A planning and design method, apparatus and storage medium

Abstract

A planning and design method, apparatus, and storage medium are disclosed. This scheme guides users to input the necessary planning and design conditions through a pre-set question chain, thereby ensuring the completeness of the information required for planning and design, and thus ensuring the completeness and accuracy of the generated planning and design scheme. Furthermore, prompting users to input specific information through the pre-set question chain avoids situations where users input missing data but do not know what to supplement, ultimately resulting in the inability to obtain a professional planning and design scheme, thus greatly improving the user experience. In summary, this scheme, based on prompting engineering technology and a large language model for user interaction, can generate complete and accurate planning and design schemes and provide users with a good user experience.

Description

Technical Field

[0001] This disclosure relates to, but is not limited to, the field of computers, and particularly to a planning and design method, apparatus, and storage medium. Background Technology

[0002] Recently, large-scale language models have become popular, such as ERNIE (Enhanced Representation through Knowledge Integration). These large language models can analyze and understand input text and process various natural language tasks through internet knowledge, achieving significant progress in the field of natural language processing. However, current large language models are generally general-purpose models and cannot meet users' needs for planning and design solutions. Summary of the Invention

[0003] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0004] This disclosure provides a planning and design method, apparatus, and storage medium that can generate complete and accurate planning and design schemes, and provide users with a good user experience during the recommendation process.

[0005] One embodiment of this disclosure provides a planning and design method, including:

[0006] Select the corresponding preset question chain based on the task type of the planning and design selected by the user;

[0007] The large language model outputs the questions in the preset question chain to the user one by one, and extracts keywords based on the user's input data for each question; the extracted keywords are then used to search the preset planning and design database to obtain planning and design condition information.

[0008] Based on the aforementioned planning and design conditions information, a pre-generated planning and design knowledge graph is searched to confirm the design materials;

[0009] Based on the confirmed design materials, generate the planning and design scheme corresponding to the task type.

[0010] An embodiment of this disclosure also provides a planning and design system, including a reasoning module, a planning and design function module, and a conventional function module, wherein:

[0011] The reasoning module is configured to determine the type of function to be executed, which includes planning and design functions and regular dialogue functions.

[0012] The planning and design function module is configured to perform planning and design functions, and to perform them in accordance with the planning and design method as described in any embodiment of this disclosure;

[0013] The regular function module is configured to perform regular dialogue functions.

[0014] An embodiment of this disclosure also provides a planning and design apparatus, including: a memory and a processor;

[0015] The memory is used to store the program used for planning and design;

[0016] The processor is configured to read the program for planning and design and execute the planning and design method as described in any embodiment of this disclosure.

[0017] An embodiment of this disclosure also provides a non-transient computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, is capable of implementing the planning and design method as described in any embodiment of this disclosure.

[0018] Compared with related technologies, the planning and design method, apparatus, and storage medium provided in this disclosure utilize Prompt Engineering technology. This involves setting preset question chains corresponding to different planning and design task types, and outputting the questions in the preset question chains to the user one by one using a large language model. Then, keywords are extracted from the user's input data for each question, and a preset planning and design database is searched based on the extracted keywords to obtain user-approved planning and design conditions. Finally, design materials are selected and a planning and design scheme is generated based on the user-approved planning and design conditions. In summary, the planning and design method, apparatus, and storage medium of this disclosure can interact with the user based on a large model and preset question chains, confirm the user's planning and design needs, and thus ensure the generation of accurate planning and design schemes that meet the user's expectations.

[0019] Other features and advantages of the embodiments of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the embodiments of this application. The objects and other advantages of the embodiments of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0020] The accompanying drawings are used to provide an understanding of the technical solutions of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the technical solutions of this disclosure and do not constitute a limitation on the technical solutions of this disclosure.

[0021] Figure 1This is a flowchart of a planning and design method according to an embodiment of the present disclosure;

[0022] Figure 2 This is a flowchart of a plant planning and design method according to an embodiment of the present disclosure;

[0023] Figure 3 This is a flowchart of the pre-defined problem chain related part of a plant planning and design method according to an embodiment of the present disclosure;

[0024] Figure 4 This is a flowchart illustrating the generation of planning and design schemes in a plant planning and design method according to an embodiment of the present disclosure;

[0025] Figure 5 This is a schematic diagram of a planning and design system according to an embodiment of the present disclosure;

[0026] Figure 6 This is a schematic diagram of a planning and design device according to an embodiment of the present disclosure. Detailed Implementation

[0027] This disclosure describes several embodiments, but these descriptions are exemplary and not limiting, and it will be apparent to those skilled in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with, or may replace, any feature or element of any other embodiment.

[0028] This disclosure includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features, and elements disclosed in this disclosure may also be combined with any conventional features or elements to form a unique inventive scheme as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive schemes to form another unique inventive scheme as defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this disclosure may be implemented individually or in any suitable combination. Therefore, the embodiments are not limited except by the limitations imposed by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.

[0029] One embodiment of this disclosure provides a planning and design method, such as... Figure 1 As shown, the following steps may be included:

[0030] Step S101: The large language model selects the corresponding preset question chain based on the task type of the planning and design selected by the user;

[0031] In step S102, the large language model outputs the questions in the preset question chain to the user one by one. After outputting each question, the following operations are performed: extract the first keyword based on the user's input data for the current question; search the preset planning and design database based on the extracted first keyword to obtain the planning and design condition information corresponding to the current question.

[0032] Step S103: The large language model searches the pre-generated planning and design knowledge graph based on the planning and design condition information corresponding to each question in the preset question chain to confirm the design material selection;

[0033] Step S104: The large language model generates a planning and design scheme corresponding to the task type based on the confirmed design materials.

[0034] This embodiment's planning and design method utilizes Prompt Engineering technology. It sets up preset question chains corresponding to different planning and design task types, and outputs each question in the preset question chain to the user through a large language model. Then, it extracts keywords from the user's input data for each question, searches a preset planning and design database based on the extracted keywords, and obtains user-approved planning and design conditions. Finally, it determines the design materials and generates a planning and design scheme based on these conditions. This method, on the one hand, guides users to input the necessary planning and design conditions through the preset question chain, ensuring the completeness of the required information and thus the completeness and accuracy of the generated planning and design scheme. On the other hand, prompting users to input specific information through the preset question chain avoids situations where users input missing data without knowing what to supplement, ultimately preventing them from obtaining a professional planning and design scheme, thus greatly improving the user experience. In summary, this embodiment's planning and design method, based on Prompt Engineering technology and a large language model interacting with the user, can generate complete and accurate planning and design schemes and provides a good user experience.

[0035] It should be noted that the preset question chain in this embodiment can be set based on the suggestions of experts in the target field, and can be adjusted and optimized according to actual planning and design needs. The preset question chain based on expert suggestions can ensure that the planning and design condition information required to generate a professional planning and design scheme for the target field is obtained, thereby enabling the generation of a professional planning and design scheme for the target field.

[0036] In an exemplary embodiment of this disclosure, the step of selecting a preset problem chain corresponding to the task type selected by the user for the planning and design further includes:

[0037] When a user opens the dialog interface of the planning and design system, the large language model is activated. The large language model outputs a preset first question, which includes a variety of preset planning and design task types.

[0038] The large language model determines the task type based on the user's input data for the first question.

[0039] In one example of this embodiment, after confirming the task type based on the user's input data for the first question, the method further includes:

[0040] Output a pre-set second question, which includes various pre-set planning and design features;

[0041] The unique design concept was confirmed based on the user's input data for the second question.

[0042] The aforementioned distinctive design concepts will be used as planning and design condition information.

[0043] The distinctive design concepts may include low-carbon design concepts (or carbon neutral design concepts), cost control design concepts, Internet of Things design concepts, smart garden design concepts, etc., but are not limited to these and can be adjusted according to actual needs.

[0044] This embodiment, by pre-setting some distinctive design concepts, can help users add some cutting-edge design elements to the general planning and design scheme, so that the generated planning and design scheme can keep up with the times.

[0045] In an exemplary embodiment of this disclosure, obtaining the planning and design conditions information corresponding to the current problem includes:

[0046] 41. Output the obtained planning and design conditions information to the user for confirmation, and extract the second keyword based on the user's input data on the planning and design conditions information;

[0047] 42. When it is confirmed that the user agrees with the planning and design conditions information based on the extracted second keyword, the planning and design conditions information shall be used as the planning and design conditions information corresponding to the current question;

[0048] If, based on the extracted second keyword, it is confirmed that the user does not agree with the planning and design conditions information, the current question is re-outputted, and the planning and design conditions information is searched again in the planning and design database based on the third keyword extracted from the user's new input data, or based on a combination of the first and second keywords; then the process returns to step 41. It should be noted that the search for planning and design conditions information in the planning and design database is not limited to the above two methods and can be adjusted according to needs. For example, it can also be searched again based on the second keyword, or by replacing the search strategy based on the first keyword, etc.

[0049] In this embodiment, the system further confirms with the user whether each planning and design condition meets the user's needs. Based on the user's input data, the system analyzes the user's specific needs and feedback, and adjusts the planning and design condition information in real time to ensure that each confirmed planning and design condition meets the user's specific needs. This ensures that personalized planning and design schemes are generated for different users, better meet their needs, gain their approval, and greatly improve user satisfaction and user experience.

[0050] To make this embodiment clearer, let's take the application of the planning and design method of this embodiment to plant planning and design as an example. For instance, one question in the preset question chain could be asking about the area of ​​the site for which plant design is needed. Based on the keyword extracted from the received user input data, "10 hectares" can be used as a planning and design condition. Then, based on the site area "10 hectares", the number of plants corresponding to that site area, "30", can be obtained from the preset planning and design database as a planning and design condition. Another example is that another question in the preset question chain could be asking about the location of the site for which plant design is needed. Based on the keyword extracted from the received user input data, "Beijing" can be used as a planning and design condition. Then, based on the site location "Beijing", the evergreen-to-deciduous ratio of plants corresponding to the "North China region" where the site is located, "4:6", can be obtained from the preset planning and design database as another planning and design condition. Yet another example is... Another question could be to inquire about the type of land use for which plant design is required. Based on the keywords extracted from the received user input data, such as "urban land" and "water body," the system can then retrieve the corresponding plant type ratio ("tree:shrub:herb:aquatic plant ratio of 3:2:1:1") from a pre-set planning and design database, based on the land use type "urban land / water body," as another planning and design condition.

[0051] In an exemplary embodiment of this disclosure, the step of searching a pre-generated planning and design knowledge graph based on the planning and design condition information to confirm the design material selection includes:

[0052] Generate index instructions based on the planning and design conditions information;

[0053] The pre-generated planning and design knowledge graph is searched using the indexing instructions to confirm the design materials.

[0054] In this embodiment, after analyzing the user-input data through a large language model to obtain planning and design conditions, index instructions are generated based on these conditions. Finally, material selection can be confirmed based on these index instructions. This process, compared to the traditional method where users need to generate planning and design schemes by inputting command lines, is more in line with human operating habits, more convenient, and provides a better user experience.

[0055] In an exemplary embodiment of this disclosure, the preset planning and design database further includes a material list template; the planning and design scheme includes a material list;

[0056] The step of generating a planning and design scheme corresponding to the task type based on the confirmed design materials includes:

[0057] The first bill of materials template is confirmed from the bill of materials template according to the index instruction, and the first constraint condition corresponding to the first bill of materials template is confirmed.

[0058] Based on the first constraint, the confirmed design material selection is output to obtain the material list, and the form of the material list is the same as that of the first material list template.

[0059] To make the format of the material list in this embodiment clearer, let's take the application of the planning and design method of this embodiment to plant planning and design as an example. For instance, the constraints corresponding to the matched first material list template can be: the header titles are in the following order: Chinese name, scientific name, plant type, ornamental characteristics, plant height / cm, diameter at breast height, crown width, price, and carbon sequestration. Each type of plant is listed as a row in the table, and the relevant information for different types of the plant is written into columns that match the header content. For example, the Chinese name of the plant is written in the first column, the scientific name in the second column, and so on. The specific generated first material list for plants can be found in Table 1 in the detailed process of the plant planning and design method below.

[0060] In an exemplary embodiment of this disclosure, the preset planning and design database further includes a design specification template; the planning and design scheme further includes a design specification.

[0061] The step of generating a planning and design scheme corresponding to the task type based on the confirmed design material selection also includes:

[0062] The first design specification template is confirmed from the design specification template based on the bill of materials and the planning and design conditions information;

[0063] The design specification is generated based on the first design specification template, the bill of materials, the planning and design conditions information, and the first keywords.

[0064] It should be noted that the bill of materials template and design specification template in this embodiment were created under the guidance of professionals in the target field (such as academic professors and industry experts), which ensures that the final bill of materials and design specifications are professional, that is, the final planning and design scheme is professional.

[0065] Compared to traditional large language models that only generate text answers, the planning and design method in this embodiment generates both a bill of materials and a design specification. This allows users to clearly understand the materials needed, and the design specification explains the design scheme, showing users how to use the selected materials and the advantages of the generated planning and design scheme. This greatly enhances the user experience.

[0066] To make the material list format of this embodiment clearer, taking the application of the planning and design method of this embodiment to plant planning and design as an example, the design description may include: design descriptions of color and seasonal phases, layering and space, and uses and functions, etc., but is not limited to these. The large language model of this embodiment will analyze user needs based on the user's selected task type, distinctive design concepts, and other input data to generate design descriptions, ensuring the practicality of the design descriptions. For example, the generated design descriptions may be as follows:

[0067] Color and Seasons:

[0068] To create a romantic atmosphere that matches the theme, I suggest choosing plants with pure, soft colors. In spring, choose some early-blooming plants, such as irises (Iris tectorum), whose pale purple flowers complement the moon goddess theme. In summer, introduce some white or pale yellow flowers, such as white pears (Pyrus bretschneideri) and hibiscus (Philadelphus pekinensis), whose fresh and lovely flowers add a touch of coolness to the garden. In autumn, choose some red-leaved plants, such as weeping willows (Styphnolobium japonicum 'Pendula'), whose beautiful red leaves complement the moonlit garden.

[0069] Hierarchy and Space:

[0070] Based on the size and theme of the courtyard, I suggest establishing several zones within it, each with plants of varying heights and shapes to create depth and visual interest. At the entrance and main areas of the courtyard, choose tall deciduous trees such as elm (Ulmus pumila) and purple acacia (Amorpha fruticosa). Their tall, dense foliage will form a green barrier and provide some privacy. At the boundaries and corners, select smaller deciduous trees or shrubs such as hawthorn (Crataegus pinnatifida) and rhododendron (Rhododendron simsii). Their medium height and abundant flowers will add depth and color to the courtyard.

[0071] Uses and functions:

[0072] Depending on the intended use of the private villa garden, I recommend creating an open lawn area in the courtyard for activities and relaxation. Around the edges of the lawn or the main area, choose some ground cover plants such as petunias (Petunia hybrida) and irises (Iris tectorum), which can cover the ground, increase the green area of ​​the courtyard, and add beautiful flowers. Additionally, an aquatic plant area can be created on one side or in a corner of the courtyard, choosing aquatic plants such as water onions (Scirpus Validus) to enhance the ecological atmosphere of the courtyard. Finally, for some supporting structures or walls, choose some vines, such as Virginia creeper (Parthenocissus tricuspidata), which can climb and cover the supporting structures, providing some screening and decorative effects for the courtyard.

[0073] In summary, based on the information you provided, I have designed a plant arrangement scheme suitable for your private villa garden in Tianjin, themed "Moon Goddess." Regarding color and seasonality, plants with soft colors that match the theme have been selected. In terms of layering and space, plants of varying heights and shapes have been used to enhance the sense of depth in the garden. Regarding purpose and function, lawns, aquatic plants, and vines have been rationally arranged according to the size and intended use of the garden. I hope this plant arrangement scheme meets your needs and brings a beautiful landscape effect to your garden.

[0074] In an exemplary embodiment of this disclosure, the knowledge graph is obtained according to the following steps:

[0075] Collect data in the target domain, including basic knowledge data and planning scheme data;

[0076] The data in the target domain is preprocessed to generate target data; wherein, the preprocessing of the data in the target domain includes: data cleaning, data deduplication, data error correction, and data annotation.

[0077] A knowledge graph is constructed based on the target data.

[0078] This embodiment addresses the problem of inaccurate planning and design recommendations for target domains due to the limitations of generalization in existing technologies. By collecting target domain data containing expert opinions and professional knowledge, and preprocessing the target domain data through cleaning and deduplication, the data is then organized, classified, and summarized to form a professional knowledge graph. This allows for the retrieval of design material selection information for the target domain from the professional knowledge graph based on user input data, thereby improving the accuracy of retrieval and ultimately enhancing the accuracy of the generated target domain planning and design solutions, providing users with highly professional target domain planning and design solutions.

[0079] To make this embodiment clearer, taking the application of the planning and design method of this embodiment to plant planning and design as an example, the knowledge graph can include plant name, plant type, Latin name, growth habits, suitable habitat, ornamental characteristics, plant height, crown width, price, etc. The knowledge graph in this embodiment is a comprehensive, accurate, and professional knowledge graph of landscape plants formed by integrating professional knowledge in the field of landscape plant design and existing databases. This knowledge graph enables the generated plant planning and design schemes to be more professional and accurate.

[0080] In an exemplary embodiment of this disclosure, the large language model is obtained according to the following steps:

[0081] Collect and calculate basic knowledge data and planning scheme data, and store them in a database;

[0082] The collected data undergoes preprocessing such as data cleaning, deduplication, error correction, and annotation.

[0083] A large language model is trained based on preprocessed data to obtain a pre-trained large language model.

[0084] The large language model is fine-tuned based on the fine-tuning dataset, including updating the parameters of the large language model using a gradient descent optimization algorithm.

[0085] Using the fine-tuning dataset as training data, fine-tuning training is performed on the pre-trained large language model, including updating the parameters of the large language model using the gradient descent optimization algorithm;

[0086] Test whether the large language model meets the preset requirements; for example, the fine-tuned large language model can be tested on an independent test set to evaluate its performance.

[0087] It should be noted that the fine-tuning dataset described in this embodiment is a professional fine-tuning dataset generated by collecting expert solutions and standard solutions written by professionals.

[0088] This embodiment fine-tunes a large language model using a high-quality fine-tuning dataset, and employs a gradient descent optimization algorithm to update the model's parameters during the fine-tuning process, ultimately obtaining a high-performance large language model. This large language model can provide highly specialized planning and design solutions for target domains, offering design suggestions to professional designers in those domains.

[0089] In an exemplary embodiment of this disclosure, the method may further include:

[0090] Collect user usage records and satisfaction levels;

[0091] The large language model is optimized based on user usage records and satisfaction levels.

[0092] This embodiment collects user satisfaction and usage data to understand the true performance of the large language model, and then optimizes the large language model based on the collected user data.

[0093] In an exemplary embodiment of this disclosure, user input data can be entered in the form of text, and the generated planning and design information can also be output to the user in the form of text. This embodiment does not limit the specific input and output methods.

[0094] When text is input and output in the form of a dialogue in this embodiment, the implementation can combine data storage to store the dialogue in a database. Through continuous dialogue, more information about the user is collected, and the collected information is used to optimize the large language model, so that the answers generated by the large language model better meet the user's expectations.

[0095] In summary, the planning and design method disclosed herein utilizes prompting engineering techniques combined with deep learning technology and large language model fine-tuning to help users clarify planning and design conditions, thereby improving design efficiency and generating accurate, professional, and user-relevant planning and design solutions for the target domain, greatly enhancing the user experience.

[0096] The following example, using the above planning and design methods applied to plant planning and design, illustrates how to use these methods.

[0097] One embodiment of this disclosure provides a plant planning and design method, such as... Figure 2 As shown, the method may include the following steps:

[0098] Step S210: When the user opens the dialog interface of the planning and design system, a preset first question is output, which includes a variety of preset plant planning and design task types to inquire about the type of plant planning and design task the user needs. Among them, the preset plant planning and design task types may include: plant material list design scheme, design description scheme, and complete planning and design scheme, but are not limited to these. The preset plant planning and design task types can be expanded according to actual needs.

[0099] The dialogue process for this step can be as follows:

[0100] LA chat: Which type of plant planning and design scheme do you need? Plant material list design scheme / design specification scheme / complete planning and design scheme?

[0101] User: I need a complete planning and design scheme.

[0102] Step S211: Confirm the task type selected by the user based on the user's input data in response to the first question.

[0103] For example, keywords can be extracted from user input data and matched with task types. If a match is found with one and only one task type, then the user's selection of that task type is confirmed.

[0104] It should be noted that if a unique plant planning and design task type cannot be identified based on the user input data (e.g., the user input data cannot match any task type, or more than one task type is successfully matched), the process will return to step S210, which involves continuing to ask the user for the desired plant planning and design task type. When multiple task types are successfully matched, the first question can only include the successfully matched task types for the user to further select from.

[0105] For example, if the user input data is detected to be an interrogative sentence (such as containing a question mark, or a preset word expressing a question like "what"), then the explanatory text for various task types is output and the process returns to step S210.

[0106] Step S220: Output a pre-set second question, which includes various pre-set characteristic design concepts for plant planning and design, to ask the user whether they need to incorporate characteristic design concepts when conducting plant planning and design. These pre-set characteristic design concepts may include: low-carbon design concepts (or carbon-neutral design concepts), cost control design concepts, Internet of Things design concepts, and smart garden design concepts, but are not limited to these. The content of the pre-set characteristic design concepts can be expanded according to actual needs.

[0107] The dialogue process for this step can be as follows:

[0108] LA chat: Would you like to add any of the following design concepts? (e.g., low-carbon design concept / cost control design concept / IoT design concept / smart landscape design concept)

[0109] User: I need low-carbon design concepts.

[0110] Step S221: Based on the user's input data in response to the second question, confirm whether the user needs to incorporate a unique design concept. If the user needs to incorporate a unique design concept, confirm the unique design concept selected by the user. It should be noted that if, based on the user's input data, it is determined that the user does not need to incorporate a unique design concept, proceed to step S230.

[0111] Step S230: Determine the corresponding preset question chain based on the task type selected by the user; output the questions in the preset question chain to the user one by one, and extract keywords based on the user's input data for each question; search the preset planning and design database based on the extracted keywords to obtain planning and design condition information. The planning and design condition information may include site area, site location, site land use type, site theme, climate conditions, number of plant species, ratio of evergreen to deciduous plants, proportion of plant types, plant types, and site color. This disclosure does not limit the content of the planning and design condition information and it can be configured according to actual needs.

[0112] Optionally, it can be further confirmed whether each planning and design condition information meets the user's needs, including: organizing the planning and design condition information into natural language form through a large language model and outputting it to the user; extracting keywords based on the user's input data for each planning and design condition information; if the user agrees with the planning and design condition information based on the extracted keywords, then the planning and design condition information is taken as the final planning and design condition information; if the user does not agree with the planning and design condition information based on the extracted keywords, then the planning and design condition information is regenerated based on the extracted keywords and organized into natural language form through a large language model and output to the user for confirmation.

[0113] This step outputs the questions in the preset question chain to the user one by one, and confirms the planning and design conditions for each question with the user. By reaching a consensus on the planning and design conditions for each question with the user, a planning and design scheme that meets the user's expectations can be generated based on the planning and design information approved by the user, which can improve the user experience.

[0114] For example, when the user selects the task type as "complete planning and design scheme", such as Figure 3 As shown, step S230 may include the following steps S231-S234:

[0115] Step S231: Output the first question in the preset question chain, confirm the site area for the plant planning design based on the user's input data for the first question, and obtain the recommended number of plant species to be planted for the site area from the preset constraint template. For example, the preset constraint template includes "3 hectares or less: 10 species", "3-10 hectares: 20 species", and "10 hectares or more: 30 species".

[0116] Optionally, it can be further confirmed whether the recommended number of plant species based on the site area meets the user's needs. For example, if the site area cannot be confirmed (e.g., the user input data is detected as an interrogative sentence, such as containing a question mark, or preset words expressing questions such as "what"), or if the user is not satisfied when confirming with the user (e.g., containing the keyword "not satisfied"), then step S231 is re-executed.

[0117] The dialogue process for this step can be as follows:

[0118] LA chat: What is the area of ​​the site for which you need to design the plant arrangements?

[0119] User: 10 hectares.

[0120] LA chat: We recommend using 30 kinds of plants. Does that meet your needs?

[0121] User: Not suitable, 40 kinds of plants are required.

[0122] LA chat: We suggest using 40 plants. Does this meet your needs? (If not, please provide your suggestions.)

[0123] User: Matches.

[0124] Step S232: Output the second question in the preset question chain, confirm the site location for the plant planning and design based on the user's input data for the second question, and obtain the recommended ratio of evergreen to deciduous plants for the site location from the preset constraint template.

[0125] Optionally, it can be further confirmed whether the recommended evergreen to deciduous ratio of plants based on the site location meets the user's needs. For example, if the site location confirmed by the user's input data for the second question is "Beijing", then "Beijing" can be matched to "North China" in the preset location categories "East China\South China\North China\Central China\Southwest China\Northwest China\Northeast China". Then, the preset evergreen to deciduous ratio of plants "4:6" corresponding to "North China" can be output from the database. Finally, the site location "North China" and "evergreen to deciduous ratio of plants 4:6" are organized into natural language by the large model and output to the user for confirmation.

[0126] The dialogue process for this step can be as follows:

[0127] LA chat: Where is the design site you need plant design located?

[0128] User: Located in Beijing.

[0129] LA chat: We have selected commonly used plants in North China for you, and suggest an evergreen to deciduous ratio of 4:6. Does this meet your needs? (If not, please provide your suggestions.)

[0130] User: Matches.

[0131] Step S233: Output the third question in the preset question chain, confirm the site use type of the plant planning design based on the user's input data for the third question, and obtain the recommended proportion of plant types for the site use type from the preset constraint template.

[0132] Optionally, it can be further confirmed whether the recommended plant type ratio based on the site's land use type meets the user's needs. For example, if the keywords extracted from the user's input data for the third question are "urban land" and "water body", the keywords "urban land" and "water body" can be matched with the preset site land use classifications "urban land\water body\mountain\farmland\hills\grassland\desert\wasteland\grassland\forest" to determine the site land use type as "urban land\water body". Then, referring to the database output, the plant type ratio corresponding to "urban land\water body" can be "tree:shrub:herb:aquatic plant ratio of 3:2:1:1". The site land use type "urban land\water body" and "tree:shrub:herb:aquatic plant ratio of 3:2:1:1" can be organized into natural language form by the large model and output to the user for confirmation.

[0133] The dialogue process for this step can be as follows:

[0134] LA chat: What type of site do you need for your plant design?

[0135] User: Primarily urban land, with some water bodies.

[0136] LA chat: We have selected commonly used plants for urban and aquatic habitats for you, suggesting a ratio of trees:shrubs:herbs:aquatic plants of 3:2:1:1. Does this meet your needs? (If not, please provide your suggestions.)

[0137] User: Matches.

[0138] Step S234: Output the fourth question in the preset question chain, confirm the site theme of the plant planning and design based on the user's input data for the fourth question, and obtain the recommended plant types and site colors for the site theme from the preset constraint template.

[0139] Optionally, it can be further confirmed whether the recommended plant types and site colors based on the site theme meet the user's needs.

[0140] The dialogue process for this step can be as follows:

[0141] LA chat: What theme do you need for your plant design?

[0142] User: Subject: Moon.

[0143] LA chat: We have selected plants related to moonlight culture for you, such as osmanthus, evening primrose, epiphyllum, and moonflower. We suggest the main colors of these plants are green, white, purple, light purple, blue, and light blue. Does this meet your needs? (If not, please provide your suggestions.)

[0144] User: Matches.

[0145] Step S240: Generate an index instruction based on the planning and design conditions information, and determine the selected plants and their related information from the pre-generated knowledge graph using the index instruction. For example: if the number of plants confirmed in the above steps is "40", the site location is "North China", the ratio is "evergreen trees: deciduous trees: evergreen shrubs: deciduous shrubs: vines: herbaceous flowers and ground cover, format reference 5:5:5:4:3:2", the site theme keywords are "moonlight, green, white, purple, light purple, blue, light blue", and the output additional column is "carbon sink value (i.e., low-carbon design concept) yes, price (i.e., cost control design concept) yes, IoT design concept no", then the format of the finally generated index instruction can be "40; North China; 5:5:5:4:3:2; moonlight, green, white, purple, light purple, blue, light blue; carbon sink value yes, price yes, IoT design no".

[0146] Step S250: Generate a plant planning and design scheme. For example... Figure 4 As shown, step S250 may include the following steps S251-S252:

[0147] Step S251: Match a suitable plant material list template from the preset plant material list templates according to the indexing instructions, and obtain the constraints of the matched plant material list template; output the retrieved plant information under the constraints, and output a plant material list with the same format as the matched plant material list template. For example, the constraints corresponding to the matched plant material list template can be: the header titles are in the following order: Chinese name, scientific name, plant type, ornamental characteristics, plant height / cm, diameter at breast height, crown width, price, and carbon sequestration. Each type of plant is listed as a row in the table, and the relevant information of different types of the plant is written into columns that are consistent with the header content. For example, the Chinese name of the plant is written into the first column, the scientific name into the second column, and so on. The specific output plant material list can be shown in Table 1.

[0148] It should be noted that the plant material list template may be created under the guidance of professionals (such as academic professors or industry experts).

[0149] Table 1

[0150]

[0151] Step S252: Match a suitable design specification template from the preset design specification templates based on the plant material list and planning design conditions, and generate a design specification based on the matched template and user-input keywords. The design specification template may be based on factors such as site theme (50%), site size (20%), habitat type (20%), and regional culture (10%); color and seasonality: considering the flower and leaf colors of the plants to create an aesthetically pleasing landscape effect, etc. For example, the generated design specification may be as follows:

[0152] Next is my description of the plant design for this low-carbon industrial park:

[0153] Seasonal design:

[0154] The plant design for a moonlight-themed low-carbon industrial park should fully consider the changing plant landscape throughout the four seasons to ensure that the entire park presents a beautiful moonlight atmosphere in different seasons. Specific plant selections are as follows:

[0155] Spring: Plants such as Kerria japonica, Sycamore, and Lilac create a fragrant spring atmosphere. Ash, Ginkgo, and Sophora japonica trees display their vibrant green leaves in spring, adding vitality to the park. Summer: Evergreen trees such as Willow and Robinia pseudoacacia provide ample shade, offering cool resting places for employees. Evening primrose and Pearl Bush, among other summer-flowering plants, adorn the park, creating a fresh and pleasant environment. Autumn: Plants such as Ginkgo and Pinus armandii create a scene of falling leaves, like gold, enhancing the park's autumnal charm. Evergreen trees such as Pinus bungeana and Cedrus deodara maintain their greenery, creating a picturesque scene throughout the year. Winter: Plants such as Ash and Pinus armandii create an elegant, snow-covered landscape. Winter-flowering shrubs such as Prunus triloba and Pearl Bush bring a touch of winter color to the park.

[0156] Landscape Function and Layout Design:

[0157] Shade Trees and Street Trees: Ash trees, ginkgo trees, and cypress trees are planted around main roads and office areas, forming dense shade that not only provides a cool working environment but also casts soft moonlight shadows at night. Ornamental Trees and Flower Beds: Chinese pine, purple-leaved barberry, and kerria are cleverly arranged in scenic spots and rest areas of the park, forming beautiful ornamental trees and flower beds. These areas become even more enchanting under the moonlight, attracting employees and visitors. Hedges and Bank Protection Trees: Cypress trees and willows are used as hedges and bank protection trees in the park, not only serving a greening purpose but also providing isolation and protection at night, increasing the park's safety. Water Features and Reflections: Small ponds or artificial lakes are incorporated into the park design, creating beautiful reflections under the moonlight. Water features not only enhance the landscape's depth but also bring a greater sense of tranquility to the park.

[0158] Low-carbon design:

[0159] First, from a carbon sequestration performance perspective, prioritizing native and drought-tolerant plants is crucial. These plants are adapted to local climate and soil conditions, thus requiring less water and fertilizer. This not only reduces energy consumption for irrigation and fertilization but also helps increase the storage of organic carbon in the soil. I favor planting large trees such as sycamores and plane trees, which can absorb and store significant amounts of carbon throughout their life cycle. Planting different heights and types of plants, such as trees, shrubs, and ground cover, can increase the park's carbon storage capacity. Second, from a low-carbon perspective, practices such as organic farming can be considered. Avoiding chemical fertilizers and pesticides in favor of organic fertilizers and biopesticides can reduce carbon emissions and increase the storage capacity of organic carbon in the soil. Secondly, from a low-carbon perspective, the following practices can be considered: organic farming practices can avoid the use of chemical fertilizers and pesticides, instead using organic fertilizers and biological pesticides, which can reduce carbon emissions and increase the storage of organic carbon in the soil. Secondly, renewable materials, such as recycled wood and recycled plastics, can be used in park and facility construction, which can reduce the energy consumption and carbon emissions required for the production of new materials. Furthermore, discarded industrial materials, such as railway tracks and old machinery, can be reused as design elements for parks, which not only reduces the need for new materials but also adds a historical industrial atmosphere to the parks. Finally, we can collect and reuse rainwater, for example, by designing rainwater harvesting systems to collect rainwater for irrigation and other uses, which can reduce dependence on groundwater and municipal water supplies, thereby reducing energy consumption and carbon emissions. Regarding the transportation system, I incorporate green transportation concepts, providing bicycle parking facilities and walking paths within the park to encourage visitors to use public transportation, bicycles, or walking, which can reduce the use of private cars, thereby reducing carbon emissions.

[0160] Summarize:

[0161] In summary, when selecting plants, I considered their connection to the industrial era, their environmental impact, and their seasonal landscape effects. I also fully considered the moonlight theme while integrating the plants' landscape functions and layout. Furthermore, I comprehensively considered both carbon sequestration performance and low-carbon effects to ensure that the park, while providing recreational and educational functions, also makes a positive contribution to the environment. I hope these suggestions can help you design a park that is both characteristic of the industrial era and environmentally friendly and low-carbon.

[0162] Do these perspectives meet your needs? (If not, please provide your suggestions.)

[0163] Step S260: Output the plant planning and design scheme to the user, and modify and improve the plant planning and design scheme according to the user's response.

[0164] The plant planning and design method in this embodiment, on the one hand, analyzes the user's design conditions based on the user's input data, fully considering factors such as the site theme, area, location, land use type, and user preferences. This makes the generated plant planning and design scheme more aligned with user needs, providing a personalized solution. On the other hand, the final generated plant planning and design information includes a plant material list and plant design descriptions, allowing users to clearly see the material requirements, seasonal and color design descriptions, spatial and layout design descriptions, and plant function and role descriptions for each plant. This helps users understand the rationale and advantages of the recommended design scheme, providing a basis for their decision and significantly enhancing the user experience. Furthermore, the preset question chain, plant material list template, and design description template in this embodiment are designed based on the advice of experts in the field, ensuring that highly professional landscape plant planning and design schemes are provided to users.

[0165] In summary, the plant planning and design method of this embodiment features integrated professional knowledge, personalized solutions, real-time interactive functions, and a high degree of professionalism. It is a plant planning and design method particularly suitable for the field of landscape architecture design, and solves the following problems existing in the current conventional method of generating planning and design schemes using large language models:

[0166] Lack of domain expertise: While ERNIE and ChatGPT can generate natural language text, they lack in-depth expertise in specific domains. In the field of landscape architecture and plant planning and design, these technologies cannot provide highly specialized solutions because they cannot understand and handle domain-specific requirements and specifications. The output text lacks practical value in terms of format and content, and cannot produce usable plant material lists, design specifications, etc.

[0167] Generalization limitations: While existing techniques perform well on natural language processing tasks, their generalization capabilities are limited. This means that when dealing with problems in specific domains, they may be affected by data scarcity or insufficient domain expertise. This is because these models primarily rely on large-scale general text data for training, and the lack of domain expertise leads to poor accuracy even with plant-based data.

[0168] Lack of interactivity: Existing technologies are typically used to generate text responses, but they lack the ability to engage in complex interactions with users. In the field of landscape architecture and plant planning and design, users often require in-depth question-and-answer sessions and discussions with the system, and existing technologies cannot provide advanced interactive features.

[0169] It should be noted that the planning and design methods disclosed herein, which "utilize Prompt Engineering technology to ensure the completeness of information required for planning and design through a pre-set chain of questions," "employ algorithms such as deep reinforcement learning in artificial intelligence technology to achieve more efficient and accurate data processing through adaptive model training," "use natural language processing and deep learning algorithms to achieve better results in data integration and analysis," and "interaction logic with users," can be applied not only to the field of plant planning and design but also to other fields, playing a greater role. For example, they can be used for intelligent recommendations of paving, recommendations and design specifications in interior design and architectural design, etc.

[0170] This disclosure also provides a planning and design system in one embodiment, see [link to relevant documentation] Figure 5 It includes a reasoning module, a planning and design function module, and a regular function module, among which:

[0171] The reasoning module is configured to determine the type of function to be executed, which includes planning and design functions and regular dialogue functions.

[0172] The planning and design function module is configured to perform planning and design functions, and to perform them in accordance with the planning and design method as described in any embodiment of this disclosure;

[0173] The regular function module is configured to perform regular dialogue functions.

[0174] In an exemplary embodiment of this disclosure, the type of function to be executed can be determined by a large language model. For example, the large language model can be trained by first learning the planning and design function, and then the keyword in the user input data can be used to identify whether the planning and setting function needs to be executed.

[0175] In an exemplary embodiment of this disclosure, after the program is started, the planning and design system can not only provide user input but also configure parameters. This allows the user to input text and parameters, such as problem descriptions and set conditions, after the program is started. The system can then parse and analyze the user-input information to determine the user's needs.

[0176] In an exemplary embodiment of this disclosure, the planning and design system may use Prompt Engineering to add a series of standard answers and retrieved plant data to prompt words to guide the model in generating answers that are superior to traditional large language models.

[0177] In an exemplary embodiment of this disclosure, after the planning and design system is developed, testing can be used to verify whether the system meets the requirements. For example, enumeration testing can be performed first, followed by integration testing and system testing, to ensure the stability and reliability of the planning and design system.

[0178] This embodiment of the planning and design system adds a reasoning module for judgment, assisting users in clarifying functional requirements and determining the next function to execute. The task logic has been redesigned specifically for plant planning and design problems, and deep learning technology and large language models are used to generate professional plant planning and design schemes when processing plant planning and design functions. Furthermore, this embodiment's planning and design system also has real-time interactive capabilities, tracking the dialogue context and dynamically adjusting based on user feedback to maintain contextual coherence, effectively responding to user interactions, and providing further explanations or suggestions when necessary. This improves the accuracy of the output answers and significantly enhances the user experience. This embodiment's planning and design system is an information retrieval, question-and-answer, and content generation platform that provides a one-stop intelligent consulting service for landscape plant planning and design, greatly facilitating the learning of landscape architecture planning and design.

[0179] It should be noted that the planning and design system in this embodiment is based on a cloud computing platform and includes a front-end and a back-end. The front-end and back-end can communicate using a RESTful interface.

[0180] The front-end is used for display and interaction with the user, allowing the user to input questions and displaying the answers. The front-end can be in the form of a webpage, but it's not limited to just webpages; it can be any form that can interact with the user, such as an app. When the front-end is in webpage form, it can utilize web development technologies, built using HTML5, CSS3, and JavaScript. The webpage can be deployed to web browsers or mobile applications for easy user access.

[0181] The backend includes management, data storage, and computing resource modules. The data storage module stores large-scale plant databases, user information, and interaction history, managing and backing up data in a highly secure and reliable manner. Furthermore, the backend servers can employ a distributed architecture to ensure system stability and scalability. The server design also incorporates load balancing, auto-scaling, and fault tolerance mechanisms to address the impact of high traffic and system failures, further enhancing system stability.

[0182] The planning and design system in this embodiment adopts a microservice architecture and is deployed using Docker containers to achieve efficient utilization of computing resources. Furthermore, each service can be independently upgraded and extended. For example, new components can be developed to replace existing components according to the corresponding interface specifications, and data sources and data types can be configured according to user needs.

[0183] For example, the planning and design system in this embodiment can be a plant planning and design system, in which case...

[0184] The planning and design function module in this embodiment can be a plant planning and design function module. The plant planning and design system of this example has the following beneficial technical effects:

[0185] Improved performance: By employing natural language processing and deep learning algorithms to analyze and process input data, more efficient and accurate data processing and analysis can be achieved, enabling question answering functions and thus improving performance. This addresses current shortcomings such as difficulty in searching for professional knowledge about landscape architecture plants and limitations in design creativity.

[0186] Reduced time costs: The technical solution in this embodiment adopts a text generation part based on deep learning algorithm, which realizes automated processing in the generation of landscape plant planning and design. It can significantly reduce the cost of searching for materials and writing design specifications, optimize the multi-software collaborative processing process of designers in the design process, and reduce time costs.

[0187] Increased stability: The data integration and analysis section of this embodiment uses natural language processing and deep learning algorithms to establish a landscape architecture professional knowledge base, which can perform in-depth analysis and processing of input data, thereby improving the stability and reliability of the system.

[0188] Easy to operate or use: The technical solution of this embodiment adopts natural language processing, which is more in line with human thinking habits than the traditional command line operation method, thereby improving the convenience of operation and making it easier to use;

[0189] Other useful performance enhancements: The deep learning algorithms and natural language processing techniques employed in this embodiment not only achieve better results in data integration and analysis, but can also be applied to fields such as intelligent voice interaction, natural language generation, machine translation, text classification, and image recognition, enabling more efficient and accurate data processing and analysis. Therefore, this disclosure has broad application prospects and many other useful performance enhancements.

[0190] This disclosure also provides a planning and design apparatus according to one embodiment, see [link to relevant documentation]. Figure 6 This includes: memory and processor;

[0191] The memory is used to store the program used for planning and design;

[0192] The processor is configured to read the program for planning and design and execute the planning and design method as described in any embodiment of this disclosure.

[0193] The processor in the embodiments of this disclosure can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a microprocessor, etc., or other conventional processors. The processor can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), discrete logic or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or other equivalent integrated or discrete logic circuits, or a combination of the above devices. That is, the processor in the embodiments described above can be any processing device or combination of devices that implements the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. If the embodiments of this disclosure are implemented in part in software, then instructions for software can be stored in a suitable non-volatile computer-readable storage medium, and one or more processors can be used to execute the instructions in hardware to implement the methods of the embodiments of this disclosure. The term "processor" as used herein can refer to the above-described structure or any other structure suitable for implementing the techniques described herein.

[0194] An embodiment of this disclosure also provides a non-transient computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, is capable of implementing the planning and design method as described in any embodiment of this disclosure.

[0195] In one or more exemplary embodiments described above, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or code on or transmitted via a computer-readable medium and executed by a hardware-based processing unit. The computer-readable medium may comprise a computer-readable storage medium corresponding to a tangible medium, such as a data storage medium, or a communication medium comprising any medium facilitating the transfer of a computer program from one place to another, such as according to a communication protocol. In this manner, the computer-readable medium may generally correspond to a non-transitory tangible computer-readable storage medium or a communication medium, such as a signal or carrier wave. The data storage medium may be any available medium accessible by one or more computers or one or more processors to retrieve instructions, code, and / or data structures for implementing the techniques described in this disclosure. Computer program products may comprise computer-readable media.

[0196] For example, and not as a limitation, such computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage devices, magnetic disk storage devices or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and is accessible by a computer. Furthermore, any connection may also be referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. However, it should be understood that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but rather refer to non-transient tangible storage media. As used herein, disks and optical discs include compact optical discs (CDs), laser discs, optical discs, digital versatile optical discs (DVDs), floppy disks, or Blu-ray discs, where disks typically reproduce data magnetically, while optical discs use lasers to reproduce data optically. The above combinations should also be included within the scope of computer-readable media.

Claims

1. A planning and design method, characterized in that, include: When a user opens the dialog interface of the planning and design system, the large language model is activated. The large language model outputs a preset first question, which includes a variety of preset planning and design task types. The large language model determines the task type based on the user's input data for the first question. Output a pre-set second question, which includes various pre-set planning and design features; The unique design concept was confirmed based on the user's input data for the second question. The aforementioned distinctive design concepts will be used as planning and design condition information. The large language model selects a corresponding preset question chain based on the task type of the planning and design selected by the user. The questions in the preset question chain are output to the user one by one. After each question is output, the following operation is performed: the first keyword is extracted based on the user's input data for the current question. The process involves searching a preset planning and design database based on the extracted first keyword to obtain planning and design condition information corresponding to the current problem, including:

41. Outputting the obtained planning and design condition information to the user for confirmation, and extracting a second keyword based on the user's input data regarding the planning and design condition information; 42. If the user agrees with the planning and design condition information based on the extracted second keyword, using the planning and design condition information as the planning and design condition information corresponding to the current problem; if the user does not agree with the planning and design condition information based on the extracted second keyword, re-outputting the current problem, and searching for planning and design condition information again in the planning and design database based on the third keyword extracted from the user's new input data, or searching for planning and design condition information again in the planning and design database based on a combination of the first keyword and the second keyword; returning to step 41. The large language model searches a pre-generated planning and design knowledge graph based on the planning and design condition information corresponding to each question in the preset question chain to confirm the design material selection, including: generating an index instruction based on the planning and design condition information; and searching the pre-generated planning and design knowledge graph through the index instruction to confirm the design material selection. The large language model generates a planning and design scheme corresponding to the task type based on the confirmed design materials. The preset planning and design database also includes a material list template, and the planning and design scheme includes a material list. Generating a planning and design scheme corresponding to the task type based on the confirmed design materials includes: confirming a first material list template from the material list templates according to the index instruction, and confirming a first constraint condition corresponding to the first material list template; outputting the confirmed design materials according to the first constraint condition to obtain the material list, wherein the form of the material list is the same as that of the first material list template. The preset planning and design database also includes a design specification template; the planning and design scheme also includes a design specification; the step of generating a planning and design scheme corresponding to the task type based on the confirmed design materials further includes: confirming a first design specification template from the design specification template based on the material list and the planning and design condition information; and generating a design specification based on the first design specification template, the material list, the planning and design condition information, and the first keyword.

2. The method according to claim 1, characterized in that, The large language model is obtained according to the following steps: Collect and calculate basic knowledge data and planning scheme data, and store them in a database; The collected data undergoes data cleaning, deduplication, error correction, and annotation preprocessing. A large language model is trained based on preprocessed data to obtain a pre-trained large language model. The large language model is fine-tuned based on the fine-tuning dataset, including updating the parameters of the large language model using a gradient descent optimization algorithm. Using the fine-tuning dataset as training data, fine-tuning training is performed on the pre-trained large language model, including updating the parameters of the large language model using the gradient descent optimization algorithm; Test whether the large language model meets the preset requirements.

3. A planning and design device, comprising: Memory and processor; characterized in that: The memory is used to store the program used for planning and design; The processor is configured to read the program for planning and design and execute the planning and design method as described in any one of claims 1 to 2.

4. A non-transient computer-readable storage medium storing a computer program, wherein, When the computer program is executed by a processor, it can implement the planning and design method as described in any one of claims 1 to 2.

Images