Intelligent review method and system for construction plan calculation

Abstract

This invention provides an intelligent audit method and system for construction plan calculation documents. The system includes a semantic table for audit content based on natural language processing (NLP), a manually curated system rule base, document parsing based on optical character recognition (OCR), document annotation based on computer vision (CV), and algorithm-based model training. Based on natural language processing (NLP), combined with an audit rule base and OCR recognition technology, this method can verify the results of calculation documents in different types of construction plans, review the correctness of standard references, and other functions. It also outputs an audit report for manual review, accurately eliminating calculation content risks and improving the work efficiency of technical personnel.

Description

Technical Field

[0001] The present invention belongs to the field of building construction, and in particular relates to an intelligent review method and system for construction plan calculation books. Background Art

[0002] A construction plan is the implementation plan for a project and is a document that construction company technicians routinely compile in large quantities. This includes organizational plans, staffing plans, technical plans, safety plans, material supply plans, and calculations. Calculations are a crucial component of the construction plan and are crucial to safety management during the construction process.

[0003] Typically, calculations for construction plans are compiled by on-site technicians and then submitted to technical management for proofreading and review. Due to the complexity and difficulty of verifying these calculations, this review process is often time-consuming and labor-intensive. Traditional review processes are labor-intensive, time-consuming, and inefficient for technicians. Improving work quality and reducing working hours are worthy research issues. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent review method and system for construction plan calculation documents. This method and system uses artificial intelligence to replace technical personnel in reviewing calculation documents in construction plans, thereby reducing the workload of technical personnel and improving the quality and efficiency of construction plan review.

[0005] The present invention is achieved through the following technical solutions:

[0006] An intelligent review method for construction plan calculation sheet, comprising the steps of:

[0007] Organize the manually reviewed construction plan sample materials and sort them according to different construction plan types;

[0008] Confirm the content and extraction format of sample materials and organize them into a semantic table for audit content;

[0009] Confirm the system rule base for AI auditing;

[0010] Parsing and identifying construction plan documents to be reviewed and converting document formats based on optical character recognition technology;

[0011] Annotate construction plan documents to be reviewed based on computer vision technology, and extract annotations from the identified features;

[0012] Invite experts to manually review the extracted information to ensure the accuracy of the annotations;

[0013] Conduct model training on the annotated and extracted information based on various algorithm models, including comparative review of content semantic tables and rule bases;

[0014] Optimize the content to be improved after model training;

[0015] Manually modify content found to be incorrect by the system review;

[0016] Classify the status of construction plans to be reviewed;

[0017] Export system-audited and manually-audited construction plan audit reports.

[0018] In an embodiment of the present invention, the algorithms involved in the step of performing model training on the annotated extracted information include:

[0019] The structured extraction algorithm extracts annotated information with high template matching and obvious features and performs model training based on the content semantic table and rule base;

[0020] The sequence labeling module includes a series of basic sequence labeling technologies. For labeled information with unclear features, it extracts and performs model training based on the content semantic table and rule base.

[0021] In an embodiment of the present invention, the step of optimizing the content to be improved after model training includes:

[0022] Feature extraction reduces the dimension of the vector space without damaging the core information, simplifies calculations, and improves the speed and efficiency of text processing;

[0023] Call the intelligent classification module and use the following basic classifiers for different extraction scenarios, including but not limited to: support vector machine, xgboost, TextCNN;

[0024] Optimization review is carried out using a computational graph. The review rules correspond to the nodes on the computational graph. Each node is expressed by an SPO triple. The document content and extracted results flow through the computational graph in the form of data flow to complete the optimization review.

[0025] An intelligent review system for construction plan calculations, comprising:

[0026] A memory for storing program codes corresponding to the processing process of the intelligent review method for the construction plan calculation sheet as described above;

[0027] A processor is configured to run the program code.

[0028] An intelligent review system for construction plan calculations, comprising:

[0029] The review content semantic table based on natural language processing technology is used to identify and extract the format of text content based on the manual review of construction plan sample materials using natural language processing technology, and organize it into a review content semantic table according to different construction plan types;

[0030] A manually curated system rule base is used to confirm the system rule base for artificial intelligence review based on the rules for traditional manual review of construction plans;

[0031] Document parsing based on optical character recognition technology is used to parse and recognize text in construction plans to be reviewed in different document formats using optical character recognition technology;

[0032] Document annotation based on computer vision technology is used to extract the characteristic content of the construction plan to be reviewed after analysis and identification. Experts are invited to review the annotation information to ensure the accuracy of the annotation.

[0033] Algorithm-based model training is used to train models for labeled solutions to be reviewed through structured extraction algorithms and sequence labeling modules, and to optimize the content that needs to be improved after training based on the system rule base.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. Convert offline construction plan review to an online intelligent centralized review method. By solidifying the construction plan review rules and providing a content semantic table that can quickly extract target content, efficient and high-quality review of construction plan calculation documents is achieved.

[0036] 2. Based on content comparison, intelligently generate construction plan review reports and optimization results for on-site production personnel to refer to and implement;

[0037] 3. Integrate the machine review and human review processes, use artificial intelligence to accurately identify and review data, and then hand it over to manual secondary review after processing. This working method has extremely high freedom and flexibility, accurately breaks down various computing risks, and improves review efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 The present invention provides a flowchart of an intelligent method for reviewing construction plan calculation documents. DETAILED DESCRIPTION

[0040] Name explanation:

[0041] NLP, short for Natural Language Processing, is a discipline that studies language-related aspects of human-computer interaction. Based on the technical difficulty of implementation, these systems can be categorized into three types: simple matching, fuzzy matching, and paragraph comprehension.

[0042] OCR: The full name is Optical Character Recognition, and its Chinese name is optical character recognition. It refers to the process in which an electronic device (such as a scanner or digital camera) examines characters printed on paper, determines their shape by detecting dark and light patterns, and then uses character recognition methods to translate the shape into computer text. That is, for printed characters, the text in the paper document is optically converted into a black and white dot matrix image file, and recognition software is used to convert the text in the image into text format for further editing and processing by word processing software.

[0043] CV (Computer Vision) is a field of artificial intelligence that simulates biological vision using computers and related equipment. Its research goal is to enable computers to perceive three-dimensional environments through two-dimensional images. Computer vision uses image processing, signal processing, probability and statistics, computational geometry, neural networks, machine learning theory, and computer information processing techniques to analyze and process visual information.

[0044] CRF: The full name is conditional random field algorithm, and the Chinese name is conditional random field algorithm. It is a mathematical algorithm proposed in 2001 and is based on a probabilistic graphical model that follows Markov properties.

[0045] BiLSTM+CRF is a currently popular sequence labeling algorithm. It combines BiLSTM and CRF, allowing the model to consider the correlation between the previous and next sequences like CRF, while also having the feature extraction and fitting capabilities of LSTM.

[0046] Transformer: is a model that uses the attention mechanism to increase the speed of model training.

[0047] SVM: The full name is Support Vector Machine, and the Chinese name is Support Vector Machine. It is a type of generalized linear classifier that performs binary classification on data in a supervised learning manner. Its decision boundary is the maximum margin hyperplane solved for the learning samples.

[0048] Xgboost: It is an optimized distributed gradient boosting library designed to be efficient, flexible, and portable. It implements machine learning algorithms under the Gradient Boosting framework. XGBoost provides parallel tree boosting (also known as GBDT, GBM), which can quickly and accurately solve many data science problems. The same code runs on major distributed environments (Hadoop, SGE, MPI) and can handle problems beyond billions of examples.

[0049] TextCNN: It uses a convolutional neural network for text classification;

[0050] SPO triple: Subject, Predicate, Object, which is the basic unit of knowledge representation, abbreviated as SPO <Subject - subject, Predicate - predicate, Object - object>.

[0051] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0052] The present invention provides an intelligent review method and system for construction plan calculation sheets. Based on natural language processing technology (NLP), combined with a review rule library and OCR recognition technology, it can achieve functions such as result verification of calculation sheets in different types of construction plans and review of the correctness of code citations, and output a review report for manual review, accurately breaking through the risks of calculation content and improving the work efficiency of technical personnel.

[0053] To achieve the above object, the technical solution adopted by the present invention is:

[0054] Example 1:

[0055] An intelligent review system for construction plan calculation books, including a review content semantic table based on natural language processing technology (NLP), a system rule base based on manual organization, document parsing based on optical character recognition technology (OCR), document annotation based on computer vision technology (CV), model training based on algorithms, etc. The review content semantic table based on natural language processing technology (NLP) is characterized in that: on the basis of manual review of construction plan sample materials, natural language processing technology (NLP) is used to identify and extract the format of text content, and the review content semantic table is organized according to the preparation basis (paragraph), project overview (paragraph), architectural design overview (table), calculation book (paragraph), etc.; the system rule base based on manual organization is characterized in that: according to the rules of traditional manual review of construction plans, the system rule base of artificial intelligence review is confirmed, such as the compliance of the preparation basis, the completeness of the project overview, the standardization of architectural design parameters, the accuracy of the calculation book results, etc.; the system rule base based on optical character The document parsing feature of the recognition technology (OCR) is: using optical character recognition technology (OCR) to parse and identify the text of the construction plans to be reviewed in different document formats; the document annotation feature based on computer vision technology (CV) is: for the characteristic content of the construction plans to be reviewed after parsing and identification, document annotation is extracted by using computer vision technology (CV), and experts are invited to review the annotation information to ensure the accuracy of the annotation; the algorithm-based model training feature is: through the structured extraction algorithm and sequence annotation module, the model training of the annotated plans to be reviewed is carried out, and the content to be improved after training is optimized based on the system rule library.

[0056] The working principle of the present invention is: through natural language processing technology (NLP), an audit content semantic table is established on the basis of sample materials, and a system rule base is organized based on the manual audit situation; after the construction plan to be audited is parsed and annotated using optical character recognition technology (OCR) and computer vision technology (CV), the model of the construction plan to be audited is trained and improved by comparing the audit content semantic table and the system rule base; in addition, by connecting the machine review and human review processes, manual review of the machine review results can be realized, thereby improving the freedom, flexibility and accuracy of the calculation book audit work.

[0057] Example 2: See Figure 1 The intelligent review steps of the construction plan calculation sheet of the present invention are as follows:

[0058] 1) The first step is to organize the manually reviewed construction plan sample materials according to different construction plan types, such as cantilever scaffolding construction plan, construction organization design construction plan, high-support formwork construction plan, etc.

[0059] 2) The second step is to confirm the content and extraction format of the sample materials and organize them into a semantic table of audit content, such as preparation basis (paragraph), project overview (paragraph), architectural design overview (table), calculation sheet (paragraph), etc.

[0060] 3) The third step is to confirm the rules for manual review, such as the compliance of the preparation basis, the completeness of the project overview, the standardization of architectural design parameters, the accuracy of calculation results, etc., and establish a system rule base;

[0061] 4) The fourth step is to parse and identify the construction plan documents to be reviewed and convert the document format based on optical character recognition (OCR) technology;

[0062] 5) In the fifth step, the construction plan documents to be reviewed are annotated using computer vision (CV) technology, and the feature content after analysis and recognition is annotated and extracted;

[0063] 6) In the sixth step, experts are invited to manually review the extracted information to ensure the accuracy of the annotations;

[0064] 7) The seventh step is to train the model based on various algorithm models for the extracted information, including comparative review of the content semantic table and the rule base;

[0065] 8) The eighth step is to optimize the content to be improved after model training;

[0066] 9) Step 9: Manually modify any content found to be incorrect by the system review;

[0067] 10) Step 10: Classify the status of the construction plans to be reviewed, such as recognition failure, recognition in progress, recognition results, unreviewed, review in progress, system reviewed, manual review, etc.

[0068] 11) The eleventh step is to export the system-audited and manually-audited construction plan audit reports for easy review by the review experts and subsequent archiving.

[0069] Furthermore, in the seventh step, the model is trained on the annotation extraction information. The algorithms involved include:

[0070] 1) Structured extraction algorithm: For labeled information with high template matching and obvious features, it extracts and performs model training based on the content semantic table and rule base;

[0071] 2) Sequence labeling module, including a series of basic sequence labeling technologies: BiLSTM-CRF, CRF conditional random field, and Transformer. For labeled information with unclear features, it extracts and performs model training based on the content semantic table and rule base.

[0072] Furthermore, in the eighth step, the steps for establishing the optimization mechanism after model training are as follows:

[0073] 1) Feature extraction: reducing the dimension of the vector space without damaging the core information, simplifying calculations, and improving the speed and efficiency of text processing;

[0074] 2) Call the IDPS intelligent classification module to target different extraction scenarios, such as chapter location, paragraph extraction, and table extraction, using basic classifiers including but not limited to support vector machines (SVMs), xgboost, and TextCNN.

[0075] 3) Use the computational graph to conduct optimization review. The review rules correspond to the nodes on the computational graph. Each node is expressed by an SPO (Subject, Predicate, Object) triple. The document content and extracted results flow through the computational graph in the form of data flow to complete the optimization review.

[0076] Example 3:

[0077] An intelligent review system for construction plan calculations according to an embodiment of the present invention includes:

[0078] A memory for storing program codes corresponding to the processing of the intelligent review method for the construction plan calculation sheet according to the above embodiment;

[0079] A processor is configured to run the program code.

[0080] The processor may be a DSP (Digital Signal Processing) digital signal processor, an FPGA (Field-Programmable Gate Array) field programmable gate array, an MCU (Microcontroller Unit) system board, an SoC (system on a chip) system board, or a PLC (Programmable Logic Controller) minimum system including I / O.

[0081] Compared with the prior art, the present invention has the following advantages:

[0082] 1. Convert offline construction plan review to an online intelligent centralized review method. By solidifying the construction plan review rules and providing a content semantic table that can quickly extract target content, efficient and high-quality review of construction plan calculation documents is achieved.

[0083] 2. Based on content comparison, intelligently generate construction plan review reports and optimization results for on-site production personnel to refer to and implement;

[0084] 3. Integrate the machine review and human review processes, use artificial intelligence to accurately identify and review data, and then hand it over to manual secondary review after processing. This working method has extremely high freedom and flexibility, accurately breaks down various computing risks, and improves review efficiency.

[0085] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It is apparent to those skilled in the art that various changes, modifications, substitutions, and variations to these embodiments may be made without departing from the principles and spirit of the present invention, and these changes and modifications still fall within the scope of protection of the present invention.

[0086] From the foregoing it will be seen that the present invention is well adapted to attain all of the objects and aims set forth above, as well as other advantages which are obvious and inherent to the structure. It will be understood that certain features and subcombinations are useful and may be employed without reference to other features and subcombinations. This is within the scope of the present invention.

Claims

1. An intelligent review method for construction plan calculation, characterized in that: Including steps: Organize the manually reviewed construction plan sample materials and sort them according to different construction plan types; Confirm the content and extraction format of sample materials and organize them into a semantic table for audit content; Confirm the system rule base for AI auditing; Parsing and identifying construction plan documents to be reviewed and converting document formats based on optical character recognition technology; Annotate construction plan documents to be reviewed based on computer vision technology, and extract annotations from the identified features; Invite experts to manually review the extracted information to ensure the accuracy of the annotations; Conduct model training on the annotated and extracted information based on various algorithm models, including comparative review of content semantic tables and rule bases; The algorithms involved in the step of training the model on the annotated extracted information include: The structured extraction algorithm extracts annotated information with high template matching and obvious features and performs model training based on the content semantic table and rule base; The sequence labeling module includes a series of basic sequence labeling technologies. For labeled information with unclear features, it extracts and performs model training based on the content semantic table and rule base. Optimize the content to be improved after model training; The step of optimizing the content to be improved after model training includes: Feature extraction reduces the dimension of the vector space without damaging the core information, simplifies calculations, and improves the speed and efficiency of text processing; Call the intelligent classification module and use the following basic classifiers for different extraction scenarios, including but not limited to: support vector machine, xgboost, TextCNN; Use computational graphs to conduct optimization audits. Audit rules correspond to nodes on the computational graph. Each node is expressed through an SPO triple. Document content and extracted results flow through the computational graph in a data flow to complete the optimization audit. Manually modify content found to be incorrect by the system review; Classify the status of construction plans to be reviewed; Export system-audited and manually-audited construction plan audit reports.

2. An intelligent review system for construction plan calculations, characterized by: include: A memory for storing program codes corresponding to the processing process of the intelligent review method for the construction plan calculation sheet as claimed in claim 1; A processor is configured to run the program code.

Images