtains the cross sections of structural members as output. The learning unit 125 stores the generated learning model in the memory unit 13.

[0040] FIG. 9 is a flowchart showing an example of processing in the information processing device 100 according to this embodiment. In step S101, the information processing device 100 acquires building quantities. Next, the information processing device 100 executes the process of step S102. In step S102, the information processing device 100 documents the building quantities. Next, the information processing device 100 executes the process of step S103. In step S103, the information processing device 100 converts the building quantities into a distributed representation. Next, the information processing device 100 executes the process of step S104. In step S104, the information processing device 100 estimates the cross section of the structural member by inputting the building quantities in the distributed representation into the learning model.

[0041] As such, the information processing device 100 according to this embodiment includes an acquisition unit (building quantity acquisition unit 121) that acquires building quantities, a building quantity text generation unit 122 that converts the building quantities into text containing the building quantities, a building quantity distributed expression unit 123 that converts the textual building quantities into a distributed expression, and a structural member cross section estimation unit 124 that estimates the structural member cross section of a target member based on the distributed expression of the building quantities.

[0042] In this way, by applying natural language processing technology to building frames, it becomes possible to mathematically and unify the features related to building frames and their components.

[0043] Next, a modified example will be described.

[0044] In the modified example, a case will be described in which the estimation accuracy of the frame member cross section estimation NN in the first embodiment is improved. Specifically, an example in which TanimotoScore is used will be described. TanimotoScore has already been proven to be extremely effective in assessing the similarity of buildings. However, the previous version of TanimotoScore incorporated the concept of distribution width for structural member lengths, and calculated the discrete structural member lengths by representing the member length distribution in a distributed manner, converting them into a kind of fixed-length vector. The natural language processing described above can convert component lengths into fixed-length vectors without the concept of distribution width. Therefore, the distribution of component lengths used in the evaluation of the TanimotoScore can be used to express the building frame as a sum of fixed-length vectors. In other words, the building frame can be expressed as a fixed-length vector. In this modification, the normalized fixed-length vector (with a vector length of 1) of the building frame defined in this way is called the TanimotoVector.

[0045] The TanimotoVector can be calculated as shown in FIG.

[0046] Specifically, the building quantity acquisition unit 121 further acquires TanimotoVector g as the building quantities. The building quantity documentation unit 122 documents the building quantities such that the TanimotoVector of the building to which the component cross section belongs is g, the building height is a meters, its installation level is b meters, the beam (column) component length is c meters, the end joint type is d, and the component length with the end jointed is e meters, f meters.

[0047] The building quantity distributed representation unit 123 performs vector addition of the TanimotoVector to the target component V vector to generate an added target component V vector. The building quantity distributed representation unit 123 performs a frame component cross section estimation NN by associating a OneHot vector of the component cross section index with the added target component V vector. An example of this is shown in Figure 11.

[0048] In this way, according to the modified example, by adding TanimotoVector to the fixed-length vector used in the cross-section estimation NN, information about the building frame to which the member belonged is added, thereby improving the accuracy of cross-section estimation of the frame member.

[0049] Note that a part of the information processing device 100 in the above-described embodiment may be realized by a computer. In this case, a program for realizing this control function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed to realize the control function. Note that the "computer system" here refers to a computer system built into the information processing device 100, and includes hardware such as an OS and peripheral devices.

[0050] Furthermore, "computer-readable recording media" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording media" may also include devices that dynamically store programs for a short period of time, such as communication lines used when transmitting programs over networks like the Internet or communication lines like telephone lines, or devices that store programs for a fixed period of time, such as volatile memory within computer systems that serve as servers or clients in such cases. Furthermore, the above-mentioned programs may be programs that realize some of the aforementioned functions, or may be programs that can realize the aforementioned functions in combination with programs already stored in the computer system.

[0051] Furthermore, part or all of the information processing device 100 in the above-described embodiment may be realized as an integrated circuit such as an LSI (Large Scale Integration). Each functional block of the information processing device 100 may be individually implemented as a processor, or part or all of the functional blocks may be integrated into a processor. Furthermore, the integrated circuit implementation method is not limited to LSI, and may be implemented using a dedicated circuit or a general-purpose processor. Furthermore, if an integrated circuit implementation technology that can replace LSI emerges due to advances in semiconductor technology, an integrated circuit based on that technology may be used.

[0052] One embodiment of the present invention has been described in detail above with reference to the drawings, but the specific configuration is not limited to that described above, and various design changes and the like can be made within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0053] 100 Information processing device 11 Communications Department 12 Control Unit 121 Building Quantity Acquisition Department 122 Building Quantity Documentation Department 123 Building Quantity Distributed Representation Unit 124 Frame member cross section estimation section 125 Learning Department 13 Storage section 14 Input section 15 Output section SYS Information Processing System

Claims

1. an acquisition unit for acquiring building quantities; a building quantity documentation unit for documentation of the building quantities into a document including the building quantities; a building quantity distributed representation unit that distributes the documented building quantities; a frame member cross section estimation unit that estimates a frame member cross section of a target member based on the distributed representation of the building quantities; Equipped with Information processing system.

2. The building quantities include at least the height of the building, the installation level of the target component, the length of the beam, the type of joint at the end, and the length of each jointed component. The information processing system according to claim 1 .

3. The distributed representation is vectorization. The information processing system according to claim 2 .

4. the structural member cross-section estimation unit estimates the structural member cross-section of the target member by inputting the vectorized building quantities into a learning model that has learned correspondence relationships between the cross-sections of structural members, the building quantities, and the vectorized building quantities; The information processing system according to claim 3 .

5. an acquisition unit for acquiring building quantities; a building quantity documentation unit for documentation of the building quantities into a document including the building quantities; a building quantity distributed representation unit that distributes the documented building quantities; a frame member cross section estimation unit that estimates a frame member cross section of a target member based on the distributed representation of the building quantities; Equipped with Information processing device.

6. An information processing method executed by a computer of an information processing device, an acquisition step of acquiring building quantities; a building quantity sentence generation step for generating sentences including the building quantities; a building quantity distributed representation step of distributing the documented building quantities; a frame member cross section estimation step of estimating a frame member cross section of a target member based on the distributed representation of the building quantities; An information processing method comprising:

7. The computer of the information processing device an acquisition step of acquiring building quantities; a building quantity sentence generation step for generating sentences including the building quantities; a building quantity distributed representation step of distributing the documented building quantities; a frame member cross section estimation step of estimating a frame member cross section of a target member based on the distributed representation of the building quantities; A program to execute.

Images