Building process support system, method, and program

Abstract

To enable various human resources to quickly create building model data as they like.SOLUTION: When a grid label 21 for specifying coordinates on a page 6 of a terminal 2 by character codes is included in an inputted message 11 of a building structure, a position information process 22 for making a natural language process 30 generate an instruction code 31 with the grid label 21 as position information of the structure, and a three dimensionalization process 15 for display-controlling each structure of a building model datum 41 read from a building model DB40 on a canvas 8 in the page 6 as a three dimensional datum 13 of a position by the grid label 21 are provided.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a construction process support system, method and program that improves the construction process by enabling natural language messaging. [Background technology]

[0002] In order to move architectural designs from hand-drawn to CAD and then to more usable architectural models, systems to support designers and people involved in the construction process, as well as architectural models such as BIM (Building information Modeling), have been proposed. For example, Patent Document 1 discloses a method (paragraphs 0039 and 0045) in which a floor plan is specified by specifying cells in a grid pattern (FIG. 11). Patent Document 2 discloses a method for generating a design drawing using fuzzy inference in response to an abstract linguistic input (FIG. 2(a)) (paragraph 0026). Patent Document 3 discloses a method for combining machine learning and BIM to generate multiple BIMs in response to input of images and text indicating the exterior and interior design (paragraph 0058), evaluate whether each BIM model matches the content of the images and text (paragraph 0060), and select the BIM model with the highest evaluation (paragraph 0061), thereby designing the exterior and interior design to match the individual's sense of the image or text (paragraph 0020). Patent Document 4 discloses a method for positioning a floor plan on a geometric grid (paragraph 0016) by arranging design elements on a grid (FIGS. 1 and 4, paragraph 0011). Patent Document 5 discloses a method of using a learning model to interact with a user about the contents of a design drawing regarding changes to dimensions or the delivery of large musical instruments after the design drawing has been registered.

[0003] In addition, the concept of BIM and examples in Japan are reported in non-patent literature. Non-patent document 1 discloses a methodology and examples of project management using BIM. For example, BIM does not have a scale and models at a 1:1 ratio, but as has become a specification of the American Institute of Architects, levels of development (LoD) are defined, and scales that are suitable for each of the five stages of development are proposed: conceptual design, design and development, architectural documentation, fabrication drawings, completion and construction, and operation (p. 31). The use of architectural model data such as BIM not only reduces administrative burdens such as reporting, but also refines designs to be cost-effective and enables design changes to be made at optimal costs as the project progresses. This is explained as utilizing three-dimensional data as 3D, 4D by adding a time axis, and 5D by adding a cost axis (pp. 36-40). This non-patent document 1 also discloses the IFC (Industry Foundation Class) schema for BIM documentation, for example, for information exchange in the IFC model according to the openBIM standard (pp. 74-87). Regarding terminology, there is the buildingSMART Data Dictionary (bSDD), a service provided by buildingSMART, which provides translations of the following terms into English, German, French, Japanese, etc. (p.89). Objects (walls, doors, floors) Attributes (fire prevention section, cost) Actors (designers, piping engineers, BIM managers) Phases (design and development, bidding, construction) BIM related (clash detection, cost calculation) Classification system (Omniclass, Uniclass, eBKP)

[0004] Regarding the introduction of BIM, Non-Patent Document 1 reports that, based on the results of an informal survey conducted in the United States, there was an average 25% to 50% drop in productivity when new BIM tools were introduced, and it took an average of 3 to 4 months to return to the previous production level (pp. 102-103). Operational level activities are broad and varied and include everything from model creation, design analysis, inventory preparation (e.g., quantity takeoff), progress reporting, and tracking of changes and defects in the field.

[0005] Non-patent document 2 discloses an example of BIM usage from the perspective of facility management. For example, in a verification of a model development case aimed at data-driven building operation, it was disclosed that operational BIM selected and discarded information that was unnecessary for operation from the information created during design and construction (pp. 16-17). In addition, examples of BIM use are disclosed, such as quantity estimation for large-scale apartment building repair work, construction history management, and the use of elevation drawings that only show piping for emergency response (pp. 20-23).

[0006] In addition, although not limited to the field of architecture, his work is known for its application of large-scale language models and explanations of mathematical understanding. For example, Non-Patent Document 3 discloses methods for named entity recognition, summary generation, sentence embedding, and question answering as examples of using large-scale language models. As an example of question-answering, the book discloses example program code that sends a question such as "What is the highest mountain in Japan?" as an API request to OpenAI's ChatGPT, and displays the reply on the device (p. 261). It also discloses a method for calling the API with the question message as an argument, by defining a string of characters such as "You will now be asked to answer the quiz," or "We will give you a question, so please briefly output only the answer," as a prompt on how to answer the question before answering it (pp. 266-267).

[0007] Non-Patent Document 4 discloses a research example of mathematical analysis of a large-scale language model. For example, on page 37, an example of a conventional machine learning method is given: approximation of any function using trigonometric functions in Fourier transform. In this case, a neural network with only two layers is sufficient, and no multi-layers are required. However, large-scale language models are multi-layered, and the analysis suggests that the success of multi-layer neural networks may be due to their ability to represent functions with non-smooth jump structures (p. 39). Multi-layer neural networks also have smaller approximation errors for functions whose smoothness or period changes discontinuously. Multilayer neural networks using large-scale language models can approximate the complexity of natural language, which cannot be approximated by ordinary trigonometric functions, and can store it as a set of parameters in a vector space. Although it is difficult to fully control the information processing that determines the parameters of a multilayer neural network, it is known that if you provide the correct answer and then perform a process to reduce the error in the reverse direction (backpropagation), the values ​​of the large-scale multilayer parameter set will become values ​​that closely approximate the training data.

[0008] Large-scale language models can convert natural human sentences into other languages ​​or codes, or generate plausible, highly probable sentences that follow a question, by probabilistically predicting the words (tokens) that will appear next in a sentence. Currently, dialogue with large-scale language models has evolved to the point where it is no longer easy to determine whether the person is a real person. On the other hand, when asked a question about something they have not learned, they generate sentences that appear in a formally plausible style but have completely different content, or sentences that are non-existent answers (hallucinations, hallucinations).

[0009] The least squares method is also a technique for approximating a function by minimizing the error with the data, but it is known that the resulting formula cannot be used to make predictions when applied to events outside the range of the sample data.Even with the backpropagation method, which reduces errors, it can be assumed that hallucination will occur outside the training data, which is technically unavoidable. Non-Patent Document 3 states regarding hallucinations that "even when the (large-scale language) model does not have the knowledge necessary to answer, it generates tokens (omitted) and behaves in such a way that it outputs (as a hallucination) the names of restaurants that appear to be plausible in the training corpus" (p. 76). At the time of filing, no method had been proposed to fundamentally eliminate hallucination in large-scale language models, and the patent attempts to prevent hallucination from occurring by providing some kind of guidance in individual applications and operating them within the scope of training data. Large-scale language models can also generate numbers that are similar in meaning or form as part of a sentence (tokens) without actually performing any calculations, making it difficult to even calculate the sum of numbers that appear in a chronological order in a chat.

[0010] If architectural model data such as BIM could be used for architectural design and the entire construction process, costs could be reduced and reports could be automated, allowing people working in the construction industry to focus on what they should be doing and reducing stress. However, introducing BIM and other technologies presents high hurdles not only in terms of cost but also in terms of human resources. As mentioned in the above patent documents, there are proposals to utilize language input, grids, BIM, machine learning, etc. in architectural design, but architectural model data has not yet penetrated into construction sites or construction processes, and not only has productivity not improved, but there is also a labor shortage due to stress caused by a large amount of unnecessary work.

[0011] Although it is known that utilizing architectural model data throughout the entire value chain of construction, including post-construction operations, would increase efficiency, the entry hurdles for inputting architectural model data and digitizing it for usability are extremely high, and this approach has not yet taken hold. Furthermore, there is no known method for applying large-scale language models to support the creation and design of architectural model data that can be converted into BIM, which is not limited to exterior and interior design. Simply using large-scale language models makes it difficult to consistently generate consistent architectural model data, as the handling of numerical values ​​such as dimensions is unstable. It is also impossible to control the possibility of generating hallucination data that is too unnatural for an architectural model, which in turn results in verification taking more time. [Prior art documents] [Patent documents]

[0012] [Patent Document 1] Japanese Patent Application Publication No. 08-006989 [Patent Document 2] Japanese Patent Application Publication No. 08-293036 [Patent Document 3] Japanese Patent Application Publication No. 2019-200721 [Patent Document 4] Special Publication No. 2020-514905 [Patent Document 5] Patent No. 7440053 [Non-patent literature]

[0013] [Non-Patent Document 1] "The BIM manager: A practical guide to BIM project management" by Mark Baldwin, edited and translated by buildingSMART Japan, buildingSMART Japan, June 2022 [Non-patent document 2] "BIM Utilization Case Studies for Facility Management," by the Japan Facility Management Association and the BIM / FM Research Group, published by the Japan Facility Management Association in July 2022. [Non-patent document 3] "Introduction to Large-Scale Language Models," edited by Ikuya Yamada, Masatoshi Suzuki, Kosuke Yamada, and Linghan Li, published by Gijutsu Hyoronsha in August 2023. [Non-patent document 4] Masaaki Imaizumi, "Approaching the Principles of Deep Learning: A Mathematical Challenge," Iwanami Shoten, April 2021 Summary of the Invention [Problem to be solved by the invention]

[0014] [Technical Issues] The above-mentioned conventional examples and their combinations have the disadvantage that it is difficult to widely spread the input of architectural model data. Furthermore, the above-mentioned conventional example has the disadvantage that it is not possible to create (design) architectural model data that is consistent as an architectural model while using a large-scale language model. For example, even if one tries to apply large-scale language models to architectural design, the handling of coordinate values ​​and units tends to be inconsistent, making it difficult to put into practical use.

[0015] [Objective of the Invention] The objective of the present invention is to provide a construction process support system, method and program that can flexibly respond to the needs of clients and improve the entire construction process by enabling a variety of people to quickly create architectural model data as they wish. [Means for solving the problem]

[0016] [Point of View] The inventor of this invention has conducted various thoughts, research and experiments, gained insight into the challenges facing the construction industry, and discovered the relationship that in order to eliminate inefficiencies associated with construction and allow people to focus on traditional, essential and creative work, it is first necessary to make it possible to create architectural model data that is consistent as an architectural model simply by giving instructions naturally. Therefore, we came up with the idea that in order to accumulate consistent architectural model data while utilizing the inherent functions of large-scale language models to enable design in natural language, we first needed to devise a way to handle the coordinate values ​​of structures.

[0017] [Means for solving the problem 1] Therefore, the first group of the present invention corresponding to the first embodiment comprises a drawing processing unit (10) connected to a user's terminal (02) and controlling the display of a page (06) that serves as a user interface, a controller (20) that controls an input message (11) input as natural language data from the terminal (02) to the drawing processing unit (10), a natural language processing unit (30) that generates an instruction code (31) that describes architectural model data (41) by referring to the input message (11), and an architectural model database (40) that stores the architectural model data (41) corresponding to the instruction code (31) in a searchable manner. The controller (20) is provided with a location information processing (22) that causes the natural language processing unit (30) to generate the instruction code (31) that sets the grid label (21) as location information of the structure when the grid label (21) that specifies the coordinates on the page (06) by a character code is included in the input message (11) of the architectural structure. Furthermore, the drawing processing unit (10) is configured to include a chat processing unit (14) that controls the display of the input message (11) in the chat area (07) of the page (06), and a three-dimensionalization processing unit (15) that controls the display of three-dimensional data (13) of the position of each structure of the architectural model data (41) read from the architectural model database (40) on a canvas (08) within the page (06) based on the grid label (21) of the structure. This solved the technical challenge of creating consistent architectural model data using only natural instructions.

[0018] [Means for solving the problem 2] The second group of the present invention corresponding to the second embodiment includes a complementation process (23) in which the controller (20) causes the natural language processing unit (30) to refer to predetermined standard data (51) of architecture, and complements the architectural model data (41) that is missing in the input message (11) about the structure with a context corresponding to the conversation with the user using the natural language data. The drawing processing unit (10) is configured to include a surface processing unit (16) that processes the surface of the three-dimensional data (13) in accordance with the material of the structure designated by the user or the material that has been complemented. This solves the technical problem of making it possible to create just the right amount of architectural model data.

[0019] [Solution to Problem 3] The third group of the present invention corresponding to the third embodiment is configured such that the controller (20) has an input prompting process (28) that causes the natural language processing unit (30) to refer to design process data (54) predetermined as a standard design process, and generates a reply message (12) that prompts input of the content to be designed following the state of the architectural model data (41). This solved the technical challenge of enabling a diverse range of people to carry out natural, high-quality design using standard processes without stress. [Effects of the Invention]

[0020] When the meaning of the terms in each claim is interpreted in consideration of the description and drawings of this specification and the inventions according to each claim are identified, the present invention functions as described below and has the following advantageous effects in relation to the above-mentioned background art, etc.

[0021] [Effect 1 of the Invention] In the construction process support system of problem-solving means 1, when the input message (11) includes a grid label (21), the location information processing (22) of the controller (20) causes the natural language processing (30) to generate an instruction code (31) that uses the grid label (21) as location information for the structure, and the natural language processing unit (30), under the control of the controller (20), references the input message (11) and uses the grid label (21) to generate an instruction code (31) that describes architectural model data (41). The drawing processing unit (10) converts each structure of the architectural model data (41) read from the architectural model database (40) into three-dimensional data (13) of the position of each structure according to the grid label (21) of the structure by its three-dimensional processing (15) on the canvas (08) in the page (06), and controls the display. Therefore, the location information can be handled as a character code called a grid label rather than as a numerical value in the input message (11) in natural language, the instruction code (31) for operating the architectural model database (40), the architectural model data (41), and the three-dimensional data (13), and the location information can be handled while maintaining the same without being changed by natural language processing from the input message (11) to the three-dimensional data (13). Therefore, the natural language processing unit (30) accepts the creation of architectural model data in the input message (11) in natural language, while strictly handling coordinate values, and as a result, it is possible to design and accumulate architectural model data that is consistent as an architectural model. In this way, the present invention can accumulate consistent architectural model data while using the natural language processing unit (30) as a component, making it easier to carry out trial-and-error designs that meet the client's needs and requests, allowing humans to focus on the work they should be doing, and accumulating architectural model data (41) that increases the satisfaction of clients and other clients.

[0022] [Effect of the Invention 2] In the construction process support system of the problem-solving means 2, the complementation process (23) causes the natural language processing unit (30) to refer to the standard data (51) to complement the architectural model data (41) that is lacking in the input message (11) regarding the structure, so that the user does not need to input the fixed details of the structure's attributes, and can create architectural model data using only the minimum necessary essential content in the input message (11) in natural language. Then, a three-dimensional processing (15) generates a set of structures as three-dimensional data (13) using the coordinate values ​​of the grid labels (21). At this time, a surface processing (16) processes the surface of the structure according to the material of the structure specified or supplemented by the user, so that the surface of the three-dimensional data (13) is processed with the material of each structure based on the latest information of the architectural model data (41), and therefore the architectural model can be visually confirmed as three-dimensional data (13) with a surface texture that visually indicates that concrete, wood, etc. have been specified. In this way, the interpolation process (23) interpolates from the standard data (51), so the user can register the architectural model data that is consistent with the minimum of instructions. This allows the creation of architectural model data (41) using natural language, while accumulating architectural model data (41) that is of sufficient quality as an architectural design, can be utilized throughout the entire value chain of the construction process, and can be useful in operations after completion.

[0023] [Effect of the Invention 3] In the architectural process support system of the problem-solving means 3, the input prompting process (28) causes the natural language processing unit (30) to refer to the design process data (54) and generates a reply message (12) that prompts input of the content to be designed following the state of the architectural model data (41). This makes it possible to prompt input in a standard and desirable order for the design process, and also to ask the user for judgment so that there are no missing items in the architectural model data (41). The user can naturally know the state of the latest architectural model data (41) according to the inputs made up to now through dialogue via messages and the contents to be input, and can consider the essential items in the standard order of design in the desired order without stress. This reduces unnecessary work during the design phase, allowing a diverse range of people to carry out natural, high-quality designs using standard processes without stress, providing users with a productive and fulfilling work experience. Furthermore, by answering questions in a natural language orderly manner, without the need for advanced IT or CAD training, it is possible to create architectural model data (41) that is consistent with the data. This allows a diverse range of people to quickly create architectural model data as they wish, thereby enabling the widespread use of architectural model data (41) throughout the construction industry. [Brief explanation of the drawings]

[0024] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of a construction process support system that controls grid labels (Example 1). [Figure 2] FIG. 2 is an explanatory diagram showing an example of a page (Example 1). [Figure 3] FIG. 3 is an explanatory diagram showing an example of architectural model data (Example 1). [Figure 4] FIG. 4 is a flowchart showing the process of creating architectural model data using natural language and grid labels (Example 1). [Figure 5] FIG. 5 is a block diagram showing an example of the configuration of a construction process support system that references standard design data (Example 2). [Figure 6] FIG. 6 is an explanatory diagram showing an example of standard design data (Example 2). [Figure 7] 7 is a sequence diagram showing the process of creating architectural model data by interpolation processing with reference to standard design data (Example 2). [Figure 8] FIG. 8 is an explanatory diagram showing an example of an error message (Example 2). [Figure 9] 9 is a flowchart showing an example of processing based on an error message (Example 2). [Figure 10] FIG. 10 is a block diagram showing an example of the configuration of a construction process support system that references design process data (Embodiment 3). [Figure 11] FIG. 11 is an explanatory diagram showing an example of a page before placing a living room on the canvas (Example 3). [Figure 12]FIG. 12 is an explanatory diagram showing an example of a page in which a living room is arranged on a canvas (Example 3). [Figure 13] FIG. 13 is an explanatory diagram showing an example of a page in which the position of the living room has been changed (Example 3). [Figure 14] FIG. 14 is an explanatory diagram showing an example of a page in which a pillar is placed in the corner of the living room (Example 3). [Figure 15] FIG. 15 is an explanatory diagram showing an example of a page in which a reply about the state of the pillar in the living room is sent to the chat box (Example 3). [Figure 16] FIG. 16 is an explanatory diagram showing an example of a page from which one of the running heads has been deleted (Example 3). [Figure 17] Figure 17 is a management design sheet that explains the industrial applicability of this embodiment (Examples 1, 2, 3, and 4). DETAILED DESCRIPTION OF THE INVENTION

[0025] Four embodiments are disclosed as modes for carrying out the invention. The first embodiment is a construction process support system, method, and program that uses grid labels 21. The second embodiment is a construction process support system, method, and program that performs the complementation process 23 by referring to the standard data 51 shown in Fig. 5 etc. Examples of using the conversation log 71 and the error message 61 are also disclosed as the second embodiment. The third embodiment is a construction process support system, method, and program that promotes user input by referring to design process data 54 shown in FIG. 10 and the like. The fourth embodiment has the functions of the first to third embodiments, and further has practical functions added thereto. Examples 1 to 4 are collectively referred to as embodiments.

[0026] The system, method and program inventions according to the present embodiment are software-related inventions, and require hardware resources for their implementation. Each unit and process in the embodiment is a group of programs that can be executed by the processor of one or more computers. The computer has a processor (arithmetic unit), memory (main storage device and auxiliary storage device), bus, input / output device, network control, etc. A program may directly execute instructions from a processor, but it can also be written in a programming language that is based on an operating system, browser, or various APIs, and executed as is, or compiled into executable code. The system and method according to this embodiment may be executed on one computer or on multiple computers connected via a network 4. For example, the controller 20 can be implemented on the operating system of a computer connected to the display of the terminal. Display control of the three-dimensional data 13 by the drawing processing unit 10 can be implemented by an API of a script language that runs on the browser of the terminal 2. The natural language processing unit 30 can implement its basic functions by accessing a server that provides a service for large-scale language models via the API. The architectural model database 40 may also be installed on a computer connected to the terminal 2, or may be provided via the network 4 using a cloud service.

[0027] The controller 20, the drawing processing unit 10, the natural language processing unit 30, and each of the processes 14, 15, 22, etc. each perform information processing using a computer processor or the like. In other words, each unit and process uses hardware resources such as a processor and memory to temporarily store input data, perform some kind of calculation, and output the data. Those skilled in the art will be able to create and use a program that reproduces an embodiment of the invention by referring to the disclosure of this embodiment while using the original functions of a computer, operating system, browser, API, and hardware resources. For example, in the embodiment, the natural language processing unit 30 uses limited functions that can be easily used by a person skilled in the art at the time of filing the application, and therefore the natural language processing unit 30 can be realized by accessing a general chat service with a large-scale language model via an API and adding the configuration of the embodiment. Since the elements of the embodiments all use hardware resources, there is an inventive step in utilizing hardware resources.

[0028] The controller 20, each part, and each process in the embodiments and examples solves the above-mentioned problems by organically cooperating while utilizing hardware resources, going beyond the original functions that operating systems, browsers, databases, APIs, large-scale language models, etc. normally have at the time of filing. For example, at the time of filing, the inherent functions of a large-scale language model are that it can interact with a user in natural language and generate executable machine-readable code such as programs and queries. In contrast to this original function, in the embodiment, a useful and novel configuration is achieved by combining the reproducible element technologies disclosed in this specification, such as the grid labels 21, instruction codes 31, architectural standard data 51, and error messages 61 of each example, without using a special large-scale language model, special pre-learning (fine-tuning), or artisan-like prompts (input messages 11), and this new system configuration solves each of the above-mentioned problems. These techniques will be disclosed in detail below as four examples. Database is abbreviated as DB. [Example]

[0029] Referring to FIG. 1, the construction process support system of the first embodiment includes a drawing processing unit 10, a controller 20, a natural language processing unit 30, and a construction model DB 40. The drawing processing unit 10 is connected to the user's terminal 2 via a network 4 and controls the display of a page 6 that serves as a user interface. The controller 20 controls an input message 11 input as natural language data from the terminal 2 to the drawing processing unit 10. In this embodiment, the natural language data is input, for example, by text input or voice input. The text input or voice input may be performed by a human user, or a system that reads an existing design drawing and expresses it in natural language may be used as a virtual user, and the virtual user may input text or voice. The input message 11 may be input using data or signals that can input natural language, such as gaze, electromyogram, or electroencephalogram. The controller 20 controls the input message 11 by, for example, receiving the input message 11 via the network 4 and analyzing the contents. The natural language processor 30 refers to the input message 11 and generates an instruction code 31 that describes the architectural model data 41 . The architectural model database 40 stores architectural model data 41 in a searchable manner according to instruction codes 31. The instruction codes 31 are queries that operate the architectural model database 40, and these queries update the architectural model data 41 in a specific structure as shown in Fig. 3. Therefore, the instruction codes 31 can be considered as codes in a markup language that describe the architectural model data 41 in a machine-readable manner.

[0030] In this first embodiment, the controller 20 includes a location information processor 22 that handles grid labels 21 . Grid labels 21 are labels that specify coordinates on page 6 using character codes, and are character code labels that specify the coordinate values ​​of the dividing lines that divide the x and y coordinates at equal intervals. It is advisable to specify coordinate values ​​using labels such as B2 by combining the letters A, B, and C with the numbers 1, 2, and 3. It will be smoother if a specific sequence of character codes that can be handled with regular expressions, such as B2 or A3, can be treated as grid labels 21. On the other hand, it is also possible to train the system to recognize characters at positions that refer to location information as the context of input message 11 as grid labels 21.

[0031] When this grid label 21 is included in the input message 11 of an architectural structure (object), the location information processing 22 causes the natural language processing 30 to generate an instruction code 31 that designates this grid label 21 as location information of the structure. The "generation" control by the location information processing 22 may be performed by having the natural language processing unit 30 learn the grid label 21 and its processing when the natural language processing unit 30 is started, or by determining whether the input message 11 contains a grid label 21, and if so, adding to the input message 11 that the grid label 21 is location information, thereby causing the natural language processing unit 30 to generate the grid label 21. In the instruction code 31, the character code used by the user, such as B2 or A3, may be written as is as the location information, or may be converted into another expression. If the absolute value of the grid label 21 is fixed, the description of the grid label 21 common to the input message 11, reply message 12, instruction code 31, and architectural model data 41 can be used as the location information as is. If the position of a structure deviates from the grid label 21, it is advisable to treat the offset amount from the grid label 21 as a numerical value or a ratio. When treating it as a ratio, the grid label 21 may be hierarchized, such as dividing the area between A and B into multiple sections, such as the b in "Ab3," and making it the position of one of the b sections. In the example shown in Figure 3, the pillar is located at the position of grid label 21, and the offset amount is 0 [mm]. When specifying a position using multiple grid labels 21, such as the living room, the position can be specified with a single offset amount for the entire area, or the offset amount from the position of each grid label 21 can be handled.

[0032] The input message 11 is written in natural language, and no specific format is required except for the grid label 21. Natural language refers to the language used by humans on a daily basis, including written and spoken language. The natural language processing unit 30 (large-scale natural language model) converts this natural language into a form that can be understood by a computer. Specifically, it analyzes the text and voice input by the user and generates a reply (response) or action based on the content. This allows the user to interact with the construction process support system through intuitive communication.

[0033] However, large-scale language models are unstable in the handling of defined data items and in the management of physical values ​​such as length, and therefore the original functions of large-scale natural language models are not capable of stably generating architectural model data 41. For this reason, in the first embodiment, the creation and design of architectural model data 41 is stably supported by linking the functions of the drawing processing unit 10 and the controller 20 with the natural language processing unit 30 (large-scale language model).

[0034] Referring to Figure 2, page 6 is an area displayed on the display of terminal 2, and includes a chat area 7 that displays input messages 11, etc., and a canvas 8 that visually displays coordinates (location information) in a grid label 21 that is easy for users to understand. In the example shown in Figure 2, grid labels 21 specify coordinates in the horizontal direction (x direction) of the figure in alphabetical order A, B, C, and in the vertical direction (y direction) of the figure in numerical order 1, 2, 3. For example, a point in xy coordinates can be specified by combining x and y labels to "D3." For example, an instruction (input message 11) to "place the living room at positions D3 to G5" specifies the coordinate value of the living room on the plane. The living room is an example of three-dimensional data 13 (an object of an architectural structure).

[0035] The drawing processing unit 10 shown in FIG. 1 includes a chat processing unit 14 and a three-dimensional processing unit 15. The chat process 14 controls the display of an input message 11 in the chat area 7 of the page 6. The chat process 14 alternately displays the input message 11 and a reply message 12 (system response) in the chat area 7, thereby displaying a dialogue with the construction process support system (system). The three-dimensional processing 15 controls the display of three-dimensional data 13 of the positions of each structure in the architectural model data 41 read from the architectural model DB 40 according to the grid labels 21 of the structures on the canvas 8 within the page 6. In Figure 2, the living room is displayed in three dimensions.

[0036] When page 6 is displayed on the browser of terminal 2, the rendering processing unit 10 can execute a program using an API capable of interactively rendering two-dimensional and three-dimensional computer graphics, such as WebGL (Web Graphics Library). The original function of a program using WebGL is to interactively display 3D object data, but the rendering processing unit 10 of the first embodiment is characterized by information processing that converts architectural model data 41 read from an architectural model DB 40 into three-dimensional data and controls the display of grid labels 21 on canvas 8 as position information.

[0037] A database (DB) comprises a recording medium that electromagnetically and physically holds data representations (bits), a cache, an input / output control unit that controls the input and output of data, and a database management system that logically manages the IDs and addresses of data such as records.

[0038] The Architectural Model DB40 can use relational databases (RDBs) that store data in a data structure that can be expressed in an ER diagram and perform registration, update, reading, etc. using SQL statements while maintaining ACID properties, or NoSQL databases that handle large amounts of data flexibly and quickly while sacrificing consistency.

[0039] In contrast to the general original functions of a database, the architectural model DB 40 is characterized in that it stores data in the data structure of architectural model data 41 specific to the embodiment as shown in Fig. 3 and responds to access requests. Therefore, even if a database management system is not used, simply storing text files in XML or JSON format, for example, corresponds to the database of the embodiment.

[0040] Because the natural language processing unit 30 processes probabilistic information, if the architectural model DB 40 is an RDB, the number of formal errors related to database operations will increase, but it will be possible to record data in a data structure with clearly defined keys. If a NoSQL database or temporary JSON files or XML files are used to temporarily assign meaning to data by tag for each item, sacrificing duplication and consistency, errors will be reduced when processing the input message 11, but it will be difficult to "always maintain formal consistency" as architectural model data 41. However, if the structure shown in Figure 3 can be ensured, "consistency as an architectural model" can be maintained.

[0041] As various embodiments, for example, in applications where architectural model data 41 is designed by trial and error, the architectural model DB 40 may be constructed using NoSQL or XML, and once a group such as a floor is completed, the data may be re-stored in a separate RDB, while ensuring formal consistency. The architectural model DB 40 may be realized solely with NoSQL by enhancing the function of reviewing the consistency of the architectural model data 41 according to the progress of the design. The system, method, and program of the first embodiment can be fully implemented with NoSQL even for large-scale buildings. When collaboration or synchronization with other RDBs is required, synchronization will be smoother if the architectural model DB 40 is an RDB. For input synchronization, if existing two-dimensional design drawing data is stored in an RDB, the architectural model DB 40 of the embodiment may also be an RDB, and the data may be read into the architectural model DB 40 using RDB inter-database reading, and then converted into three dimensions using natural language. Of course, existing two-dimensional drawing data may be written in JSON or the like, and read by the natural language processing unit 30 or the like to generate the instruction code 31 in the format shown in FIG. 3. In any case, before updating the architectural model data 41, the three-dimensional data 13 is displayed to the user and visual confirmation is requested, so consistency in terms of the content of the architectural model data 41 can be accumulated sequentially regardless of the database format. Depending on the embodiment, it may be advisable to separately implement a review process for determining whether or not there is interference, which would be cumbersome to visually check. Furthermore, in Example 1, the architectural model database 40 stores architectural model data 41 corresponding to the instruction code 31 in a searchable manner, and therefore the natural language processing unit 30 can generate reports, two-dimensional drawings, etc. that refer to the architectural model data 41 that the designer is finalizing by using a natural language processing (RAG) technique that refers to external information. For example, when automatically generating reports in natural language or three-dimensional data 13 about progress and content, such as weekly reports from designers or work reports from people involved in the construction process that refer to architectural model data 41, it is possible to search for architectural model data 41 with reliable numbers, preventing hallucination and generating content with sufficient accuracy for business use.

[0042] An architectural model is a set of data structures that can represent buildings and structures in a three-dimensional digital format on a computer, and is a collection of structures that are 3D objects. 3D objects (structures) are three-dimensional object information such as building components, fixtures, equipment, piping, and materials that make up an architectural model, and in this embodiment, meaningful spaces such as living rooms and kitchens are also treated as structures.

[0043] Referring to FIG. 3, the architectural model data 41 has a structure including an ID, a name, a position, a shape (height), a type, and a material for each structure. In terms of hierarchical structure, floors are at the top, and structures (objects) belong to each floor. The ground level at height 0 [mm] is id1, the first floor is id2, and IDs are assigned sequentially for each type of structure, such as "room | place" belonging to the first floor. In the example shown in Figure 3, all structures can be uniquely identified by floor level, structure type, and ID. As an alternative method, consecutive IDs can be assigned to all structures, but with the configuration shown in Figure 3, the ID is displayed when the structure is displayed in three-dimensional data 13, and when used in dialogue with the user, the ID values, such as pillar number 1, pillar number 2, etc., can be used directly in natural language dialogue and displayed on the screen.

[0044] 3 records values ​​of attributes (size, height, etc.) specified in the input message 11 or automatically assigned by the completion process 23 of the second embodiment, but items that should be specified for the structure even if they do not have a value may be stored as having no value or as a blank value. With such a data structure, the natural language processing unit 30 can identify items that are currently missing for each structure by referring to the latest architectural model data 41 (which can be used in the input prompting process 28 of the third embodiment).

[0045] The instruction code 31 is an operation instruction to the database generated by the natural language processing unit 30, and in this embodiment, it is a code for operating the architectural model DB 40 in order to describe an architectural model (structure). This instruction code 31 is generated in response to an input message 11, which is a request from the user, and is passed to the controller 20. The controller 20 executes this instruction code 31 on the architectural model DB 40 to operate the architectural model DB 40 and add, update, delete, etc. the architectural model data 41 .

[0046] The instruction code 31 is a query to the database, and if the architectural model DB 40 is a relational database, it is an SQL statement, but if it is a noSQL database, there are various types. The instruction code 31 is a query to the database and also a code in a language that describes the architectural model. There are various ways to describe the architectural model, such as BIM's IFC (Industry Foundation Classes) file, but as shown in Figure 3, a description system specific to the embodiment can be adopted, which has a hierarchical structure and prepares the type of structure.

[0047] The instruction code 31 may be executable code that reads a JSON file, updates the target, and outputs a string that becomes a JSON file again, or it may be a query that updates a specific item in a database. The natural language processing unit 30 is a large-scale language model with functionality at the time of filing, and by providing the format of the instruction code 31 and architectural model data 41 at startup, it can output the instruction code 31 corresponding to the input message 11 in natural language. Error handling when database operation using the instruction code 31 fails is disclosed in Example 2 (e.g., Figures 8 and 9).

[0048] In this embodiment, the natural language processing unit 30 (large-scale language model) is treated as a commercially available component that converts an input message 11 written in natural language into an instruction code 31. Converting natural language into code is an inherent function of a large-scale language model, and in this embodiment, the system is configured to generate the instruction code 31 for problem-solving while implementing conditions specific to the construction process as data structures and processing contents. As an innovation in each embodiment, a configuration is proposed in which the natural language processing 30 reads not only the input message 11 but also the standard data 51, individual standard data 52, and conversation log 71 in the second embodiment, and the design process data 54 in the third embodiment, in order to generate the instruction code 31, thereby enabling the creation of architectural model data 41 that is consistent without any excess or deficiency.

[0049] The controller 20 operates the architectural model DB 40 . The controller 20 executes an operation (processing 0) on the architectural model DB 40 in accordance with the instruction code 31 generated by the natural language processing unit 30 in order to realize the request of the input message 11 issued by the user in natural language. Specifically, perform the following operations. Retrieving Data: Reading specific information from a database Adding (inserting) data: Adding new information to the database Update data: Change existing information Data Purging: Removing unnecessary information from the database This database operation is performed in accordance with the data structure of the architectural model data 41 shown in FIG.

[0050] Referring to FIG. 4, the construction process support method of Example 1 includes a drawing processing step S10, an input message receiving step S11, a location information processing step S12, a natural language processing step S13 (generation of instruction code 31), an architectural model data storage step S14, and a three-dimensionalization process S15.

[0051] 4, in the construction process support method of the first embodiment, first, the drawing processing unit 10 controls the display of a page 6 that is connected to the user's terminal 2 and serves as a user interface (drawing processing step S10). The drawing processing unit 10 then receives an input message 11 that is input in natural language data in a chat area 7 of the page 6 via the network 4, and controls the display of this input message 11 in the chat area 7 (input message receiving step S11). This input message receiving step S11 is part of the function of the chat processing 14 of the controller 20. The display in the chat area 7 is preferably executed after the input message 11 is received, around the time when the three-dimensional data 13 is displayed, and before the reply message 12 is displayed. In the first embodiment, the generation, display, and display control of the reply message 12 shown in FIGS. 5 and 7 are not essential.

[0052] When the input message 11 contains a grid label 21 that specifies coordinates on the canvas 8 within the page 6 using character codes, the controller 20 generates an instruction code 31 that uses the grid label 21 as location information for the structure (location information processing step S12). Then, the natural language processing unit 30 references the input message 11 and generates the instruction code 31 that describes the architectural model data 41 by natural language processing. At this time, the natural language processing unit 30 specifies the location information of the structure to be described using the grid label 21 in the input message 11 (natural language processing step S13). Next, the controller 20 executes the instruction code 31 to store the architectural model data 41 in the architectural model DB 40 in a searchable manner (architectural model data storage step S14).

[0053] Next, the drawing processing unit 10 converts the architectural model data 41 read from the architectural model DB 40 into three-dimensional data 13 and controls the display on the canvas 8. At this time, the three-dimensional processing 15 generates the three-dimensional data 13 of each structure at the position according to the grid label 21 of each structure (three-dimensional processing step S15).

[0054] Each of these steps can be realized by having a computer processor execute a group of programs (procedures) that cause the computer to operate as each step. In the embodiments, a process expressed as a "step" executes a "procedure" by the corresponding program. Because the disclosure of each step also discloses the procedure by the program that is executed by the processor, redundant disclosure of steps and procedures will be omitted, but in the disclosure of the embodiments, steps can be interpreted as procedures and procedures can be interpreted as corresponding to steps. For example, the system and method of Example 1 can be implemented by a construction process support program that is executed by one or more computers and causes the computer to execute a drawing processing procedure, an input message receiving procedure, a location information processing procedure, a natural language processing procedure, an architectural model data storage procedure, and a three-dimensionalization processing procedure. Details for each procedure are the same as those disclosed for the corresponding steps by name.

[0055] Furthermore, system processes correspond to the corresponding steps and procedures. For example, location information processing 22 is step S12 in Figure 4, and can be realized by executing a program for location information processing procedures. However, while processes are system configurations and the order of processing is arbitrary, steps and procedures differ in that they also have their own unique processing order. Processes, steps, and procedures with the same name share a common role to play in the entire system or method. The relationships between the treatments, steps, and procedures are the same for the second to fourth embodiments.

[0056] As described above, according to the first embodiment, the user can place structures on the canvas 8 using instructions in natural language and grid labels 21, so that people who are unfamiliar with design, people who are unfamiliar with IT, and design experts can all create architectural model data 41 without having to learn complicated operating methods, etc. This lowers the barrier to entry for inputting and registering architectural model data 41, which is the part of the construction process that most needs improvement. Furthermore, since the natural language processing unit 30 generates instruction codes 31 using grid labels 21, it is possible to describe architectural model data 41 using instruction codes 31 with a fixed format while responding to a variety of input messages 11 in natural language, and it is possible to maintain the consistency of the architectural model data 41. Furthermore, since the three-dimensional data 13 is generated from the architectural model data 41 and the display is controlled independently of the natural language processing unit 30, it is easy for the user to visually check whether the content of the architectural model data 41 registered by interpreting the natural language input message 11 is as intended. Furthermore, since the architectural model DB 40 stores the architectural model data 41 in a searchable manner, in combination with the natural language processing 30, it is possible to automatically generate reports and the like that have accurate numerical values ​​and are suitable for business use. Therefore, in Example 1, by storing architectural model data 41 in a searchable database while enabling design in natural language, it is possible to provide a foundation for improving the work of many people involved in a series of processes, from design to actual construction (construction), and subsequent maintenance and management of equipment and fixtures, until the building is completed, and even after it is completed, including its operation. In this way, by using the construction process support system, method and program of the first embodiment, the user can create consistent architectural model data 41 using only natural instructions. [Example]

[0057] Complementary process 23 in Example 2 Referring to FIG. 5, in the construction process support system of the second embodiment, in addition to the configuration of the first embodiment, the controller 20 includes a complementing process 23 and the drawing processing unit 10 includes a surface processing 16. In the example shown in FIG. 5, the controller 20 is provided with a standard data table 50 that stores standard data 51 for the building. The completion process 23 causes the natural language processing unit 30 to refer to predetermined standard data 51 for architecture, and completes the architectural model data 41 that is missing from the input message 11 about the structure with the context corresponding to the conversation with the user using natural language data. Then, the natural language processing unit 30 can include in the instruction code 31 content that is not included in the input message 11 but can be completed by referring to the standard data 51.

[0058] The surface processing 16 processes the surface of the three-dimensional data 13 according to the material of the structure specified by the user or supplemented. The three-dimensionalization processing 15 of the drawing processing unit 10 can identify the texture of the surface of the three-dimensional data 13 according to the material (concrete, wood, etc.) of the structure supplemented by the standard data 51, and render it on the canvas 8. This allows the user to visually confirm information about the material of the structure that they input or supplemented.

[0059] For example, if input data 11 is received assuming a reinforced concrete (RC) structure, and the material of the pillars is not specified, the completion process 23 will specify concrete as the pillar material from the standard data 51 and conversation log 71. Then, the surface processing 16 will render the surface with a concrete texture as three-dimensional data 13. From the user's perspective, this means that the standard required information has been automatically completed with the minimum necessary input, and the user can easily visually confirm this with the 3D data (three-dimensional data 13) displayed on canvas 8, assuming the context of the conversation with the system.

[0060] Referring to FIG. 6, the standard data 51 for the design of the second embodiment stores attributes such as conditions and sizes for each type of structure such as floor level, pillars, walls, and windows. The completion process 23 causes the natural language processing unit 30 to refer to this standard data 51, thereby completing the attributes (size, etc.) of the structure that are not explicitly specified by the user in the input message 11, and generating the instruction code 31 that specifies the architectural model data 41. For example, the user can accumulate architectural model data 41 including height information by simply specifying the position of the pillar in relation to the grid label 21 or an existing structure, without specifying the height of the pillar. In other words, this interpolation process 23 can eliminate excesses and deficiencies in the architectural model data 41. By eliminating deficiencies with the interpolation process 23, the minimum input is encouraged, and input of excessive details is naturally suppressed. Furthermore, when recording the design basis to increase the reliability of the architectural model data 41 and improve data maintainability, it is advisable to record the data items and values ​​supplemented under the control of the supplementation process 23 as part of the architectural model data 41 or in association with it. Compared to conventional line drawings where the meaning of a line cannot be formally and uniquely determined, the architectural model data 41 can manage attribute data such as whether it is a living room or a pillar. Furthermore, if it is possible to record that a data item is supplemented, it is possible to record not only the meaning of the line but also the design basis for how the line was selected.

[0061] Furthermore, if a typical large-scale language model available at the time of filing is used as the natural language processing unit 30, the instruction code 31 can be generated by the complementation process 23 without detailed coding by passing the content shown in Fig. 6 as a text file. Fine tuning may be performed before the entire construction process support system is launched, but the latest standard data 51 can be used by having the controller 20 input the standard data 51, along with instructions on how to use it, into the natural language processing unit 30 as text data when launching the natural language processing unit 30 via an API.

[0062] Then, the complementation process 23 identifies the material of the structure according to the context of the conversation (dialogue), and the surface process 16 automatically identifies the color and texture of the surface of the structure in the three-dimensional data 13 according to the material without any instructions from the user. This allows the user to check whether the three-dimensional data 13, which is three-dimensional and has a color and texture close to the finished image, including the complemented content, is as intended, and therefore allows for a design with overall consistency with only minimal instructions. In the configuration of the second embodiment, the organic combination of the surface processing 16 and the interpolation processing 23 using the standard data 51 allows the user to create just the right architectural model data 41 through dialogue in natural language. In other words, the user can create an architectural model 41 with standard content while checking it in the three-dimensional data 13, simply by entering the minimum necessary instructions in natural language text or voice in the chat area 7 of Example 2. This is a result that cannot be achieved simply by using a large-scale language model, and can be achieved by a new system configuration that combines elemental technologies such as the drawing processing unit 10, controller 20, position information processing 22, complementation processing 23, standard data 51, and a configuration that causes the natural language processing unit 30 to generate instruction codes 31, as shown in Figure 5.

[0063] Individual standard data 52 of Example 2 5 again, the construction process support system of the second embodiment may include a conversation log database 70. In this example, the controller 20 includes a reply process 24 and an extraction process 25.

[0064] In response to an update of the architectural model data 41 in the architectural model DB 40, the reply processing 24 causes the natural language processing unit 30 to generate a reply message 12 to the user related to the content of the update. The conversation log database 70 stores time-series data of the input message 11 and the reply message 12 as a conversation log 71. Then, the extraction process 25 causes the natural language processing unit 30 to refer to the conversation log 71 and extract individual standard data 52 for each attribute of the user in accordance with the configuration of the standard data 51. In this example, the complementation process 23 may preferentially apply individual standard data 52 for each user or each organization to which the user belongs, rather than general standard data 51. For example, if both an offset amount for each grid label 21 and a collective offset amount for all of a plurality of grid labels 21, such as the living room, are registered as offset amounts for the architectural model data 41, the reply process 24 may control the display of a reply message 12 explaining the current offset amount in the chat area 7. By being able to explain the contents of the architectural model data 41 with this reply process 24, it is possible to naturally convey to the user how the architectural model data 41, including the supplemented content, is registered in an integrated manner with the display of the three-dimensional data 13.

[0065] As shown in Fig. 6, the standard data table 50 may store standard data exclusive to Company A as individual standard data 52. The individual standard data 52 has the same set of standard data 51 as is, but has numerical values ​​and the like overwritten. For example, the floor height is 2800 [mm] in the standard data 51, but is overwritten to 3000 [mm] in the exclusive data for Company A. The individual standard data 52 may also include items not included in the standard data 51. In examples with individual standard data 52, the interpolation process 23 may refer to the individual standard data 52 as standard data 51 or similarly to interpolate the architectural model data 41.

[0066] The second embodiment, which uses this individual standard data 52, includes a conversation log database 70, and the extraction process 25 extracts the individual standard data 52. This makes it possible to extract standard data related to the design and construction process for each user, each company or department to which the user belongs, or each individual project from past conversation logs. In particular, when the individual standard data 52 includes numerical values, by preparing the data as individual standard data 52 rather than having the natural language processing unit 30 learn the data further, it is possible to control the complementation by natural language conversation without any fluctuations, and it is possible to stabilize the operation and performance of the entire system at a high level while using a large-scale language processing model. By making the individual standard data 52 text data in a natural language as shown in FIG. 6, it can be handled as a manual that is easy to understand for both the natural language processing unit 30 and the user. The user or the organization may extract the individual standard data 52, approve it at an internal meeting, etc., and store the confirmed content in the standard data table 50. If the design basis can be recorded as user input, standard data 51, or individual standard data 52, the traceability of design decisions can be significantly improved, and the reasons for such decisions can be shared throughout the construction process.

[0067] The individual standard data 52 is customized for each company based on the standard data 51 (basic manual), and directly contributes to improving the productivity of each company. The standard data 51 and the individual standard data 52 are referenced when completing dimensions and the like when there are no detailed instructions from the user. In addition to this completion, they can also be used as reference data when proposing dimensions to the user. In Example 2, individual standard data 52 can be utilized, so that design is not dependent on individuals, dimensional standards can be unified for each organization, the entire construction process can be standardized, waste can be eliminated, productivity can be increased, and stress on workers can be reduced.

[0068] A detailed example of the standard data 51 and individual data 52 shown in Figure 6 is disclosed. It has a hierarchical structure in the order of ○■●. Numerical values ​​overwritten by individual standard data 52 are indicated in double brackets (()). A company A's dedicated manual for XX is individual standard data 52 for a special project. If there is a range for the numerical value, the range may be displayed to the user to prompt input, or the natural language processing unit 30 may be made to select a value within that range that is consistent with other structures. In either case, if there is a discrepancy in visual confirmation of the three-dimensional data 13, it may be updated by the user, and if there is a refresh process or review process for the architectural model data 41, a process to check the consistency between structures may be executed.

[0069] Standard Data 51 ○Floor level ■ The floor height (z) of the first floor is 500 mm or more. ■The floor height per floor is 2800mm ○ Pillar ■ Height up to the next floor level ■Size For wooden structures, 120mm For reinforced concrete (RC) construction, 300-600mm For steel frame construction, 150-200mm ○Wall ■ Height up to the next floor level ■Thickness For wooden structures, 100-150mm In the case of reinforced concrete construction The exterior wall is 180-200mm The inner wall is 120 mm For steel structures, 100-150mm ○Window ■ Height from the floor For waist-high windows, 1000mm In the case of sweeping, it should be at the same height as the floor. For high-side windows, the ceiling height is minus 100mm and the window height is ■The height of the window itself For waist-high windows, 1100mm For sweeping, 200mm For high windows, choose 300, 450, or 600 mm. ■ The width of the window itself For waist-high windows, 1650mm For sliding windows, 1690mm For sweeping, 1500-2600mm Door ■ Height ●For entrance doors, 2000 2400mm For interior doors, 1800-2000mm ■Width ● For entrance doors, 780-900mm For parent and child doors or double doors, 1220 1690mm For interior doors, 700-800mm The toilet door is 600mm ○Curtain wall ■Glass thickness ●6 12mm ○Floor ■ Thickness (height) ●15mm ○Handrails ■ Height ●1100mm

[0070] Company A's dedicated manual (individual standard data 52) ○Floor level ■ The floor height (z) of the first floor is 500 mm or more. ■The floor height per floor is ((3000)) mm ○ Pillar ■ Height up to the next floor level ■Size For wooden structures, 120mm For reinforced concrete (RC) construction, 300 ((500)) mm For steel frame construction, ((180)) mm ○Wall ■ Height up to the next floor level ■Thickness For wooden structures, 100-150mm In the case of reinforced concrete construction The exterior wall is 180-200mm The inner wall is ((150)) mm For steel structures, ((120))mm ○Window ■ Height from the floor For waist-high windows, ((900)) mm In the case of sweeping, it should be at the same height as the floor. For high-side windows, the ceiling height is calculated as follows: ((150)) mm - window height ■The height of the window itself For waist-high windows, 100mm For sweeping, 200mm For high windows, choose 300, 450, or 600 mm. ■ The width of the window itself For waist-high windows, 1650mm For sliding windows, 1690mm For sweeping, 1500-2600mm Door ■ Height ●For entrance doors, 2000 2400mm For interior doors, 1800-2000mm ■Width ● For entrance doors, 780-900mm For parent and child doors or double doors, 1220 1690mm For interior doors, 700-800mm ○Curtain wall ■Glass thickness ●((10))mm ○Floor ■ Thickness (height) ●15mm ○Handrails ■ Height ●((1200))mm

[0071] Company A's dedicated manual (individual standard data 52) Floor Level ■ The floor height (z) of the first floor is 500 mm or more. ■The floor height per floor is ((2900)) mm ○ Pillar ■ Height up to the next floor level ■Size For wooden structures, 120mm For reinforced concrete (RC) construction, 300 ((500)) mm For steel frame construction, ((180)) mm ○Wall ■ Height up to the next floor level ■Thickness For wooden structures, 100-150mm In the case of reinforced concrete construction The exterior wall is 180-200mm The inner wall is ((150)) mm For steel structures, ((120))mm ○Window ■ Height from the floor For waist-high windows, ((800)) mm In the case of sweeping, it should be at the same height as the floor. For high-side windows, the ceiling height is calculated as follows: ((120)) mm - window height ■The height of the window itself For waist-high windows, 100mm For sweeping, 200mm For high windows, choose 300, 450, or 600 mm. ■ The width of the window itself For waist-high windows, 1650mm For sliding windows, 1690mm For sweeping, 1500-2600mm Door ■ Height ●For entrance doors, 2000 2400mm For interior doors, 1800-2000mm ■Width ● For entrance doors, 780-900mm For parent and child doors or double doors, 1220 1690mm For interior doors, 700-800mm ○Curtain wall ■Glass thickness ●((12))mm ○Floor ■ Thickness (height) ●15mm ○Handrails ■ Height ●1100mm

[0072] Construction process support method of Example 2 (complementary processing 23, individual standard data 52) Referring to Figure 7, the construction process support method of Example 2 newly includes a complementation processing step S20, a reply processing step S21, a conversation log recording processing step S22, an extraction processing step S23, and a surface processing step S24 in addition to the configuration of the construction process support method of Example 1 shown in Figure 4. In FIG. 7, the same steps and procedures as those shown in FIG. 4 of Example 1 are indicated by dotted lines. 7, in the construction process support method of Example 2, similarly to Example 1 (FIG. 4), the drawing processing unit 10 controls the display of page 6 (drawing processing step S10), receives an input message 11 input in natural language data in the chat area 7 of page 6, and controls the display of the message in the chat area 7 (input message receiving step S11). When a grid label 21 is included in the input message 11, the controller 20 generates an instruction code 31 that uses the grid label 21 as location information of the structure (location information processing step S12). Particularly in the second embodiment, the complementing process 23 causes the natural language processing unit 30 to refer to the standard data 51, and complements the architectural model data 41 that is missing in the input message 11 regarding the structure with the context corresponding to the conversation with the user using natural language data (complementing process step S20). Then, the natural language processing unit 30 refers to the input message 11 and generates an instruction code 31 that describes the architectural model data 41 by natural language processing, and particularly in the second embodiment, adds numerical values ​​and the like of items that are in the standard data 51 but not in the input message 11 received so far as the complementing process 23 (natural language processing step S13). Next, the controller 20 executes the instruction code 31 to store the architectural model data 41 in the architectural model DB 40 in a searchable manner (architectural model data storage step S14).

[0073] The chat processing 14 of the drawing processing unit 10 controls the display of an input message 11 in the chat area 7 (input message receiving step S11). In the second embodiment, the reply processing 24 of the controller 20 generates a reply message 12 to the user related to the content of the update in response to an update of the architectural model data 41 to the architectural model DB 40 (reply processing step S21). The natural language processor 30 generates a reply message 12 in natural language that describes the execution result of the instruction code 31. For example, it may be a sentence such as "XX has been placed at the position of the grid label."

[0074] In the second embodiment, the controller 20 stores the input message 11 and the reply message 12 in chronological order in the conversation log database 70 (conversation log recording processing step S22). Then, the extraction processing 25 of the controller 20 extracts individual standard data 52 from the conversation log 71 for each attribute of the user in accordance with the configuration of the standard data 51. For example, the controller 20 may read the conversation log 71 and request the natural language processing unit 30 to generate, as individual standard data 52, the contents in which values ​​of the standard data 51 have been overwritten or items that are not included in the items of the standard data 51 but have been added.

[0075] In the second embodiment, the surface processing 16 processes the surface of the three-dimensional data 13 according to the material of the structure specified or completed by the user (surface processing step S24). For example, the surface of the three-dimensional data 13 of the structure is rendered with a texture corresponding to the material name of the structure (object). Then, the three-dimensional processing 15 generates the three-dimensional data 13 of each structure at the position according to the grid label 21 of each structure (three-dimensional processing step S15).

[0076] 7, individual standard data 52 is extracted from the conversation log and updated each time an input message 11 is entered, so that it can be immediately used in the next input message 11. On the other hand, to ensure greater stability as a standard, the conversation log 71 may be read out when the design of a floor or the design of the entire architectural model is completed, and the contents of the standard data 51 items that have been overwritten, as well as the contents that are not included in the standard data 51 items but are specifically specified in the repeated input message 11, may be extracted.

[0077] Example 2 Error message 61 Next, in a second embodiment, an example in which an error message 61 is used will be disclosed. 5 again, with regard to error processing, the construction process support system of the second embodiment is provided with an error message table 60 in the controller 20. The controller 20 is also provided with a re-creation process 26 and an inexecutable process 27.

[0078] The error message table 60 stores error messages 61 in advance. The error messages 61 are natural language data messages that indicate an error resulting from the execution of the instruction code 31, such as an insufficient specification of a data item or the non-existence of a target specified by the instruction code 31. The instruction codes 31 are codes that the natural language processing unit 30 generates from the input message 11, etc., under the control of the controller 20.

[0079] When the execution result of the instruction code 31 is an error due to insufficient specification of a data item, the re-creation process 26 causes the natural language processing unit 30 to refer to the error message 61 and re-create the instruction code 31. On the other hand, if the execution result of the instruction code 31 is an error indicating that the target does not exist, the unexecutable process 27 displays the error message 61 on the page 6 .

[0080] Insufficient specification of a data item corresponds to, for example, the case where the position of a pillar is required but not specified. Non-existence of a target corresponds to the case where a pillar with the ID to be deleted is not registered in the architectural model DB 40.

[0081] The execution result of the instruction code 31 is the operation result of the architectural model DB 40. The operation result is the result obtained after the controller 20 operates the architectural model DB 40 according to the instruction code 31, and branches the error processing. The database operation result is important information for generating a response message to a user request and for stably and appropriately operating the database. In the second embodiment, the controller 20 identifies an error message 61 in natural language by referring to an error code output by the architectural model DB 40 as a normal function.

[0082] Referring to FIG. 8, the error messages 61 in the second embodiment are divided into two types: a recreate type and an inexecutable type. The recreate type occurs when a required item is not specified and is not completed even when the completion process 23 is performed by referring to the standard data 51. The error message 61 may be something like "The required field 'XXX' has not been specified." This error message 61 is not an error code output by the architectural model DB 40, but is an error message 61 that is predetermined according to the error code.

[0083] The XX in the error message 61 displays the items required for adding or updating each structure (object). For example, the placement position, height, material, etc. In other words, even if the user does not specify the details in the input message 11, the natural language processing unit 30 will complete the information in accordance with control from the controller 20, but an error will occur if the natural language processing unit 30 does not complete the information. For example, if the placement position and material are included in the instruction code 31 but the height is not included in the instruction code 31, by generating the instruction code 31 again using this re-creation type error, the natural language processing unit 30 can generate the instruction code 31 by complementing it from the standard data 51. For important items such as placement locations, where you don't know where to place them in the first place, it's a good idea to ask the user about an unexecutable error.

[0084] In the second embodiment, the error messages 61 are created in advance by classifying them into a recreate type or an inexecutable type depending on the type of structure (object) while also referring to the items in the standard data 51. By dividing them into these two types, it is possible to prevent the natural language processing unit 30 from repeatedly issuing unsolvable questions, while minimizing the items that the user must specify, thereby enabling robust and effective construction process support.

[0085] As an inexecutable type, error message 61 in #11 is "An unsupported object type '____' was specified." ____ represents an object type such as column or wall. This error occurs when an unsupported object type is specified in the instruction code 31 of the natural language processor 30. The natural language processor 30 generates a reply message 12 informing the user that the object is unsupported.

[0086] The error message 61 in #12 is "Level with ID XX not found.", where XX is the level (floor) ID. For example, the controller 20 or the architectural model DB 40 automatically assigns a unique ID, such as ID0 for the ground level and ID1 for the first floor level. When a user places a new object and specifies a floor (level) that does not yet exist, this error #12 will occur. For example, if the third floor is not set and the input message 11 from the user is "Please add a column to C3 on the third floor level," the process will be as follows: [1] The natural language processing unit 30 or the controller 20 checks the level ID of the floor (third floor) specified by the user. [2] If the controller 20 subsequently determines that the level of the specified floor does not yet exist, it transmits an error message 61 to the natural language processing unit 30 stating “Level ID3 not found” and causes the natural language processing unit 30 to generate a reply message 12. The natural language processing unit 30 generates a reply message 12 such as "The third floor level has not yet been set, so pillars cannot be placed."

[0087] Furthermore, when the natural language processing unit 30 infers a level ID from the context of a conversation with the user, it may generate an instruction code 31 that includes a non-existent level ID. In this case, the controller 20 also generates an error #12. If the controller 20 transmits an error message 61 stating "Level ID3 not found" to the natural language processing unit 30 and generates a reply message 12, even if an inference error (hallucination) occurs in the large-scale language model (natural language processing unit 30), no inconsistency in the architectural model data 41 will occur, and design can continue within the normal range using the natural language input message 11.

[0088] The error message 61 in #13 is "The specified object ID △△ for 〇〇 does not exist." 〇〇 displays the object type (column or wall). △△ displays the ID (number) of the object type specified by the user. For example, when deleting or updating an existing object, an error occurs if the specified object ID does not exist. For example, if the input message 11 from the user is "Please delete column ID 100," this error will occur if a column with ID 100 does not exist. The natural language processing unit 30 generates instruction code 31 to delete the column with ID 100, and the controller 20 executes instruction code 31. Then, the architectural model DB 40 returns an error code indicating that the object does not exist, because the column with ID 100 does not exist. The controller 20 transmits this error message 61 of #13 to the natural language processing unit 30 for this error code indicating that the object does not exist. In response, the natural language processing unit 30 generates a reply message 12 such as "A pillar with ID 100 does not exist." If the target does not exist, the controller 20 may directly control the display of an error message 61 in the chat area 7 as the reply message 12, but when the controller 20 transmits the error message 61 to the natural language processing unit 30, the natural language processing unit 30 generates a reply message 12 that also includes the error message 61 depending on the context, and may also generate, for example, a suggestion of what to do next in response to the error message 61.

[0089] If the user does not specify an object by ID, but instead specifies it by the floor and grid label 21, such as "pillar C3 on the 3rd floor," and instructs it to be moved or deleted, and the C3 pillar does not exist, the same error processing will occur as in #12.

[0090] In Example 2, errors are divided into those that may be completed normally without any additional information from the user and those that are not. If there is a possibility of successful completion, the natural language processing unit 30 is requested to recreate the instruction code 31, and if there is no possibility of successful completion, the user is notified of an error regarding the non-existence of the target. By creating an error message 61 for this purpose and storing it in the error message table 60 in advance, it is possible to realize a system that is robust to a variety of input messages 11 without relying on the capabilities of the natural language processing unit 30.

[0091] Referring to Figure 9, the construction process support method of Example 2 for processing error messages 61 newly includes a response determination processing step S30, an error message identification processing step S31, an error classification processing step S32, a re-creation processing step S33, and an absence processing step S34 following the construction model data storage step S14 of the construction process support method shown in Figure 4 of Example 1 and Figure 7 of Example 2.

[0092] 9, similarly to the first embodiment, the natural language processing unit 30 refers to the input message 11 and generates, by natural language processing, an instruction code 31 describing the architectural model data 41. At this time, in the second embodiment, as the complementation process 23, numerical values ​​and the like of items that are in the standard data 51 but not in the input message 11 received so far are added as the complementation process 23 (natural language processing step S13). Next, the controller 20 executes the instruction code 31 to store the architectural model data 41 in the architectural model DB 40 in a searchable manner (architectural model data storage step S14).

[0093] In the second embodiment for processing the error message 61, the controller 20 checks the processing result (response code or error code) from the architectural model DB 40 and determines whether the response is normal (response determination processing step S30). If the add operation, update operation, delete operation, etc. is successful, the response is determined to be normal and the process proceeds to the processes (S21, S24) described in the first and second embodiments. In the example shown in FIG. 9, if the response is not normal, the error code or the like is referenced to identify the error message 61 shown in FIG. 8 (error message identification process S31). If the execution result of the instruction code 31 is an error due to insufficient specification of a data item (error classification processing step S32), the natural language processing unit 30 is made to refer to the error message 61 and recreate the instruction code 31 (recreation processing step S33). Returning to step S13, the natural language processing unit 30 identifies the content of the insufficient specification from the error message 61 and regenerates the instruction code 31 (S13), and the controller 20 executes the recreated instruction code 31 (S14). If the recreated instruction code 31 does not complete normally (S30), the controller 20 may determine that an object does not exist error has occurred (S32).

[0094] In step S32, if the execution result of instruction code 31 is an error indicating that the target does not exist, either from the beginning or as a result of the loop (S32), an error message 61 notifying that the target does not exist is displayed on page 6 (non-existence processing step S34). The controller 20 may control the display of this target non-existence error message 61 directly in the chat area 7, or the error message 61 may be passed to the natural language processing unit 30 once, and the natural language processing unit 30 may generate a reply message 12 including the error message 61 (S21).

[0095] The re-creation process 26 shown in Fig. 5 corresponds to steps S31, S32, S33 to S13 in the example shown in Fig. 9. The inexecutable process 27 shown in Fig. 5 corresponds to steps S31, S32, S34, and S21 in the example shown in Fig. 9. [Example]

[0096] 10, in the construction process support system of the third embodiment, the controller 20 includes an input prompting process 28. Also, a standard data table 50 stores design process data 54 that is predetermined as a standard design process. The input prompting process 28 reads the design process data 54 from the standard data table 50 and causes the natural language processing unit 30 to refer to it, and causes the natural language processing unit 30 to generate a reply message 12 prompting input of the content to be designed following the state of the architectural model data 41.

[0097] The design process data 54 is a proceduralization of textbook-like design content, and in the scope of the explanation of Example 3, it is the order in which the design of, for example, the selection of construction method, floor level setting, floor plan (arrangement of structures and human movement), pillars, doors, windows, etc. is decided. In the third embodiment, for elements (types of structures) of the architectural model data 41, a standard design sequence can be defined as the design process data 54 to facilitate smooth design. In addition, by identifying the construction method, such as steel-reinforced concrete construction, reinforced concrete construction, traditional wooden frame construction, or wooden frame wall construction (two-by-four construction), it is possible to identify the standard materials for the pillars and walls. By defining the construction method in the standard data 51 and design process data 54 and having the natural language processing unit 30 refer to these data, even if the material is not specified in the input message 11, it is possible to complement it with the concrete material when generating a pillar.

[0098] In Example 3, the natural language processing unit 30 is made to refer to the design process data 54 to generate a reply message 12, which can report the execution result ("The living room has been registered") as well as generate a suggestion of the next item to input ("Where should the kitchen be placed?"). By configuring the input prompting process 28 to generate this reply message 12 by referring to the design process data 54 rather than through additional learning by the natural language processing unit 30, it is possible to prevent the reply message 12 from deviating from the architectural model data 41 to be designed, and to continue to ensure the consistency of the architectural model data 41.

[0099] The construction process support method including the input prompting process 28 of the third embodiment may include a new input prompting process step S40 (not shown) either integrally with or subsequent to the reply process step S21 shown in FIGS.

[0100] 7 again, in step S14, the natural language processing unit 30 stores the architectural model data 41 in the architectural model DB 40 using the generated instruction code 31. Then, the reply processing 24 generates a reply message 12 in accordance with the latest status of successful storage or update of the architectural model data 41 (S21). In the third embodiment, the input prompting processing 28 refers to the design process data 54 and generates a reply message 12 prompting input of the content to be designed following the current status of the architectural model data 41 (S40, not shown). For example, by specifying in the design process data 54 the order in which the living room is followed by the kitchen, and once the layout of one floor has been decided, the design of pillars and windows is prompted, the natural language processing unit 30 can generate a reply message 12 suggesting what should be input next. The format of reply message 12 should be something like, "XX has been placed. Please check. Next, where would you like to place △△?"

[0101] Next, a specific processing example in the third embodiment will be described with reference to Fig. 11 to Fig. 16. This content has some parts in common with the first and second embodiments. Case: Adding floor levels and placing compartments and columns Assuming an apartment building constructed of reinforced concrete (RC) In this example, for ease of explanation, a virtual procedure for placing a pillar after placing a living room is described as the design process data 54. In reality, it is desirable to place the pillar after placing all the partition objects such as the living room and kitchen. In this embodiment, the name of the system that is the subject of the reply message 12 is "ACIMUS."

[0102] -Adding (creating) floor levels User: "Please add a first floor level." System "1st floor floor level" has been added at "500mm". Let us know what you want to do next. For example, you can create a living room, kitchen, or other area on the first floor." Adding (creating) a parcel object User: "Next, please place the living room on C3 F5." (Figure 11, Chat box 7) System (ACIMUS) "The living room has been placed from C3 to F5. Please confirm. Next, how about placing a column in the corner of your living room? Also, please let us know if there is anything else you would like to add." (Figure 12)

[0103] ·Change (update) the parcel object User: "Please change the living room to D3 G5 position." System: "Living room has been moved to D3 G5. Please confirm. Second, do you want to put a column in the corner of your living room?”

[0104] Adding (creating) a pillar object User: "Then, please place a pillar in the corner of the living room." (Figure 13) System: "A pillar has been placed in the corner of the living room. Please check. Next, how about placing a 200mm wide wall between the columns? Or, let us know if there's anything else you'd like to add." (Figure 14)

[0105] Importing pillar objects User: "How many pillars do you have now?" System: There are currently four pillars, located as follows: 1. D3 2. D5 3. G3 4. G5 Is there anything else you would like to add or change?

[0106] Delete the pillar object User: "Then please delete the G5 column." (Figure 15) System: "I've deleted the G5 column. Do you have any other instructions for additions or changes?" (Figure 16) The three-dimensional data 13 indicated by the reference numeral 13a in Fig. 15 is a pillar G5 with a grid label 21, which has been deleted as shown in Fig. 16. The pillar may be identified by the grid label 21, or the ID of the pillar shown in Fig. 3 may be added by the three-dimensional processing 15 or the like and displayed on the canvas 8, and the pillar may be identified and interacted with by the ID number.

[0107] In Example 3, the natural language processing unit 30 knows the location of the living room from the previous exchange, so it can place the pillars at D3, D5, G5, and G3 where they should be placed by simply saying "the corner of the living room." He also suggests what to do next in reply message 12. Furthermore, even if specific floor level heights or column sizes and heights are not specified, the dimensions of the structure are supplemented with general dimensions (values ​​in standard data 51). In this way, in this embodiment, it is possible to add, change, read, and delete structures (objects) that are consistent and just the right amount through natural language conversation. In this regard, simply using a large-scale language model would not allow three-dimensional data 13 (3D objects) to be displayed on the canvas 8, and it would be difficult for the user and the system (large-scale language model) to execute the coordinates without any misunderstanding. Furthermore, if contexts and assumptions arise between the user and the system and the coordinate values ​​are expressed in a special natural language, it would be impossible to construct a consistent architectural model data 41. In this embodiment, first, by using grid labels 21, it is possible to stably handle location information while utilizing the original functions of the large-scale language model (natural language processing unit 30) (chat in natural language and code generation). Furthermore, by preparing standard data 51 and error messages 61 and incorporating them into the system components to operate the entire system, users can create architectural model data 41 that is perfectly consistent based on natural language dialogue. This improvement in human creativity allows the system to listen to the client's detailed needs and requests, allowing work to be focused on the essential tasks of construction. At the same time, users who are just beginning to learn design and architecture, users unfamiliar with IT, and experienced designers who have the standard data 51 memorized can all enjoy creating designs using natural ideas in natural language, improving the burdensome process of inputting architectural model data 41 in the construction industry. By using the systems of each embodiment, architectural model data 41 that meets the client's needs can be created in a variety of natural languages, and as the number of construction projects that can use architectural model data 41 increases, the productivity improvements envisioned by BIM and other technologies can be realistically implemented at construction sites.

[0108] As described above, in the third embodiment, the input prompting process 28 causes the natural language processing unit 30 to refer to the design process data 54 and generates a reply message 12 prompting the user to input the content to be designed following the state of the architectural model data 41. This makes it possible to prompt the user to input the content in a standard and desirable order for the design process, and also makes it possible to ask the user to make a decision so that there are no missing items in the architectural model data 41. The user can naturally know the contents to be input following the state of the latest architectural model data 41 according to the inputs made up until now through dialogue by messages, and can consider essentially essential matters in a desired order in the standard order of design without stress. This reduces unnecessary work during the design phase, allowing a diverse range of people to carry out natural, high-quality designs using standard processes without stress, providing users with a productive and fulfilling work experience. Furthermore, by answering questions in a natural language orderly manner, it is possible to create architectural model data 41 that is consistent and complete, without requiring advanced study of IT or CAD. This allows a diverse range of people to quickly create architectural model data 41 as they wish, and as a result, it is possible to widely utilize architectural model data 41 in the construction industry. [Example]

[0109] 1 (Example 1), 5 (Example 2), and 10 (Example 3) again, basic operations and extended examples relating to common processing etc. will be disclosed as Example 4. The description of Example 4 is not suitable for direct reference when interpreting the meaning of the terms in the claims, and the disclosures of Examples 1 to 3 should be referred to first.

[0110] The terminal 2 is a device through which a user accesses the system of each embodiment, and corresponds to a PC, tablet, smartphone, etc. Chatting and operations are performed through a user interface that is displayed and controlled by the page 6. Page 6 is the user interface of the front server that functions as the drawing processing unit 10, and is an input / output device for designing architectural models through dialogue in natural language. The input natural language is displayed in the chat area 7, and the generated three-dimensional data (3D object) is drawn on the canvas 8.

[0111] When inputting into the chat area 7, messages (language) can be input either by keyboard or by voice. The input language (sentence, input message 11) and the corresponding generated sentence (reply message 12) received from the controller 20 are displayed. The controller 20 can be made stand-alone as a controller server.

[0112] The canvas 8 draws three-dimensional data 13 (3D object) received from the controller 22. By operating this canvas 8, the following functions can be implemented as functions of the drawing processing unit 10 without using natural language processing 30, but this is not essential. Importing BIM files (e.g. IFC files) - Display and output the generated architectural model data 41 in 2D drawings such as floor plans, sections, and elevations - Display / hide of various structures (objects) that make up the architectural model data 41 and certain segments Add, update, and delete structures directly through user data entry, rather than through input messages 11.

[0113] Controller 20 (controller server, control unit) It is the core of the system in each embodiment, and by coordinating with the drawing processing unit 10 (front server), natural language processing unit 30 (large-scale natural language processing model), architectural model DB 40, and conversation log DB 70 (conversation log database 70), it comprehensively realizes the functions of the entire system. The controller 20 can implement the following functions: ■ Message processing Receives an input message 11 input by the user from the drawing processing unit 10 Send the input message 11 to the natural language processor 30 A reply message 12 (response message) from the natural language processing unit 30 to the user is sent to the drawing processing unit 10 ·Operation of database (architectural model DB40) Receives instruction code 31 from the natural language processor 30 to operate the database (read, add, update, delete) Executes operations on the database based on the instruction code 31 from the natural language processor 30 Sending the database operation results to the natural language processing unit 30 If the database operation results in failure, refer to the error message table 60 and send an error message 61 to the natural language processing unit 30, requesting regeneration of the database operation instruction code 31 or generation of an inoperable reply message 12. ■Data conversion Converting the architectural model data 41 into data in a format that can be displayed on the canvas 8 of the drawing processing unit 10 ■ Management and review process Control of complex processing flows depending on the situation and refresh processing using architectural model data41 -Error handling for improperly placed structures as part of the refresh process. Detects collisions (interference) between structures according to instructions from review and refresh processes. For example, whether a girder is positioned in such a way that it will collide with the position of a column. -Detection of inappropriate structures (objects) that do not conform to the standard data 51 or design process data 54 according to instructions for review and refresh processes. For example, a girder is not connected to a column.

[0114] Natural language processing unit 30 (large-scale natural language processing server) The natural language processing unit 30, under the control of the controller 20, analyzes the input message 11 from the user, determines the appropriate action (generating the instruction code 31 for database operation or the reply message 12), and also makes suggestions to the user. This is a natural language processing model. The natural language processing unit 30 is a server separate from the controller 20, and can be accessed using an API provided by the large-scale natural language processing server. Depending on the scale of the large-scale natural language processing, the natural language processing unit 30 may be executed locally on the computer on which the controller 20 is executed.

[0115] The natural language processing unit 30 realizes the following functions in accordance with the control of the controller 20. Receives user input message 11 from controller 20 -Analyze the contents of user input message 11 Determining the need to operate the DB40 architectural model Generate instruction code 31 to operate architectural model DB40 ·Generating a reply message to the user 12 - Proposal of necessary additional information when the input message from the user 11 is insufficient (complementary process 23 referring to standard data 51) - Suggesting the next action to take after completing the current task (input prompt process 28) Send reply message 12 to controller 20 Sends instruction code 31 to controller 20 to operate architectural model DB40 Receives the operation results of the architectural model DB40 from the controller 20 If the operation of the architectural model DB40 fails, you will receive the reason for the failure (error message 61). If the operation of the architectural model DB 40 fails and is a re-creation type, the instruction code 31 for operating the database is regenerated based on the error message 61 (by the re-creation process 26). If the operation of the architectural model DB 40 fails and is inexecutable, a reply message 12 is generated based on the error message 61, indicating that the target does not exist (by the inexecutable process 27).

[0116] Architectural model DB40 (data storage unit) A device (database) that reads, adds, updates, and deletes data in response to instruction codes 31 received from the controller 20, and stores (saves) data on structures (objects of the architectural model). The architectural model DB 40 has the following functions. Add (save) data Loading data ·Data Update Data deletion - Data type consistency verification [Industrial Applicability]

[0117] Figure 17 shows a business design sheet that shows the industrial applicability of an invention conceived by the applicant. The future image is a vision, and while it has technical backing, it is not a disclosure of technology but a business direction. The business design sheet shown in Figure 17 is the concept of the applicant or the company to be established. First, by initially releasing the systems and services related to this invention, we will provide architectural designers with technology that can generate consistent 3D architectural model data41 in natural language. This is the value we will provide to customers and the social environment, and to achieve this, we will identify what can be done now as a waypoint toward the future by understanding architecture and researching the applicability of AI technology, and then implement it into the system. The applicant's dual expertise in architecture and AI is a valuable resource for this. For example, the applicant focuses on the stress of workers in the construction industry as a cause of the labor shortage. The current situation in the construction industry, where people are overwhelmed with unnecessary reports, unnecessary considerations, and complicated adjustments, and are unable to concentrate on the work they should be doing, is causing stress for everyone involved. Reporting and coordination can be resolved through digitalization, but the input of architectural model data 41, which is the starting point, has not been widely adopted, and the amount of architectural model data 41 that can be used throughout the construction process has not increased, preventing progress in improving the efficiency of reporting and coordination work at each construction site. In contrast, the construction process support system, method, and program according to the present invention uses large-scale language models as components, and through cooperation between the drawing processing unit 10 and the controller 20, can provide architectural designers with an environment for creating architectural model data 41 that is perfectly consistent in natural language. Architectural model data 41 that specifies whether something is a pillar and what material it is made of, rather than just line drawings, is prepared from the design stage, and as the number of available construction sites increases, it will become the foundation for improving the efficiency of the entire construction process.

[0118] In the future vision shown in Figure 17, the applicant will provide an environment where people connected with those working in the construction industry, such as building material manufacturers, furniture and home appliance manufacturers, condominium management associations, and people involved in repair work, can each concentrate on the essential work of construction. To provide this value, the applicant will promote the use of the construction process support system, method, and program according to the present invention and this embodiment. The construction process support system and its architectural model data 41 of this invention will link the entire construction process chain, automate reporting, and automatically create coordination plans, eliminating inefficient and stressful work for people working in the construction industry and those connected to it. The resources required for this are the natural language design support provided by the present invention and this embodiment, the accumulation of architectural model data 41, and networks with construction companies and building material manufacturers. Technology alone cannot solve all industrial issues, and relational capital such as networks between business operators in the process is also necessary, but the architectural model data 41 is attractive when building networks. At present, it is difficult to create that appeal (architectural model data 41), but the architectural process support system of this invention and this embodiment allows a variety of personnel to easily create architectural model data 41 that meets the needs and requests of clients with the minimum necessary instructions in natural language, with an accuracy that can be confirmed in three-dimensional data 13, so it is expected that architectural model data 41 will become widely accepted. The construction process support system of the present invention and this embodiment can directly and indirectly contribute to meeting construction needs similar to selecting custom-made clothes in the construction industry, which has realized a three-day work week, for example. Note that the content shown in Figure 17 is a vision regarding industrial applicability from a management perspective, and even if there are businesses in the future that do not implement a three-day work week, this does not narrow the scope of the rights of this invention. When interpreting the terms in the claims of the present invention, reference should be made to the corresponding descriptions in Examples 1 to 3, if necessary, and the meaning and overall scope of the terms should not be interpreted by directly referring to Example 4 or the description of industrial applicability. [Explanation of symbols]

[0119] 2. Terminal 4 Network Page 6 7. Chat Area 8. Canvas 10. Drawing processing section 11 Input Message 12 Reply Message 13 Three-dimensional data 14 Chat Processing 15 Three-dimensional processing 16 Surface Treatment 20 Controller 21 Grid Labels 22 Location Information Processing 23 Complementary Processing 24 Reply Processing 25 Extraction process 26 Re-creation process 27 Unexecutable Processing 28 Input prompt processing 30 Natural Language Processing Unit 31 Instruction Code 40 Architectural Model Database (Architectural Model DB) 41 Architectural model data 50 Standard Data Tables 51 Standard Data 52 Individual Standard Data 54 Design Process Data 60 Error Message Table 61 Error Message 70 Conversation Log Database (Conversation Log DB) 71 Conversation Log

Claims

1. a drawing processing unit that is connected to a user's terminal and controls the display of a page that serves as a user interface; a controller for controlling an input message input in natural language data from the terminal to the drawing processing unit; a natural language processing unit that generates instruction codes describing architectural model data by referring to the input message; an architectural model database that stores the architectural model data corresponding to the instruction code in a searchable manner; the controller includes a location information processing unit that, when a grid label specifying a coordinate on the page by a character code is included in the input message of an architectural structure, causes the natural language processing unit to generate the instruction code that sets the grid label as location information of the structure; a chat process in which the drawing processing unit controls display of the input message in a chat area of ​​the page; a three-dimensionalization process for displaying and controlling three-dimensional data of the positions of each structure of the architectural model data read from the architectural model database on a canvas within the page according to the grid labels of the respective structures; A construction process support system characterized by:

2. the controller includes a complementing process that causes the natural language processing unit to refer to predetermined standard data of architecture, and complements the architectural model data that is insufficient in the input message regarding the structure with a context corresponding to a conversation with the user using the natural language data; the rendering processing unit includes a surface processing unit that processes a surface of the three-dimensional data according to a material of the structure designated by the user or completed by the user; 2. The construction process support system according to claim 1.

3. the controller includes a reply process that causes the natural language processing unit to generate a reply message to the user related to the content of the update in response to an update of the architectural model data to the architectural model database; a conversation log database for storing time-series data of the input messages and the reply messages as a conversation log is provided in addition to the architectural model database; The controller includes an extraction process that causes the natural language processing unit to refer to the conversation log and extract individual standard data for each attribute of the user in accordance with the configuration of the standard data.

3. The construction process support system according to claim 2.

4. the controller is provided with an error message table in which an error message in natural language data is recorded in advance, the error message indicating an insufficient specification of a data item or non-existence of a target specified by the instruction code in relation to an error in the execution result of the instruction code generated by the natural language processing unit; a re-creation process in which, when the execution result of the instruction code is an error due to the insufficient specification of the data item, the controller causes the natural language processing unit to refer to the error message and re-create the instruction code; and a non-executable process for displaying an error message on the page when the execution result of the instruction code is an error indicating the non-existence of the target.

2. The construction process support system according to claim 1.

5. the controller includes an input prompting process that causes the natural language processing unit to refer to design process data predetermined as a standard design process and generates a reply message that prompts input of content to be designed following the state of the architectural model data, 5. A construction process support system according to claim 1, 2, 3 or 4.

6. 1. A method executed on one or more computers, comprising: a drawing processing step of connecting to a user's terminal and controlling the display of a page serving as a user interface on the terminal; an input message receiving step of receiving an input message input in natural language data into a chat area of ​​the page via a network and displaying the input message in the chat area; a position information processing step of generating an instruction code for specifying the grid label as position information of the structure when the input message includes a grid label specifying coordinates on the canvas within the page using a character code; a natural language processing step of generating, by natural language processing, an instruction code describing architectural model data that identifies position information of the structure by the grid label, with reference to an input message including the grid label; an architectural model data storage step of retrievably storing the architectural model data corresponding to the instruction code; a three-dimensional processing step for displaying and controlling three-dimensional data of the positions of each structure of the architectural model data read from the architectural model database on the canvas within the page according to the grid labels of the respective structures, A construction process support method comprising:

7. A program to be executed by one or more computers, a drawing processing procedure for connecting to a user's terminal and controlling the display of a page serving as a user interface on the terminal; an input message receiving step of receiving an input message input in natural language data into a chat area of ​​the page via a network and displaying the input message in the chat area; a location information processing procedure for generating, when the input message includes a grid label specifying coordinates on a canvas within the page using a character code, an instruction code for specifying the grid label as location information of the structure; a natural language processing procedure for generating, by natural language processing, an instruction code describing architectural model data that identifies position information of the structure by the grid label, by referring to an input message including the grid label; an architectural model data storage step of retrievably storing the architectural model data corresponding to the instruction code; a three-dimensional processing procedure for displaying and controlling three-dimensional data of the positions of each structure of the architectural model data read from the architectural model database on the canvas within the page according to the grid labels of the respective structures; causing the computer to execute A construction process support program characterized by:

Images