Project proposal level-aware knowledge construction and parameter tracing method and device

Abstract

The present application relates to the field of project proposal data processing, and provides a project proposal hierarchical perception knowledge construction and parameter tracing method and device, which comprises page layout area detection on the project proposal, post-processing correction based on a wind power professional symbol confusion table and a domain language model; layout features of each paragraph are extracted from the semantic area of the post-processing correction, and a document structure tree is reconstructed accordingly; based on the reconstructed document structure tree, knowledge block segmentation is performed according to a knowledge block length and a preset maximum threshold; named entity recognition is performed on each knowledge block obtained by the segmentation, forming a wind power knowledge association graph for parameter tracing and joint retrieval; the neighbor context of each knowledge block and the wind power knowledge association graph thereof is jointly encoded to obtain a context-enhanced semantic vector, which is written into a vector database as metadata in synchronization with a global entity index, an entity type, an association edge and a weight thereof, so that multi-modal knowledge association can be established and vectorization retrieval can be supported.

Description

Technical Field

[0001] This invention relates to the field of project proposal data processing, and in particular to a method and apparatus for hierarchical knowledge construction and parameter tracing of project proposals. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The wind power project proposal is a crucial foundational document for wind power project initiation, investment calculation, and scheme comparison. It typically includes multimodal content such as wind resource assessment reports, meteorological tower observation data, topographic maps, turbine selection parameters, power generation prediction models, and economic analysis tables. This type of document features a mixed layout of text, tables, drawings, formulas, and figure captions, and is densely packed with technical terminology. Furthermore, there are numerous implicit parameter references and semantic relationships between different chapters. Existing knowledge base construction schemes for general-purpose documents mostly rely on general OCR and fixed-length segmentation strategies, making them difficult to directly adapt to wind power engineering documents. In addition, general embedding models have limited ability to represent concepts and parameter transmission chains in the wind power field, leading to unstable results in subsequent semantic retrieval, parameter tracing, and question-answering recall. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and apparatus for constructing hierarchical knowledge and tracing parameters in project proposals. This method and apparatus can perform high-precision structural analysis of wind power project proposals, maintain semantic integrity, establish multimodal knowledge associations, and support vectorized retrieval.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a method for hierarchical knowledge construction and parameter tracing in project proposals.

[0006] In one or more embodiments, a method for hierarchical knowledge construction and parameter tracing of project proposals is provided, including: The layout area of ​​the project proposal page is detected to obtain the semantic regions, their location coordinates and confidence scores, and post-processing corrections are performed based on the wind power professional symbol confusion table and domain language model; The layout features of each paragraph are extracted from the semantic region corrected by post-processing, and hierarchical judgment is made accordingly, thereby reconstructing the document structure tree; Based on the reconstructed document structure tree, the segmentation priority sequence is set as chapter title - section title - clause title - blank line - newline - period - space, and knowledge blocks are segmented according to the knowledge block length and the preset maximum threshold. Named entity recognition is performed on each knowledge block obtained by segmentation, wind power-related entities are extracted and a global entity index is established, and association edges are established based on the semantic similarity between any two knowledge blocks as weights, forming a wind power knowledge association graph for parameter tracing and joint retrieval. The context of each knowledge block and its neighboring context in the wind power knowledge association graph are jointly encoded to obtain a context-enhanced semantic vector. This vector, along with the global entity index, entity type, associated edges and their weights, is synchronously written into the vector database as metadata for parameter tracing.

[0007] In one implementation, the semantic region includes a text area, a table area, a drawing area, a formula area, and a figure annotation area.

[0008] As one implementation method, during the post-processing correction based on the wind power professional symbol confusion table and domain language model, the wind power professional symbols are corrected, and the correction result with the highest score is selected; the correction score is expressed as follows: ; in, Indicates the candidate correction results; This represents the character confidence score obtained based on character obfuscation relationships; This represents the contextual matching score of the candidate results under the language model in the wind power field. This indicates the matching score between the candidate results and the wind power terminology dictionary; , , These are the weighting coefficients for each scoring item.

[0009] As one implementation method, the layout features include at least font size, bolding status, left indentation, first-line indentation, paragraph spacing, numbering style, and page number; candidate titles are first-level filtered based on the numbering style; paragraphs without numbers but meeting the title morphological features are identified and supplemented with titles; parent-child node relationships are established according to title levels, and body text, tables, formulas, and figure captions located between adjacent titles are assigned to the nearest parent title node; when numbering breaks, missing levels, or scanning defects are detected, hierarchical completion is performed using the semantic continuity of preceding and following paragraphs and the continuity of the layout.

[0010] As one implementation method, a hierarchy determination function is used for hierarchy determination. The hierarchy determination function is as follows: ; in, Represents the i-th paragraph This indicates the font and layout style characteristics of the paragraph; This indicates the indentation characteristics of the paragraph; This indicates the numbering pattern characteristics of the paragraph; Indicate the key features of this paragraph; This represents the hierarchical decision function; Paragraph The corresponding hierarchical category.

[0011] As one implementation method, if the length of the current knowledge block does not exceed the maximum threshold Lmax, it is output directly; if the length of the current knowledge block exceeds the threshold Lmax, it is divided into sub-blocks according to the current highest priority anchor point, and the same operation is recursively performed on the sub-blocks; if there are still no available segmentation positions under all priorities, it is forcibly segmented according to the number of characters; an overlap window of 10% to 20% of the maximum threshold Lmax is reserved between adjacent knowledge blocks to ensure the semantic continuity of the boundary, and the proportion of complete sentences in the block is not less than 95% as the semantic integrity control indicator.

[0012] As one implementation method, in the process of building a wind power knowledge association graph, candidate parameters are extracted from text blocks, table blocks, drawing blocks, and formula blocks. Based on the parameter name, unit, context keywords, and chapter, the candidate parameters are identified as entities and categorized into measured parameters, intermediate fitting parameters, height extrapolation parameters, equipment parameters, correction parameters, and conclusion parameters. A parameter template is established for each parameter slot in the formula, and candidate parameters are matched with the formula slots. When a candidate parameter simultaneously satisfies the conditions of name matching, unit matching, chapter semantic matching, and type matching, the corresponding candidate parameter is bound as the source parameter of the corresponding formula slot. Directed edges for parameter transmission are constructed according to the formula dependency order.

[0013] A second aspect of the present invention provides a device for hierarchical knowledge construction and parameter tracing of project proposals.

[0014] In one or more embodiments, a hierarchical knowledge construction and parameter tracing device for a project proposal includes: The page layout area detection module is used to detect the layout area of ​​the project proposal page, obtain the semantic area and its location coordinates and confidence, and perform post-processing correction based on the wind power professional symbol confusion table and domain language model. The document structure tree reconstruction module is used to extract the layout features of each paragraph from the post-processed semantic region, and to make hierarchical judgments accordingly, thereby reconstructing the document structure tree. The knowledge block segmentation module is used to segment knowledge blocks based on the reconstructed document structure tree. The segmentation priority sequence is set as chapter title - section title - clause title - blank line - newline - period - space. The knowledge block is segmented according to the knowledge block length and the preset maximum threshold. The wind power knowledge association graph construction module is used to perform named entity recognition on each knowledge block obtained by segmentation, extract wind power related entities and establish a global entity index, establish association edges based on the semantic similarity between any two knowledge blocks as weights, and form a wind power knowledge association graph for parameter tracing and joint retrieval. The vector database construction module is used to jointly encode the neighbor contexts of each knowledge block and its wind power knowledge association graph to obtain context-enhanced semantic vectors. These vectors, along with the global entity index, entity type, associated edges and their weights, are synchronously written into the vector database as metadata for parameter tracing.

[0015] A third aspect of the present invention provides a computer-readable storage medium.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the hierarchical knowledge construction and parameter tracing method of the project proposal as described above.

[0017] A fourth aspect of the present invention provides an electronic device.

[0018] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps in the hierarchical knowledge construction and parameter tracing method for project proposals as described above.

[0019] Compared with the prior art, the beneficial effects of the present invention are: To achieve the above objectives, the present invention adopts the following technical solution: This invention extracts semantic regions, their location coordinates, and confidence levels from project proposal pages and performs post-processing corrections. It then extracts the layout features of each paragraph and performs hierarchical judgment, reconstructs the document structure tree, and segments knowledge blocks. For each segmented knowledge block, it performs named entity recognition, extracts wind power-related entities, and establishes a global entity index, forming a wind power knowledge association graph for parameter tracing and joint retrieval. Finally, it jointly encodes the neighboring contexts of each knowledge block and its wind power knowledge association graph to obtain context-enhanced semantic vectors, which are then synchronously written into a vector database along with metadata. This achieves the goals of high-precision structural parsing of wind power project proposals, maintaining semantic integrity, establishing multimodal knowledge associations, and supporting vectorized retrieval. Attached Figure Description

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1This is a flowchart of the hierarchical knowledge construction and parameter tracing method for project proposals according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the hierarchical perception knowledge construction and parameter tracing device for project proposals according to an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device according to an embodiment of the present invention; Figure 4 This is a flowchart of the knowledge block segmentation process according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the wind power knowledge association map according to an embodiment of the present invention; Figure 6 This is a directed correlation subgraph for parameter tracing in an embodiment of the present invention; Figure 7 This is a schematic diagram of the association graph linkage recall retrieval in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0025] Figure 1 A schematic diagram illustrating the principle of the hierarchical knowledge construction and parameter tracing method for project proposals according to an embodiment of the present invention is provided. Based on... Figure 1 The hierarchical knowledge construction and parameter tracing method for project proposals in this embodiment may include the following steps S101 to S105.

[0026] The specific implementation process of steps S101 to S105 is as follows: Step S101: Perform layout area detection on the project proposal page to obtain the semantic region, its location coordinates and confidence level, and perform post-processing correction based on the wind power professional symbol confusion table and domain language model.

[0027] The semantic area includes the text area, table area, drawing area, formula area, and figure annotation area.

[0028] Export pages from PDFs, scanned documents, or CAD files. Use a layout detection model to classify the page into regions, outputting semantic regions such as text areas, table areas, drawing areas, formula areas, and annotation areas, along with their location coordinates and confidence scores. For wind power-specific symbols such as average wind speed, wind shear index, roughness length, and Weibull parameters, construct a wind power-specific character obfuscation table and perform post-processing corrections using a domain-specific language model to improve the accuracy of OCR recognition results in engineering contexts.

[0029] Page layout region detection can be represented as follows: divide the page image I into multi-label regions to obtain a region set R={r1, r2, ..., rn}, where each region ri belongs to one of the categories of text, table, drawing, formula or annotation; output the bounding box bi=(x1i, y1i, x2i, y2i) and category confidence for each region.

[0030] Specifically, the input page is first partitioned to obtain candidate regions for text, table, drawing, formula, and figure annotation areas. Then, the corresponding recognition path is called according to the region type: text recognition is performed in the text region, table structure restoration is performed in the table region, figure labels, coordinate annotations, and legend text are extracted in the drawing region, formula symbols and subscript / superscript structure are extracted in the formula region, and figure number and figure name are extracted in the figure annotation area. Subsequently, the recognition results are standardized for the wind power field, unifying units such as m / s, kW, MW, h, and % and mapping terms such as hub height, wind measurement height, and wake correction coefficient to preset field names. Finally, page numbers, region coordinates, region types, and recognition confidence scores are added to the corrected text results to form the basic evidence data required for subsequent structure tree reconstruction and parameter tracing.

[0031] In one or more embodiments, the layout detection model may employ the Transformer layout analysis model, or a combination of convolutional neural networks and traditional document analysis algorithms; the text recognition model may employ a general OCR engine in conjunction with a dictionary for wind power applications for correction.

[0032] In the post-processing correction process based on the wind power professional symbol confusion table and domain language model, the wind power professional symbols are corrected, and the correction result with the highest score is selected; the correction score is expressed as follows: ; in, Indicates the candidate correction results; This represents the character confidence score obtained based on character obfuscation relationships; This represents the contextual matching score of the candidate results under the language model in the wind power field. This indicates the matching score between the candidate results and the wind power terminology dictionary; , , These are the weighting coefficients for each scoring item.

[0033] Step S102: Extract the layout features of each paragraph from the post-processed semantic region, and make a hierarchical judgment accordingly, and then reconstruct the document structure tree.

[0034] The layout features include at least font size, bolding, left indentation, first-line indentation, paragraph spacing, numbering style, and page number. Candidate titles are first-level filtered based on the numbering style. Paragraphs without numbers but meeting the title morphological characteristics are identified and supplemented with titles. Parent-child node relationships are established according to title levels, and body text, tables, formulas, and figure captions located between adjacent titles are assigned to the nearest parent title node. When broken, missing, or scanned numbers are detected, hierarchical completion is performed using the semantic continuity of preceding and following paragraphs and the continuity of the layout.

[0035] A hierarchy determination function is used for hierarchy determination. The hierarchy determination function is as follows: ; in, Represents the i-th paragraph This indicates the font and layout style characteristics of the paragraph; This indicates the indentation characteristics of the paragraph; This indicates the numbering pattern characteristics of the paragraph; Indicate the key features of this paragraph; This represents the hierarchical decision function; Paragraph The corresponding hierarchical category.

[0036] In some alternative embodiments, during the structure tree reconstruction stage, hierarchy identification can employ either a rule engine based on font, indentation, and numbering rules, a supervised learning classification model, or a hybrid decision method combining both.

[0037] Step S103: Based on the reconstructed document structure tree, set the segmentation priority sequence as chapter title - section title - clause title - blank line - newline - period - space, and segment the knowledge block according to the knowledge block length and the preset maximum threshold.

[0038] like Figure 4As shown, if the length of the current knowledge block does not exceed the maximum threshold Lmax, it is output directly; if the length of the current knowledge block exceeds the threshold Lmax, it is divided into sub-blocks according to the current highest priority anchor point, and the same operation is recursively performed on the sub-blocks; if there are still no available splitting positions under all priorities, it is forcibly split according to the number of characters; an overlap window of 10% to 20% of the maximum threshold Lmax is reserved between adjacent knowledge blocks to ensure the semantic continuity of the boundary, and the proportion of complete sentences in the block is not less than 95% as the semantic integrity control indicator.

[0039] In the knowledge block segmentation stage, in addition to the title boundaries of chapters, sections, clauses, etc., the segmentation anchor points can be dynamically determined based on table titles, figure captions, formula numbers, or semantic similarity; the maximum block length and overlap ratio can be adjusted according to different application scenarios.

[0040] Step S104: Perform named entity recognition on each segmented knowledge block, extract wind power-related entities and establish a global entity index. Establish association edges based on the semantic similarity between any two knowledge blocks as weights, forming a wind power knowledge association graph for parameter tracing and joint retrieval. For example... Figure 5 As shown.

[0041] Named entity recognition is performed on each text block to extract entities such as wind measurement towers, wind speed data, wind turbines, coordinates, formulas, drawings, and code clauses, and a global entity index is established. For any two knowledge units, the semantic similarity of their context encoding vectors is calculated, and the association strength Aij = sim(hi,hj) × ωtype(typei,typej) is obtained by combining it with type compatibility weights. When the association strength exceeds a threshold, an association edge is established between the two knowledge units, thereby constructing a wind power knowledge association graph for parameter tracing and joint retrieval.

[0042] It should be noted here that the association strength can be determined by the product of semantic similarity and type weight, or further by introducing chapter distance, co-occurrence frequency, citation relationship or manual rules as additional constraints.

[0043] This invention focuses on establishing a parameter transfer chain from "wind measurement data" to "power generation estimation." To avoid establishing loose associations solely based on text similarity, this invention employs a joint design for wind power parameters, including parameter role identification, formula slot binding, and source evidence anchor point recording. For example... Figure 6As shown, the specific process includes: extracting candidate parameters from text blocks, table blocks, drawing blocks, and formula blocks during the wind power knowledge association graph process; performing entity recognition on candidate parameters based on parameter name, unit, context keywords, and chapter, classifying them into measured parameters, intermediate fitting parameters, height extrapolation parameters, equipment parameters, correction parameters, and conclusion parameters; establishing a parameter template for each parameter slot in the formula and matching candidate parameters with the formula slots; when a candidate parameter simultaneously satisfies the conditions of name matching, unit matching, chapter semantic matching, and type matching, binding the corresponding candidate parameter as the source parameter of the corresponding formula slot; and constructing directed edges for parameter transmission according to the formula dependency order.

[0044] Weibull distribution function: ; in, Indicates wind speed The corresponding probability density function, This represents the shape parameter of the Weibull distribution. This represents the scale parameter of the Weibull distribution.

[0045] Wind speed height extrapolation formula: ; in, Indicates wheel hub height Wind speed at the location, Indicates the wind reference height The measured wind speed at the location This represents the wind shear index.

[0046] Formula for estimating annual power generation: ; in, This represents the estimated annual power generation. Indicates the number of wind turbines. Represents the total number of hours worked per year. This represents the power curve of the wind turbine. This represents the wake correction factor. This indicates the availability of equipment.

[0047] The specific process of automatically associating the number of wind turbines, power curves, Weibull distribution, wake correction factor, and equipment availability with their respective source chapters or drawings is as follows: For the number of wind turbines, priority is given to searching in the chapters on unit selection schemes, installed capacity, and project overview, as well as the equipment configuration table; for the power curves, priority is given to searching in the wind turbine model technical parameter table, manufacturer sample diagrams, power curve diagrams, and their annotations; for the Weibull distribution parameters k and c, priority is given to searching in the chapters on wind energy resource analysis, wind measurement data statistics, and wind resource assessment results, as well as the corresponding statistical tables; for the wake correction factor, priority is given to searching in the chapters on wind farm layout optimization, wake analysis, and power generation correction; for the equipment availability, priority is given to searching in the chapters on equipment technical parameters, operation and maintenance assumptions, and power generation calculation boundary conditions, as well as related tables; after completing the parameter binding, source evidence anchor points are output for each parameter.

[0048] Step S105: Jointly encode the neighbor contexts of each knowledge block and its wind power knowledge association graph to obtain context-enhanced semantic vectors. Synchronously write these vectors, along with the global entity index, entity type, associated edges and their weights, as metadata into the vector database for parameter tracing.

[0049] The vector index preferably uses the HNSW approximate nearest neighbor index, but a multi-granularity index structure at the chapter and block levels can also be established. During retrieval, while returning the Top-K directly similar knowledge blocks, its associated subgraph neighbors are expanded to achieve cross-chapter and cross-modal linkage retrieval.

[0050] It is understood here that, in addition to HNSW, vector indexes can also use IVF, PQ, or other approximate nearest neighbor index structures; the association graph can be stored as vector metadata or stored separately in the graph database and invoked in conjunction with it during retrieval.

[0051] like Figure 7 As shown, the semantic embedding and vector database construction has the following specific design steps: Generate a basic semantic vector for each knowledge block; read the neighboring nodes of the knowledge block in the wind power knowledge association graph, and use their upstream parameter nodes, downstream result nodes, and auxiliary nodes of the same chapter as context enhancement information; jointly encode the basic semantic vector and the context enhancement information to obtain a context-enhanced semantic vector; synchronously write the semantic vector and metadata into the vector database, whereby the metadata includes at least chapter path, parameter role, source page number, region coordinates, formula number, figure number, table number, and associated edge weight; and establish a multi-granularity index according to chapter-level index, block-level index, and parameter-level index.

[0052] In a preferred embodiment, knowledge blocks The context-enhanced representation can be expressed as:

[0053] in, Representing knowledge blocks Context-enhanced semantic vectors Indicates the joint encoding function, This indicates the content of the current knowledge block. Representing knowledge blocks In the neighbor context of the association graph This represents the set of metadata associated with this knowledge block.

[0054] For Chapter 3, "Wind Energy Resource Analysis," of a 200MW wind farm project proposal, OCR was used to analyze and identify the coordinates of the meteorological towers and their corresponding annual average wind speed, standard deviation, and Weibull parameters k and c. Subsequently, the data was recursively divided into blocks according to "Chapter 3 → Section 3.1 → Article 3.1.1," and the wind speed statistics table, the meteorological tower location map, and the power generation formula block were linked together.

[0055] In economic analysis scenarios, the system can further establish a parameter traceability chain of "annual power generation → grid-connected power → annual revenue", enabling financial conclusions to be traced back to the source of wind resources and equipment parameters.

[0056] This invention improves the accuracy of structured analysis of tables, drawings, formulas and technical terms in wind power engineering documents by combining layout area detection with wind power professional symbol correction. In addition to being applicable to wind power project proposals, this invention can also be extended to wind resource assessment reports, feasibility study reports, EPC technical documents, and other new energy documents with multimodal engineering knowledge structures.

[0057] This invention reduces semantic truncation caused by fixed-length segmentation by using a recursive text block approach with chapters, sections, and clauses as priority anchors, thus maintaining the hierarchical logic and contextual integrity of knowledge blocks. By establishing parameter transfer chains between wind speed data, wind measurement height, hub height, turbine power curves, wake correction coefficients, and equipment availability, this invention enables chapter-level, chart-level, and formula-level traceability of key calculation parameters in wind power project proposals. Through contextual enhancement embedding of knowledge blocks and linked retrieval using parameter transfer chains, this invention improves the accuracy and completeness of wind power professional parameter queries, source tracing, and result interpretation. This invention has good engineering adaptability and can be extended to the construction of knowledge bases for various wind power engineering documents, including wind resource assessment, turbine selection, power generation calculation, and economic analysis.

[0058] like Figure 2 As shown, the hierarchical knowledge construction and parameter tracing device for project proposals provided in this embodiment of the invention can be implemented in software. The hierarchical knowledge construction and parameter tracing device for project proposals includes the following software modules: Page area detection module 201 is used to perform page area detection on the project proposal page, obtain semantic regions and their location coordinates and confidence scores, and perform post-processing correction based on the wind power professional symbol confusion table and domain language model; The document structure tree reconstruction module 202 is used to extract the layout features of each paragraph from the post-processed corrected semantic region, and make hierarchical judgments accordingly, and then reconstruct the document structure tree. The knowledge block segmentation module 203 is used to segment knowledge blocks based on the reconstructed document structure tree, setting the segmentation priority sequence as chapter title-section title-clause title-blank line-newline-period-space, and segmenting knowledge blocks according to the knowledge block length and the preset maximum threshold. The wind power knowledge association graph construction module 204 is used to perform named entity recognition on each knowledge block obtained by segmentation, extract wind power related entities and establish a global entity index, establish association edges based on the semantic similarity between any two knowledge blocks as weights, and form a wind power knowledge association graph for parameter tracing and joint retrieval. The vector database construction module 205 is used to jointly encode the neighbor contexts of each knowledge block and its wind power knowledge association graph to obtain context-enhanced semantic vectors. These vectors, along with the global entity index, entity type, associated edges and their weights, are synchronously written into the vector database as metadata for parameter tracing.

[0059] It should be noted that each module in the hierarchical knowledge construction and parameter tracing device for project proposals in this embodiment corresponds one-to-one with each step in the hierarchical knowledge construction and parameter tracing method for project proposals in the above embodiments, and their specific implementation processes are the same, so they will not be repeated here.

[0060] The structure of the electronic device according to an embodiment of the present invention will be described in detail below. Figure 3 This is a schematic diagram of the composition structure of an electronic device provided in an embodiment of the present invention. It can be understood that... Figure 3 The diagram shows only an exemplary structure of the electronic device, not the entire structure. Some or all of the structures shown may be implemented as needed.

[0061] The electronic device provided in this embodiment of the invention includes: at least one processor 301, a memory 302, a user interface 303, and at least one network interface 304. The various components in the hierarchical knowledge construction and parameter tracing device of the project proposal are coupled together through a bus system 305. It can be understood that the bus system 305 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 305 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general designated all buses as Bus System 305.

[0062] The user interface 303 may include a monitor, keyboard, mouse, trackball, click wheel, buttons, touchpad, or touch screen.

[0063] It is understood that memory 302 can be volatile memory or non-volatile memory, or both. In this embodiment of the invention, memory 302 is capable of storing data to support the operation of the terminal. Examples of this data include any computer programs used to operate on the terminal, such as operating systems and applications. The operating system includes various system programs, such as framework layers, core library layers, driver layers, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0064] In some embodiments, the hierarchical knowledge construction and parameter tracing device for project proposals provided in this invention can be implemented using a combination of hardware and software. For example, the hierarchical knowledge construction and parameter tracing device for project proposals provided in this invention can be a processor in the form of a hardware decoding processor, programmed to execute the hierarchical knowledge construction and parameter tracing method for project proposals provided in this invention. For instance, the processor in the form of a hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0065] As an example, processor 301 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0066] As an example of the hardware implementation of the hierarchical knowledge construction and parameter tracing device for project proposals provided in this embodiment of the invention, the device provided in this embodiment of the invention can be directly executed by a processor 301 in the form of a hardware decoding processor. For example, it can be executed by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the hierarchical knowledge construction and parameter tracing method for project proposals provided in this embodiment of the invention.

[0067] In this embodiment of the invention, the memory 302 is used to store various types of data to support the operation of the hierarchical knowledge construction and parameter tracing device for project proposals, or to store data for execution. Figure 1 The program code for the method shown. Examples of this data include: any executable instructions for operation on the hierarchical knowledge construction and parameter tracing device for project proposals, such as executable instructions, and programs implementing the hierarchical knowledge construction and parameter tracing method for project proposals according to embodiments of the present invention may be included in the executable instructions.

[0068] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 1 The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit, it performs the various functions defined in the apparatus of this application.

[0069] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0070] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for hierarchical knowledge construction and parameter tracing in project proposals, characterized in that, include: The layout area of ​​the project proposal page is detected to obtain the semantic area, its location coordinates and confidence level, and post-processing correction is performed based on the wind power professional symbol confusion table and domain language model; The layout features of each paragraph are extracted from the semantic region corrected by post-processing, and hierarchical judgment is made accordingly, thereby reconstructing the document structure tree; Based on the reconstructed document structure tree, the segmentation priority sequence is set as chapter title - section title - clause title - blank line - newline - period - space, and knowledge blocks are segmented according to the knowledge block length and the preset maximum threshold. Named entity recognition is performed on each knowledge block obtained by segmentation, wind power-related entities are extracted and a global entity index is established, and association edges are established based on the semantic similarity between any two knowledge blocks as weights, forming a wind power knowledge association graph for parameter tracing and joint retrieval. The context of each knowledge block and its neighboring context in the wind power knowledge association graph are jointly encoded to obtain a context-enhanced semantic vector. This vector, along with the global entity index, entity type, associated edges and their weights, is synchronously written into the vector database as metadata for parameter tracing.

2. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, The semantic regions include text areas, table areas, drawing areas, formula areas, and figure annotation areas.

3. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, In the post-processing correction process based on the wind power professional symbol confusion table and domain language model, the wind power professional symbols are corrected and the correction result with the highest score is selected. Its corrected score is expressed as follows: ； in, Indicates the candidate correction results; This represents the character confidence score obtained based on character obfuscation relationships; This represents the contextual matching score of the candidate results under the language model in the wind power field. This indicates the matching score between the candidate results and the wind power terminology dictionary; , , These are the weighting coefficients for each scoring item.

4. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, The layout features include at least font size, bolding, left indentation, first-line indentation, paragraph spacing, numbering style, and page number. Candidate titles are first-level filtered based on the numbering style. Paragraphs without numbers but meeting the title morphological characteristics are identified and supplemented with titles. Parent-child node relationships are established according to title levels, and body text, tables, formulas, and figure captions located between adjacent titles are assigned to the nearest parent title node. When broken, missing, or scanned numbers are detected, hierarchical completion is performed using the semantic continuity of preceding and following paragraphs and the continuity of the layout.

5. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, A hierarchy determination function is used for hierarchy determination. The hierarchy determination function is as follows: ； in, Represents the i-th paragraph This indicates the font and layout style characteristics of the paragraph; This indicates the indentation characteristics of the paragraph; This indicates the numbering pattern characteristics of the paragraph; Indicate the key features of this paragraph; This represents the hierarchical decision function; Paragraph The corresponding hierarchical category.

6. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, If the length of the current knowledge block does not exceed the maximum threshold Lmax, it is output directly; if the length of the current knowledge block exceeds the threshold Lmax, it is divided into sub-blocks according to the current highest priority anchor point, and the same operation is recursively performed on the sub-blocks; if there are still no available splitting positions under all priorities, it is forcibly split according to the number of characters; an overlap window of 10% to 20% of the maximum threshold Lmax is reserved between adjacent knowledge blocks to ensure the semantic continuity of the boundary, and the proportion of complete sentences in the block is not less than 95% as the semantic integrity control indicator.

7. The hierarchical knowledge construction and parameter tracing method for project proposals as described in claim 1, characterized in that, In the process of creating a wind power knowledge association graph, candidate parameters are extracted from text blocks, table blocks, drawing blocks, and formula blocks. Based on the parameter name, unit, context keywords, and chapter, the candidate parameters are identified as measured parameters, intermediate fitting parameters, height extrapolation parameters, equipment parameters, correction parameters, and conclusion parameters. A parameter template is created for each parameter slot in the formula, and candidate parameters are matched with the formula slots. When a candidate parameter simultaneously meets the conditions of name matching, unit matching, chapter semantic matching, and type matching, the corresponding candidate parameter is bound as the source parameter of the corresponding formula slot. Directed edges for parameter transmission are constructed according to the formula dependency order.

8. A device for hierarchical knowledge construction and parameter tracing in project proposals, characterized in that, The hierarchical knowledge construction and parameter tracing method based on the project proposal as described in any one of claims 1-7 includes: The page layout area detection module is used to detect the layout area of ​​the project proposal page, obtain the semantic area and its location coordinates and confidence, and perform post-processing correction based on the wind power professional symbol confusion table and domain language model. The document structure tree reconstruction module is used to extract the layout features of each paragraph from the post-processed semantic region, and to make hierarchical judgments accordingly, thereby reconstructing the document structure tree. The knowledge block segmentation module is used to segment knowledge blocks based on the reconstructed document structure tree. The segmentation priority sequence is set as chapter title - section title - clause title - blank line - newline - period - space. The knowledge block is segmented according to the knowledge block length and the preset maximum threshold. The wind power knowledge association graph construction module is used to perform named entity recognition on each knowledge block obtained by segmentation, extract wind power related entities and establish a global entity index, establish association edges based on the semantic similarity between any two knowledge blocks as weights, and form a wind power knowledge association graph for parameter tracing and joint retrieval. The vector database construction module is used to jointly encode the neighbor contexts of each knowledge block and its wind power knowledge association graph to obtain context-enhanced semantic vectors. These vectors, along with the global entity index, entity type, associated edges and their weights, are synchronously written into the vector database as metadata for parameter tracing.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the hierarchical knowledge construction and parameter tracing method for project proposals as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps in the hierarchical knowledge construction and parameter tracing method for project proposals as described in any one of claims 1-7.

Images