An engineering survey and design archive management method and system based on NLP

Abstract

The present invention proposes an engineering survey and design archive management method and system based on NLP, which relates to the technical field of engineering archive management. The present invention addresses the technical problem that the number and content of engineering survey and design archives vary greatly with the scale of the project, and the use of a unified large language model may lead to resource waste or poor recognition effect. By obtaining engineering survey and design data to be archived, calculating the scale index of the project based on the engineering survey and design data to be archived, and then selecting a model for engineering survey and design data management based on the scale index, the present invention realizes the flexible selection of NLP models according to the project scale, rationally allocates computing resources, and improves archive management efficiency and recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering archive management, and in particular to an engineering survey and design archive management method and system based on NLP. Background Art

[0002] In recent years, electronic archive management has been widely adopted in the surveying and design industry, but it faces challenges in efficiently managing diverse archives. Natural language processing (NLP), with its ability to understand and process human language, offers the potential for intelligent archive management. Through automated classification and intelligent compilation and research, NLP improves archive management efficiency and reduces costs.

[0003] However, the number and content of engineering survey and design archives often vary significantly depending on the project scale. For small projects, the number of survey and design archives is relatively small and the information is relatively simple; for large projects, the survey and design archives may contain millions of documents, covering complex geological, environmental, and design information.

[0004] In this context, if a unified large language model is used for recognition and processing, two situations may occur: one is the waste of computing resources, because for small projects, the processing power of the large language model far exceeds the requirements, resulting in idle computing resources; the other is insufficient volume. For large projects, the large language model may not be able to fully capture and process all the important information in the archives, resulting in the accuracy and completeness of the recognition results being affected.

[0005] Therefore, how to select a suitable NLP model according to the scale of engineering survey and design archives to achieve rational allocation and efficient utilization of resources has become an important technical issue facing the current engineering survey and design archive management. Summary of the Invention

[0006] The present invention provides an engineering survey and design archive management method and system based on NLP, so as to realize flexible selection of NLP models according to the project scale, rationally allocate computing resources, and improve archive management efficiency and recognition accuracy.

[0007] An engineering survey and design archive management method based on NLP specifically includes the following steps:

[0008] Step S1, data collection: obtaining engineering survey and design data to be archived;

[0009] Step S2, scale calculation: obtaining document parameters, picture drawing parameters, and video parameters based on the engineering survey and design data to be archived, and calculating the scale index of the project based on the parameters;

[0010] Step S3, model selection: selecting a model for engineering survey and design data management according to the scale index, specifically, when the scale index is less than or equal to a preset threshold, selecting a first management model; when the scale index is greater than the preset threshold, selecting a second management model;

[0011] Step S4, file identification: inputting the engineering survey and design data into the selected management model, identifying the engineering survey and design data through the management model, and obtaining archive information;

[0012] Step S5, data storage: save the archived information into a database.

[0013] Furthermore, the method for obtaining the scale indicator is:

[0014] Step S21, dividing the engineering survey and design data to be archived into documents, pictures, drawings and videos;

[0015] Step S22, obtaining the total word count W and document conversion coefficient α1 of the document; obtaining the file size S, picture-drawing conversion coefficient α2, and picture-drawing resolution adjustment factor β of the picture-drawing; obtaining the file size V, video conversion coefficient α3, and video resolution adjustment factor γ of the video;

[0016] Step S23: Calculate the scale index based on the parameters obtained in step S22.

[0017] Furthermore, the calculation formula of the scale indicator TRS is:

[0018] TRS=W / α1+S×β / α2+V×γ / α3.

[0019] Furthermore, the image resolution adjustment factor β is:

[0020] ;

[0021] Among them, ω TP is the weight of the image, N TP is the number of pictures, S TPi is the size of the i-th picture, β TPi is the resolution of the i-th picture, ω TZ is the weight of the drawing, N TZ is the number of drawings, S TZj is the size of the jth drawing, β TZj is the resolution of the jth drawing.

[0022] Furthermore, the video resolution adjustment factor γ is set to:

[0023] ;

[0024] Among them, ω Vk is the weight of the kth video, V k is the size of the kth video, γ Vk is the resolution of the kth video, N V is the number of videos.

[0025] Furthermore, the first management model is an LSTM model, and the second management model is a Transformer model;

[0026] Furthermore, the specific parameters of the first management model are set as:

[0027] The number of hidden layers is 2, the number of neurons in each layer is 256, the activation function is sigmoid function, the optimizer is Adam, the learning rate is 0.001, the loss function is the cross entropy loss function, and the batch size is 32.

[0028] Furthermore, the specific parameters of the second management model are set as:

[0029] The number of hidden layers is 3, the number of neurons in each layer is 512, the optimizer is Adam, the learning rate is 0.0001, the loss function is the cross entropy loss function, the activation function is GELU function, and the batch size is 64.

[0030] An NLP-based engineering survey and design archive management system, using the NLP-based engineering survey and design archive management method as described in any of the above items, specifically includes the following modules:

[0031] Data collection module: obtain engineering survey and design data to be archived;

[0032] Scale calculation module: connected to the data collection module, used to obtain document parameters, picture drawing parameters and video parameters based on the engineering survey and design data to be archived, and calculate the scale index of the project based on the parameters;

[0033] A model selection module, connected to the scale calculation module, is used to select a model for engineering survey and design data management based on the scale index. Specifically, when the scale index is less than or equal to a preset threshold, a first management model is selected; when the scale index is greater than the preset threshold, a second management model is selected;

[0034] File recognition module: connected to the model selection module, used to input the engineering survey and design data into the selected management model and recognize the engineering survey and design data into computer language;

[0035] Data saving module: connected to the file identification module, used to save the identification results output by the management model into the database.

[0036] An electronic device comprising: a processor and a memory, wherein the processor is configured to execute the steps of the NLP-based engineering survey and design archive management method as described above by calling a program or instruction stored in the memory.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] First, this invention intelligently identifies the size of engineering survey and design archives and automatically selects the most appropriate NLP model for processing. This achieves precise matching of computing resources, avoids wasting resources when large models are processing small projects, and ensures that the accuracy and completeness of recognition results are not affected by insufficient capacity of small models when processing large projects. This invention not only improves the efficiency of archive management, but also significantly reduces operating costs.

[0039] Secondly, the scale index calculation method proposed in this invention comprehensively considers various data types such as documents, pictures, drawings, and videos in engineering survey and design archives. By setting conversion coefficients and resolution adjustment factors, different types of data are uniformly converted into comparable identification scale values. This method provides managers with an intuitive and accurate evaluation tool, enabling them to quickly understand the overall scale and complexity of engineering survey and design archives, thereby making more reasonable archive management decisions. In addition, the calculation of scale indicators also provides an important basis for the subsequent selection of appropriate NLP models, ensuring the scientific nature and efficiency of the archive management process.

[0040] Third, the present invention has carefully designed and optimized the setting of model parameters. Different NLP models are selected for engineering survey and design data of different scales, and reasonable parameters are set for them. The scale and complexity of engineering survey and design data are fully considered. Through reasonable selection and configuration, the rational allocation of resources is achieved, the processing efficiency is improved, the model adaptability is enhanced, and the operating costs are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 Schematic diagram of the process of the engineering survey and design archive management method based on NLP in an embodiment of the present invention;

[0043] Figure 2This is a structural diagram of the engineering survey and design archive management system based on NLP in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0045] The specific embodiments of the present invention are described below with reference to the accompanying drawings (tables).

[0046] Natural language processing (NLP) is a key branch of artificial intelligence, aiming to enable computers to understand and process human language. With the rapid development of deep learning technology, neural network models have made significant progress in NLP, becoming the primary tool for processing natural language tasks. Commonly used NLP models include recurrent neural networks (RNNs), long short-term memory networks (LSTMs), convolutional neural networks (CNNs), transformers, and pre-trained models such as BERT.

[0047] Based on the above model, the present invention studies the method of engineering survey and design archive management. By obtaining the engineering survey and design data to be archived, calculating the scale index of the project based on the engineering survey and design data to be archived, and then selecting a model for engineering survey and design data management based on the scale index, the present invention realizes the flexible selection of NLP models according to the project scale, rationally allocates computing resources, and improves archive management efficiency and recognition accuracy.

[0048] Example 1

[0049] like Figure 1 As shown, the present invention proposes an engineering survey and design archive management method based on NLP, which specifically includes steps S1-S5:

[0050] Step S1, data collection: obtaining engineering survey and design data to be archived;

[0051] The engineering survey and design data to be archived is not of a single type, but rather includes many different types of data. Specifically, these data can be divided into three categories: documents, pictures and drawings, and videos.

[0052] Document materials usually contain a large amount of text information, such as engineering reports, design drawings, contract documents, etc.

[0053] Picture and drawing materials are mainly presented in the form of images, such as photos of the construction site, electronic drawings, etc., which intuitively show the actual situation and design results of the project and are important auxiliary tools for understanding the status of the project.

[0054] Video materials, through dynamic images, capture key aspects of the engineering survey and design process, or on-site conditions. Compared to photographic drawings, video materials provide richer temporal and spatial information, contributing to a more comprehensive understanding of the project.

[0055] Step S2, scale calculation: obtaining document parameters, picture drawing parameters, and video parameters based on the engineering survey and design data to be archived, and calculating the scale index of the project based on the parameters;

[0056] Existing technologies use a unified large language model to identify and process engineering survey and design data. This can lead to two problems: one is a waste of computing resources, as the processing power of a large language model far exceeds the required capacity for small projects, resulting in idle computing resources; the other is insufficient scale. For large projects, a large language model may not fully capture and process all the important information in the archive, affecting the accuracy and completeness of the recognition results. Therefore, it is crucial to select an appropriate model based on the scale of the engineering survey and design archive.

[0057] By calculating the scale index of the project, we can accurately understand the overall scale and complexity of the engineering survey and design data to be archived, which will help to select the appropriate management model based on the scale of the data, thereby achieving a reasonable allocation of computing resources, avoiding resource waste, and ensuring the accuracy and completeness of the identification results.

[0058] Furthermore, the method for obtaining the scale indicator is:

[0059] Step S21, dividing the engineering survey and design data to be archived into documents, pictures, drawings and videos;

[0060] Step S22, obtaining the total word count W and document conversion coefficient α1 of the document; obtaining the file size S, picture-drawing conversion coefficient α2, and picture-drawing resolution adjustment factor β of the picture-drawing; obtaining the file size V, video conversion coefficient α3, and video resolution adjustment factor γ of the video;

[0061] Step S23: Calculate the scale index based on the parameters obtained in step S22.

[0062] When considering the size calculation of a document, the total number of words is used as a statistical indicator to calculate the size of the document;

[0063] When considering the recognition scale of images, drawings, and videos, the present invention not only considers file size but also file resolution. Resolution reflects the clarity and level of detail of an image or video, and is crucial for assessing the recognition scale of the material.

[0064] The following is a specific example of how to obtain scale indicators:

[0065] 1. Set the conversion coefficient and resolution adjustment factor

[0066] document:

[0067] Original statistical indicators: word count W (10,000 words);

[0068] Conversion factor: α1 (assuming 1MB is equivalent to α ten thousand words);

[0069] Converted metric: W'=W / α1 (convert the number of words to an equivalent value in MB);

[0070] Picture drawings:

[0071] Original statistical indicators: file size (S, unit: MB) and resolution (R, unit: number of pixels or DPI / PPI);

[0072] Conversion factor: α2 (convert file size to MB);

[0073] Resolution adjustment factor: β (set an adjustment factor based on the resolution; the higher the resolution, the larger the β value);

[0074] Converted index: S'=S / α2×β (adjust the file size according to the resolution);

[0075] video:

[0076] Original statistical indicators: file size (V, unit: MB) and resolution (Res, usually expressed as horizontal pixels × vertical pixels);

[0077] Conversion factor: α3 (convert file size to MB);

[0078] Resolution adjustment factor: γ (set an adjustment factor based on the video resolution; the higher the resolution, the larger the γ value);

[0079] Converted index: V'=V / α3×γ (adjust the file size according to the resolution);

[0080] 2. Calculate the resolution adjustment factor

[0081] The settings for the resolution adjustment factors β and γ can be determined based on practical circumstances and experience. For example, a range of resolution intervals can be defined, with each interval assigned an adjustment factor. Higher resolutions require larger adjustment factors to reflect the complexity of identifying and processing high-resolution data.

[0082] In this embodiment, the image drawing resolution adjustment factor β is:

[0083] ;

[0084] Among them, ω TP is the weight of the image, N TP is the number of pictures, S TPi is the size of the i-th picture, β TPi is the resolution of the i-th picture, ω TZ is the weight of the drawing, N TZ is the number of drawings, S TZj is the size of the jth drawing, β TZj is the resolution of the jth drawing; the drawings contain more and more important details and information, so ω TZ >ω TP .

[0085] The value of the image and drawing resolution adjustment factor β is determined by the image and drawing resolution. Higher resolutions contain more information and sharper details, requiring more computing resources and processing time for recognition and processing. To reflect the complexity of recognizing and processing high-resolution data, the β value is set accordingly high.

[0086] By setting a reasonable β value, the above formula ensures that high-resolution images and drawings are fully processed during the recognition process, thereby improving recognition accuracy and completeness. It also provides managers with an intuitive and accurate assessment tool, allowing them to quickly understand the complexity of images and drawings and the required computing resources.

[0087] In this embodiment, the video resolution adjustment factor γ is set to:

[0088] ;

[0089] Among them, ω Vk is the weight of the kth video, V k is the size of the kth video, γ Vk is the resolution of the kth video, N V is the number of videos;

[0090] The value of the video resolution adjustment factor, γ, is closely related to the video resolution. Similar to images and drawings, high-resolution videos require more computing resources and processing time for recognition and processing. Therefore, the γ value is adjusted accordingly to reflect this complexity.

[0091] By setting a reasonable γ value, we can ensure that high-resolution videos are fully processed during the recognition process, thereby improving recognition accuracy and completeness. At the same time, this also provides managers with intuitive information about the complexity of video materials and the required computing resources, helping them make more reasonable archive management decisions.

[0092] 3. Calculate the total value of identification scale

[0093] After setting the conversion coefficient and resolution adjustment factor, the statistical indicators of all data can be converted into a unified unit (MB) and the total value of the identification scale can be calculated. The calculation formula is as follows:

[0094] Total Recognition Scale (TRS) = Σ(W'+S'+V') = W / α1+S×β / α2+V×γ / α3.

[0095] Among them, Σ represents the summation operation, and W', S', and V' represent the statistical indicators after conversion of documents, pictures and drawings, and video materials, respectively.

[0096] 4. Example Calculation

[0097] Assume that the following engineering survey and design data are available:

[0098] Documents: The total word count is 100,000 words, and the conversion coefficient α1 is set to 0.001 (i.e., 1MB is equivalent to 100,000 words);

[0099] Image and drawing materials: Total file size 5MB, resolution 300DPI (set the adjustment factor β to 1 for resolutions below 300DPI, 1.5 for resolutions between 300-600DPI, and 2 for resolutions above 600DPI);

[0100] Video material: Total file size 3MB, resolution 1920×1080 (set the adjustment factor γ to 1 for 1080P and below, 1.5 for 4K, and 2 for 8K and above);

[0101] The converted statistical indicators are:

[0102] Document data: W'=10 / 0.001=10MB;

[0103] Image and drawing materials: S' = 5 / 1 × 2 = 10MB (because the resolution is 300DPI, which is in the 300-600DPI range, the adjustment factor β is 1.5. However, to simplify the calculation, the upper limit is directly taken as 2. The file unit is already MB, so the conversion coefficient α2 = 1);

[0104] Video data: V' = 3 / 1 × 1 = 3MB (because the resolution is 1920 × 1080, which is 1080P or lower, the adjustment factor γ is 1, and the file unit is already MB, so the conversion coefficient α3 = 1);

[0105] Therefore, the total value of the scale indicator is:

[0106] TRS=10+10+3=23MB;

[0107] This value can be used as a reference indicator to measure the scale and complexity of engineering survey and design data, and takes into account the impact of resolution on the scale indicator.

[0108] Step S3, model selection: selecting a model for engineering survey and design data management according to the scale index, specifically, when the scale index is less than or equal to a preset threshold, selecting a first management model; when the scale index is greater than the preset threshold, selecting a second management model;

[0109] The purpose of selecting the preset threshold is to effectively classify engineering survey and design data into categories of different sizes or complexities. By setting a specific value (e.g., 5000), data can be clearly divided into "small" or "simple" categories (size index ≤ 5000) and "large" or "complex" categories (size index > 5000).

[0110] The first management model may be more inclined to choose LSTM or simpler RNN variants, because these models perform well when processing small-scale and low-complexity sequence data, and the parameter settings are relatively simple, which is conducive to reducing computing resource consumption.

[0111] In this embodiment, the first management model is an LSTM model, and the specific parameters are set as follows:

[0112] The number of hidden layers is 2, the number of neurons in each layer is 256, the activation function is sigmoid function, the optimizer is Adam, the learning rate is 0.001, the loss function is the cross entropy loss function, and the batch size is 32.

[0113] Specifically,

[0114] Number of hidden layers: 1 or 2. When the dataset is small and the task is relatively simple, fewer hidden layers can reduce the complexity of the model and prevent overfitting.

[0115] Number of neurons per layer: 128 or 256. On small datasets, too many neurons may lead to overfitting of the model. Therefore, choosing an appropriate number of neurons can reduce computational complexity while maintaining model performance.

[0116] Optimizer: Adam or SGD (with momentum). Adam optimizer performs well in most cases, while SGD (with momentum) may also perform well on simple tasks. In practical applications, the appropriate optimizer can be selected based on the convergence speed and stability of the model.

[0117] Learning rate: 0.001 or 0.002. In small-scale engineering survey and design archive management, a larger learning rate can accelerate the convergence process of the model, but it is also necessary to pay close attention to the model training process to prevent overfitting;

[0118] Loss function: Cross-entropy loss function. For simple tasks and small datasets, the cross-entropy loss function is usually sufficient to measure the performance of the model;

[0119] Activation function: ReLU function. ReLU function usually performs well on simple tasks and has high computational efficiency.

[0120] Batch size: 16 or 32. For small-scale engineering survey and design archive management, a smaller batch size helps the model better capture detailed features in the data, but it also increases training time. Therefore, it is necessary to select an appropriate batch size based on hardware resources and training time.

[0121] The second management model may be more inclined to choose pre-trained models such as Transformer or BERT, because these models perform well in processing large-scale and complex data, and can capture more language features through complex parameter settings to improve recognition accuracy.

[0122] In this embodiment, the second management model is a Transformer model, and the specific parameters are set as follows:

[0123] The number of hidden layers is 3, the number of neurons in each layer is 512, the optimizer is Adam, the learning rate is 0.0001, the loss function is the cross entropy loss function, the activation function is GELU function, and the batch size is 64.

[0124] Specifically,

[0125] Number of hidden layers: 3 or 4. When the dataset is large and the task is complex, increasing the number of hidden layers can help the model better capture the deep features in the data and improve the model performance.

[0126] Number of neurons per layer: 512 or 1024. Increasing the number of neurons can enhance the model's representation capabilities, but it also increases the model's computational complexity and memory consumption. On large-scale datasets, more neurons generally lead to better performance.

[0127] Optimizer: AdamW or RAdam. AdamW is a variant of the Adam optimizer that improves on weight decay and helps prevent model overfitting. RAdam is another improved Adam optimizer that excels at handling non-convex optimization problems.

[0128] Learning rate: 0.0005 or 0.0001. On large-scale datasets, a smaller learning rate is usually more stable and helps the model find a better global optimal solution;

[0129] Loss function: Cross-entropy loss function (with weights) or Focal Loss. When there is a class imbalance problem in the dataset, you can use the weighted cross-entropy loss function or Focal Loss to focus on minority class samples and improve the generalization ability of the model.

[0130] Activation function: GELU function or Leaky ReLU function. The GELU function performs well in most cases, but the Leaky ReLU function can maintain a certain gradient when processing negative inputs, which helps prevent the model from falling into a dead zone during training;

[0131] Batch size: 64 or 128. A larger batch size can speed up the training process, but it may also cause the model to fall into a local optimal solution during training. In practical applications, it is necessary to select an appropriate batch size based on hardware resources and dataset size.

[0132] The above parameter settings are optimized for the present invention. In practice, these parameters can be flexibly adjusted based on specific task requirements, dataset size, hardware resources, and experimental feedback. Furthermore, the model needs to be fully trained and validated to ensure its accuracy and stability in practical applications.

[0133] The selection of parameters for the two models was based on a study of engineering archives of varying sizes, aiming to precisely match computing resources. For small-scale engineering archives, the first management model, with its smaller size and faster processing speed, effectively utilizes resources. For large-scale engineering archives, the second management model, with its greater processing power and higher accuracy, ensures the integrity and reliability of the recognition results.

[0134] Step S4, file identification: inputting the engineering survey and design data into the selected management model, identifying the engineering survey and design data through the management model, and obtaining archive information;

[0135] The management model utilizes natural language processing (NLP) technology to deeply analyze and identify the engineering survey and design data, extracting and organizing key information from the data, including but not limited to: project name, project location, design firm, design date, project type, scale parameters (such as area, length, height, capacity, etc.), primary materials, construction requirements, safety regulations, environmental standards, etc. The extracted key information is formatted and structured according to a pre-set archiving information template or data structure, resulting in clear and complete archiving.

[0136] Ensure that the archived information contains both necessary metadata (such as file type, format, size, creation time, etc.) and the core content and characteristics of the engineering survey and design data to facilitate subsequent search, retrieval, and data analysis.

[0137] This type of archived information not only allows managers to quickly understand the key contents of engineering survey and design data, but also provides strong data support for subsequent project management, quality control, cost control, and other aspects. Furthermore, structured archived information is easier to efficiently store, retrieve, and analyze through database systems.

[0138] Step S5, data storage: save the archived information into a database.

[0139] This invention utilizes natural language processing (NLP) technology to identify and convert engineering survey and design data into computer language based on data type and scale. This process enables intelligent processing of engineering survey and design data, enabling computers to understand and process complex and diverse data, facilitating subsequent data storage, retrieval, and analysis.

[0140] This invention flexibly selects appropriate NLP models for processing based on the scale and complexity of engineering survey and design data, achieving rational allocation and efficient utilization of computing resources. This not only improves the efficiency of file management but also significantly reduces operating costs, providing a new solution for intelligent management in the field of engineering survey and design.

[0141] Example 2

[0142] like Figure 2 As shown, the present invention also proposes an NLP-based engineering survey and design archive management system, which uses the NLP-based engineering survey and design archive management method as described in any one of Example 1, including the following modules:

[0143] Data collection module: obtain engineering survey and design data to be archived;

[0144] Scale calculation module: connected to the data collection module, used to obtain document parameters, picture drawing parameters and video parameters based on the engineering survey and design data to be archived, and calculate the scale index of the project based on the parameters;

[0145] A model selection module, connected to the scale calculation module, is used to select a model for engineering survey and design data management based on the scale index. Specifically, when the scale index is less than or equal to a preset threshold, a first management model is selected; when the scale index is greater than the preset threshold, a second management model is selected;

[0146] File recognition module: connected to the model selection module, used to input the engineering survey and design data into the selected management model and recognize the engineering survey and design data into computer language;

[0147] Data saving module: connected to the file identification module, used to save the identification results output by the management model into the database.

[0148] Example 3

[0149] An electronic device, comprising:

[0150] processor and memory;

[0151] The processor is configured to execute the steps of the NLP-based engineering survey and design archive management method as described in any one of the first embodiments by calling the program or instruction stored in the memory.

[0152] Example 4

[0153] A computer-readable storage medium comprising computer program instructions, wherein the computer program instructions enable a computer to execute the steps of the NLP-based engineering survey and design archive management method as described in any one of Example 1.

[0154] The computer-readable storage medium may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0155] It should be noted that the terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present invention. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular and may also include the plural. The terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method or device comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "comprise a..." do not exclude the presence of other identical elements in the process, method or device comprising the elements.

[0156] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be a communication between the internal parts of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for managing engineering survey and design archives based on NLP, characterized in that: The specific steps include: Step S1, obtaining engineering survey and design data to be archived; Step S2: Obtain document parameters, picture drawing parameters, and video parameters based on the engineering survey and design data to be archived, and calculate the scale index of the project based on the parameters; the method for obtaining the scale index is: Step S21, dividing the engineering survey and design data to be archived into documents, pictures, drawings and videos; Step S22, obtaining the total word count W of the document and the document conversion coefficient α1; Obtain the file size S of the picture drawing, the picture drawing conversion coefficient α2, and the picture drawing resolution adjustment factor β; obtain the file size V of the video, the video conversion coefficient α3, and the video resolution adjustment factor γ; Step S23: Calculate the scale index TRS based on the parameters obtained in step S22. The formula is: TRS=W / α1+S×β / α2+V×γ / α3; ； ； Among them, ω TP is the weight of the image, N TP is the number of pictures, S TPi is the size of the i-th picture, β TPi is the resolution of the i-th picture, ω TZ is the weight of the drawing, N TZ is the number of drawings, S TZj is the size of the jth drawing, β TZj is the resolution of the jth drawing, ω Vk is the weight of the kth video, V k is the size of the kth video, γ Vk is the resolution of the kth video, N V is the number of videos; Step S3, selecting a model for engineering survey and design data management according to the scale index, specifically, selecting a first management model when the scale index is less than or equal to a preset threshold, and selecting a second management model when the scale index is greater than the preset threshold; Step S4: inputting the engineering survey and design data into the selected management model, identifying the engineering survey and design data through the management model, and obtaining archive information; Step S5: Save the archived information into a database.

2. The NLP-based engineering survey and design archive management method according to claim 1, characterized in that: The first management model is an LSTM model, and the second management model is a Transformer model.

3. The NLP-based engineering survey and design archive management method according to claim 1, characterized in that: The specific parameters of the first management model are set as: The number of hidden layers is 2, the number of neurons in each layer is 256, the activation function is sigmoid function, the optimizer is Adam, the learning rate is 0.001, the loss function is the cross entropy loss function, and the batch size is 32.

4. The NLP-based engineering survey and design archive management method according to claim 1, characterized in that: The specific parameters of the second management model are set as: The number of hidden layers is 3, the number of neurons in each layer is 512, the optimizer is Adam, the learning rate is 0.0001, the loss function is the cross entropy loss function, the activation function is GELU function, and the batch size is 64.

5. An NLP-based engineering survey and design archive management system, using the NLP-based engineering survey and design archive management method according to any one of claims 1 to 4, characterized in that: Specifically, it includes the following modules: Data collection module: obtain engineering survey and design data to be archived; Scale calculation module: connected to the data collection module, used to obtain document parameters, picture drawing parameters and video parameters based on the engineering survey and design data to be archived, and calculate the scale index of the project based on the parameters; A model selection module, connected to the scale calculation module, is used to select a model for engineering survey and design data management based on the scale index. Specifically, when the scale index is less than or equal to a preset threshold, a first management model is selected; when the scale index is greater than the preset threshold, a second management model is selected; File recognition module: connected to the model selection module, used to input the engineering survey and design data into the selected management model and recognize the engineering survey and design data into computer language; Data saving module: connected to the file identification module, used to save the identification results output by the management model into the database.

6. An electronic device, comprising: A processor and a memory, characterized in that the processor is used to execute the steps of the NLP-based engineering survey and design archive management method as described in any one of claims 1 to 4 by calling the program or instructions stored in the memory.

Images