Plug-in-based bidding document generation method and device, equipment and medium

Abstract

The invention relates to the technical field of natural language processing, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses a bidding document generation method, device and equipment based on a plug-in and a medium. Obtaining bidding and tendering file metadata; constructing a bidding document plug-in according to the bidding and tendering file metadata and the plug-in configuration information; extracting a demand description statement of the bidding and tendering file, and performing chapter generation on the demand description statement according to the bidding document plug-in to obtain a chapter bidding document text; extracting chapter feature vectors of the chapter bidding document text, and performing multi-level similarity matching in a preset document material database according to the chapter feature vectors to obtain corresponding text content; and performing text fusion on the chapter bidding document text and the text content to obtain the target bidding document. The bidding document generation quality and efficiency in the bidding and tendering project can be improved.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and in particular to a plugin-based method, apparatus, device, and medium for generating tender documents. Background Technology

[0002] With the standardization of the bidding market, the quality and efficiency of bidding document preparation have become key factors for enterprises to participate in market competition. However, traditional bidding document processing is still dominated by manual writing. Manual writing requires repeated review of bidding project documents and historical bidding materials, which is not only difficult and time-consuming to integrate data, but also prone to rework due to information omissions. At the same time, most existing bidding document processing tools only support basic format editing and lack the ability to understand the semantics of bidding documents and automatically generate them, making it impossible to achieve full-process intelligent processing from requirements analysis to content generation.

[0003] For example, in the healthcare sector, bidding projects often involve the procurement of large medical equipment and the construction of smart hospital information systems. Currently, when drafting bid documents manually, staff need to repeatedly compare data from a large number of medical industry standards, clinical operating guidelines, and historical project materials. However, due to a lack of thorough understanding of implicit requirements such as "adaptability to clinical diagnosis and treatment processes," the descriptions of equipment functions in the bid documents often fail to meet actual clinical needs, leading to inaccurate bid documents and low-quality standard drafting.

[0004] In the fintech context, bidding projects often revolve around upgrading financial systems and building intelligent risk control platforms. When manually drafting these bids, it's crucial to sift through complex financial regulatory documents and operational guidelines to identify key requirements. However, a lack of understanding of the synergy between regulatory policies and business scenarios frequently leads to mismatches between compliance statements in the bid and actual business needs. Furthermore, existing bid processing tools often only support basic format editing and lack format conversion capabilities, resulting in unclear and unreadable bid content.

[0005] Therefore, improving the quality and efficiency of bid document generation in bidding projects has become an urgent problem to be solved. Summary of the Invention

[0006] This invention provides a plug-in-based method, apparatus, device, and medium for generating tender documents, the main purpose of which is to solve the problems of low quality and low efficiency in generating tender documents in bidding projects.

[0007] Firstly, to achieve the above objectives, the present invention provides a plugin-based tender document generation method, comprising: Obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata; The tender document plugin for the tender project is constructed based on the metadata of multiple tender documents and the plugin configuration information. Extract the requirement description statements from the bidding documents, and generate chapters from the requirement description statements according to the bidding document plugin to obtain multiple chapter bidding document texts; Extract chapter feature vectors from multiple chapter tender texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector; The text of the tender document and the text content of the aforementioned chapter are merged to obtain the target tender document for the tender project.

[0008] Secondly, the present invention also provides a plug-in-based tender document generation device, comprising: The document semantic parsing module is used to obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata. The tender document plugin configuration module is used to construct the tender document plugin for the bidding project based on multiple tender document metadata and the plugin configuration information. The chapter text generation module is used to extract the requirement description statements in the bidding documents, and generate chapters based on the requirement description statements in the bidding document plugin to obtain multiple chapter bidding document texts. The tender document content matching module is used to extract chapter feature vectors of multiple chapter tender document texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector; The target tender document generation module is used to perform text fusion on the chapter tender document text and the text content to obtain the target tender document for the bidding project.

[0009] Thirdly, the present invention also provides an electronic device, the electronic device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the plug-in-based tender generation method described above.

[0010] Fourthly, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the plug-in-based tender generation method described above.

[0011] In this embodiment of the invention, semantic parsing technology can deeply understand the text content in bidding documents, transforming unstructured text into structured metadata, making the information more standardized and easier to manage, facilitating subsequent data storage and retrieval, and improving the accuracy and consistency of information processing. By accurately extracting and integrating the metadata of bidding documents, plugins tailored to the needs of bidding projects can be quickly generated, greatly improving the efficiency of bid preparation. Utilizing the plugin configuration information customization function can enhance the flexibility and applicability of the plugin, meeting different bidding scenarios and providing stable and reliable technical support for bidding projects. When extracting the requirement description statements in bidding documents, natural language processing technology can accurately identify and separate key requirement information, avoiding omissions or errors that may occur during manual extraction.

[0012] Specifically, by using the tender document plugin to generate chapters from the requirement description statements, requirements can be quickly and reasonably allocated to different chapters. This not only significantly shortens the tender document production time but also ensures the standardization and consistency of chapter division, while enhancing the relevance and professionalism of the tender document. Extracting chapter feature vectors accurately captures the core semantics and key information of the chapter text, and multi-level similarity matching quickly and accurately locates text content that highly matches the feature vectors of each chapter, effectively improving the quality and professionalism of the tender document. Precise structural analysis of the chapter tender document text and the matched text content clearly grasps the semantic logic and framework hierarchy of both, ensuring a smooth integration of content and greatly improving the efficiency and quality of tender document generation, providing strong support for the smooth progress of bidding projects. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of an application environment for a plug-in-based tender document generation method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a plugin-based tender document generation method according to an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the process of constructing a tender document plugin for a tendering project based on metadata and plugin configuration information of multiple tender documents, as provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of a plug-in-based tender document generation device according to an embodiment of the present invention; Figure 5A schematic diagram of the structure of an electronic device that implements a plug-in-based tender document generation method according to an embodiment of the present invention; Figure 6 This is another structural schematic diagram of an electronic device that implements a plug-in-based tender document generation method according to an embodiment of the present invention.

[0015] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] This application provides a plug-in-based tender document generation method. The execution entity of this plug-in-based tender document generation method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the device provided in this application: a server, a terminal, etc. In other words, the plug-in-based tender document generation method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0019] This invention discloses a plugin-based tender document generation method, which can be applied to applications such as... Figure 1 In this application environment, the client communicates with the server via a network. The server can obtain the bidding documents and plugin configuration information of the bidding project from the client and use semantic parsing technology to deeply understand the text content in the bidding documents, transforming unstructured text into structured metadata. This makes the information more standardized and easier to manage, facilitating subsequent data storage and retrieval, and improving the accuracy and consistency of information processing. By accurately extracting and integrating the metadata of the bidding documents, plugins tailored to the needs of the bidding project can be quickly generated, greatly improving the efficiency of bid preparation. The plugin configuration information customization function enhances the flexibility and applicability of the plugins, meeting different bidding scenarios and providing stable and reliable technical support for bidding projects.

[0020] Meanwhile, when extracting the requirement description statements from the bidding documents, natural language processing technology can accurately identify and separate key requirement information, avoiding omissions or errors that may occur during manual extraction. Based on the bidding document plugin, the requirement description statements are generated into chapters, allowing for the rapid and reasonable allocation of requirements to different chapters. This not only significantly shortens the bidding document production time but also ensures the standardization and consistency of chapter division, while enhancing the relevance and professionalism of the bidding documents. Extracting chapter feature vectors accurately captures the core semantics and key information of the chapter text, while multi-level similarity matching quickly and accurately locates text content that highly matches the feature vectors of each chapter, effectively improving the quality and professionalism of the bidding documents. Precise structural analysis of the chapter bidding text and the matched text content clearly grasps the semantic logic and framework hierarchy of both, ensuring a smooth integration of content and greatly improving the efficiency and quality of bidding document generation. Finally, the target bidding document is output and fed back to the client.

[0021] The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0022] Reference Figure 2 The diagram shown illustrates a flowchart of a plugin-based tender document generation method according to an embodiment of the present invention. In this embodiment, the plugin-based tender document generation method includes: S1. Obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata.

[0023] In this embodiment of the invention, the bidding documents refer to the formal documents issued by the bidding party (such as enterprises or government agencies) to potential bidders (suppliers, contractors, etc.) when conducting procurement, project outsourcing and other transaction activities. The bidding documents list in detail the project requirements, evaluation criteria, contract terms and other information.

[0024] The plugin configuration information refers to the various parameters and settings required by the tender document plugin during runtime. Plugin configuration information typically includes the plugin name, version number, function description, dependencies, etc.

[0025] Specifically, programming languages ​​such as Python and Java can be used to connect to a pre-defined bidding website using the HTTP protocol and send HTTP requests to obtain bidding documents from the bidding website.

[0026] In this embodiment of the invention, the semantic parsing of the bidding documents to obtain multiple bidding document metadata includes: The bidding documents are standardized to obtain standardized bidding documents. Extract the standardized text from the standardized bidding documents, and perform word segmentation on the standardized text to obtain multiple bidding text word segments; Semantic annotation is performed on multiple segments of the bidding text to obtain annotated bidding segments; Based on the labeled bidding and tendering word segmentation, the standardized text is context-associated to obtain contextual semantic information; The core metadata items in the labeled bidding and tendering word segment are extracted based on the contextual semantic information, and the core metadata items are stored in a structured manner to generate bidding and tendering document metadata.

[0027] In the embodiments of the present invention, the original format of the bidding documents is recognized. The original formats include format types such as PDF, Word, Excel, etc. After clarifying the original format of the bidding documents, a conversion tool or conversion library adapted thereto will be selected. For example, Apache PDFBox can be used to process PDF files, and Apache POI is suitable for Word and Excel files. Using the conversion tool or conversion library, the bidding documents are uniformly converted into an intermediate format, such as plain text or XML format, and then a standardized bidding document is obtained, ensuring the consistency and compatibility of subsequent document processing.

[0028] Specifically, by parsing the standardized bidding document to read the standardized text, text preprocessing is performed on the extracted standardized text, such as removing irrelevant characters such as spaces and line breaks, and performing operations such as unifying case.

[0029] Furthermore, a word segmentation algorithm (such as dictionary-based word segmentation, statistics-based word segmentation, etc.) is used to segment the standardized text into multiple meaningful words, that is, word segmentation of the bidding text. During the word segmentation process, some common stop words (such as "of", "is", "in", etc.) are filtered out, which can reduce the amount of data for subsequent processing.

[0030] Specifically, a pre-trained semantic annotation model (such as a sequence annotation model based on deep learning) is used to perform semantic annotation on the word segmentation of the bidding text. The semantic annotation model can usually identify the specific meaning and role of the word segmentation of the bidding text in the bidding document (such as project name, bidder, quotation, etc.). During the semantic annotation process, according to the predefined semantic annotation rules (such as BIO annotation rules), corresponding semantic labels are assigned to each word segmentation of the bidding text. The semantic labels reflect the semantic category and position information of the word segmentation of the bidding text in the bidding document.

[0031] Among them, the context analysis of the standardized text is carried out by using the semantic label information in the labeled bidding word segmentation, including identifying the key paragraphs in the bidding document and the logical relationship between the key paragraphs. On the basis of the context analysis, the entity relationships in the bidding document (such as the relationship between the project and the bidder, the relationship between the quotation and the bidder, etc.) are further extracted, which helps to understand the content of the bidding document.

[0032] Specifically, the core metadata items (such as project name, bidder name, quotation amount, etc.) in the bidding document are identified according to the key information and entity relationships in the context semantic information, and the identified core metadata items are stored in a structured manner, which can be achieved by defining corresponding data structures (such as JSON objects, XML elements, etc.), and then the metadata of the bidding document is generated, that is, a structured data set containing all the core metadata items in the bidding document.

[0033] In this embodiment of the invention, semantic parsing technology can deeply understand the text content in bidding documents, transforming unstructured bidding documents into structured bidding document metadata, thereby improving the accuracy and consistency of bidding document processing.

[0034] S2. Construct the tender document plugin for the tender project based on the metadata of the multiple tender documents and the plugin configuration information.

[0035] In this embodiment of the invention, the tender document plugin for bidding projects is a software functional module specifically designed for the bidding process. Its core function is to automatically assist in generating target tender documents based on the tender documents.

[0036] The tender document plugin integrates specific business logic and data interface specifications to read and process metadata of tender documents (such as tenderer information, technical specifications, commercial terms, qualification requirements, etc.), and fills it into preset tender document chapter templates, thereby generating tender document content that conforms to tender specifications, thus improving the accuracy, efficiency and compliance of tender document preparation.

[0037] For example, in a healthcare scenario, a medical technology company bids for a county-level medical data center platform construction project. The metadata extracted from the bidding documents includes functional requirements such as "the need to build four major databases, including a population information database and a health record database" and "support for data integration with the maternal and child health care system." The plugin configuration information covers parameters of the data quality control plugin (such as support for setting group sequence numbers and configuring job execution frequency) and interface specifications of the regional BI data analysis plugin. Based on the metadata of multiple bidding documents and the plugin configuration information, the bidding document plugin for the bidding project is constructed.

[0038] In this embodiment of the invention, constructing the tender document plugin for the bidding project based on multiple tender document metadata and the plugin configuration information includes: Extract the plugin function elements from the plugin configuration information, and filter the bidding document metadata according to the plugin function elements to obtain the target metadata corresponding to the plugin function elements; The target metadata is converted into structured standard data that is readable by the plugin according to the preset data interface specification; The structured standard data is integrated with the preset plugin framework according to the format display requirements in the plugin configuration information to generate the initial plugin; The initial plugin is functionally verified and compatibility tested to generate the tender document plugin for the bidding project.

[0039] In this embodiment of the invention, a preset parsing library or parsing tool is used to parse the plugin configuration information to extract plugin functional elements. The plugin functional elements include data processing rules and user interaction methods of the tender document plugin. The extracted plugin functional elements are matched with the bidding document metadata by comparing the bidding document metadata with the requirements in the functional elements. The target metadata corresponding to the plugin functional elements is then selected based on the matching results.

[0040] In detail, the target metadata is transformed according to the preset data interface specification. The data interface specification defines how the target metadata should be organized and formatted so that the tender document plugin can read and process it. The data interface specification includes data type conversion, data structure reorganization, etc., to obtain structured standard data, that is, a data set that conforms to the format readable by the tender document plugin.

[0041] Specifically, structured standard data is integrated with a pre-defined plugin framework, either by writing code or using an integration tool. During the integration process, it is ensured that the structured standard data can be correctly mapped to the corresponding parts of the plugin framework. The initial integrated plugin is then formatted according to the format display requirements in the plugin configuration information, including setting visual elements such as fonts, colors, and layouts, as well as defining user interaction methods.

[0042] Furthermore, the initial plugin is functionally verified by executing a series of preset test cases, which include various functions and scenarios of the initial plugin. In addition to functional verification, the initial plugin is compatibility tested, including testing the initial plugin's operation on different operating systems, browsers, or devices, to ensure that the initial plugin can work normally in various environments, thereby generating the tender document plugin for bidding projects.

[0043] In this embodiment of the invention, by accurately extracting and integrating the metadata of bidding documents, a plugin that fits the needs of bidding projects can be quickly generated, greatly improving the efficiency of bid preparation, reducing the time and errors of manual operation, and enhancing the flexibility and applicability of the plugin by utilizing the plugin configuration information customization function, meeting different bidding scenarios, providing stable and reliable technical support for bidding projects, and promoting the intelligent development of bidding work.

[0044] S3. Extract the requirement description statements from the bidding documents, and generate chapters from the requirement description statements according to the bidding document plugin to obtain multiple chapter bidding document texts.

[0045] In this embodiment of the invention, text cleaning technology is used to remove irrelevant information such as special symbols, extra spaces, and line breaks from the bidding documents. Then, the text-cleaned bidding documents are segmented into sentences. Based on common Chinese punctuation marks, such as periods, question marks, and exclamation marks, the bidding documents are divided into independent sentences. Then, keyword matching is performed on each sentence. With the help of a pre-built keyword library in the bidding field, words such as "requirements," "requirements," and "specifications" are searched and matched in the segmented sentences to quickly identify sentences that may contain requirement descriptions, thereby accurately extracting the requirement description sentences in the bidding documents.

[0046] In this embodiment of the invention, the step of generating multiple chapters of the requirement description statement based on the tender document plugin to obtain multiple chapter tender document texts includes: The requirement description statement is cleaned to obtain the target requirement statement; The target requirement statements are classified to obtain the target tender document type corresponding to the target requirement statements; The target tender document type is matched using the tender document plugin to obtain the target chapter template. Semantically match the key requirement elements in the target requirement statement with the chapter fields of the target chapter template to establish a key chapter field correspondence table; Based on the key chapter field correspondence table, the key requirement elements are filled into the corresponding chapter fields of the target chapter template to generate initial chapter content; The initial chapter content is checked for coherence, resulting in multiple chapters of the tender document text.

[0047] In this embodiment of the invention, the requirement description statement is cleaned by removing special characters, such as symbols that cannot be displayed correctly or are meaningless. Then, redundant spaces and line breaks are processed to make the requirement description statement uniform and standardized, avoid irrelevant factors from interfering with subsequent processing, and finally obtain the target requirement statement.

[0048] In detail, the target requirement statements are classified, and machine learning algorithms are used to train a model on a large number of labeled requirement description statements. The model learns the correspondence between different requirement description statements and tender types, thereby accurately obtaining the target tender type corresponding to the target requirement statement.

[0049] Furthermore, a tender document plugin is used to perform template matching for the target tender document type. Leveraging a pre-built template library within the plugin, which stores numerous chapter templates categorized by different tender document types, the plugin analyzes the characteristics of the target tender document type, such as key business requirements and formatting requirements, to quickly retrieve and match the most suitable target chapter template from the template library. Semantic matching is then performed between the key requirement elements in the target requirement statement and the chapter fields of the target chapter template. Using semantic analysis technology, the semantic information of the key requirement elements and chapter fields is analyzed to generate a table showing the correspondence between key chapter fields in the target requirement statement and the target chapter template.

[0050] Specifically, based on the correspondence in the key chapter field correspondence table, each key requirement element is accurately placed into the corresponding chapter field position to generate initial chapter content; the initial chapter content is then checked for coherence, that is, the logical relationship and semantic connection between the initial chapter content are checked to see if they are reasonable, and whether there are semantic conflicts or jumps, so as to ensure that the initial chapter content is logically coherent and smooth, thereby obtaining multiple chapter tender documents.

[0051] For example, in a fintech scenario, a fintech company participates in a bidding process for a bank's core system upgrade project. Using text cleaning technology, it parses the obtained bidding documents to accurately extract requirement descriptions, such as "the new system needs to support a transaction processing capacity of over 50,000 transactions per second and ensure stable operation 24 / 7." Then, based on a pre-built bidding document plugin containing various standard solutions and case templates for the financial industry, it automatically matches the corresponding chapter template in the bidding document plugin according to the key features of the requirement descriptions.

[0052] For example, for the requirement of transaction processing capability, the "System Performance Architecture" chapter template is matched; for the requirement of risk warning function, the "Risk Management Module" chapter template is matched; then, the requirement description statement is filled into the corresponding template, and it is expanded and improved by combining the technical parameters, implementation process and other contents in the tender document plugin, and finally generating a tender document text with multiple chapters.

[0053] In this embodiment of the invention, when extracting the requirement description statements from bidding documents, natural language processing technology can be used to accurately identify and separate key requirement information, avoiding omissions or errors that may occur during manual extraction, and greatly improving the accuracy and completeness of information extraction. By generating chapters for the requirement description statements based on the bidding document plugin, the requirement description statements can be quickly and reasonably allocated to different chapters, which not only significantly shortens the bidding document preparation time and improves work efficiency, but also ensures the standardization and consistency of chapter division, while enhancing the relevance and professionalism of the bidding document.

[0054] S4. Extract the chapter feature vectors of multiple chapter tender texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector.

[0055] In this embodiment of the invention, the multi-level similarity matching refers to matching the chapter feature vector with the document material database according to the similarity calculation rules of different levels in a preset document material database, so as to obtain the text content corresponding to each chapter feature vector, thereby improving information retrieval efficiency and reducing the tediousness and error of manual screening.

[0056] In this embodiment of the invention, extracting the chapter feature vectors of the multiple chapter tender texts includes: Each chapter of the tender document is segmented to obtain multiple independent sentence units; The multiple independent sentence units are denoised and integrated to obtain a standardized sentence set. Each sentence in the standardized sentence set is transformed into a word vector to obtain a word vector sequence for each sentence; The word vector sequence of each sentence is weighted according to a predefined attention mechanism to obtain the sentence feature vector of each sentence; Aggregate and calculate the feature vectors of all the sentences to obtain the initial feature vector of the chapter title text; The initial feature vector is normalized to obtain the chapter feature vector.

[0057] In this embodiment of the invention, based on common Chinese sentence ending markers such as periods, question marks, and exclamation marks, the chapter title text is divided into independent sentence units according to semantic integrity. Optionally, noise reduction processing is performed on multiple independent sentence units, that is, irrelevant information in the independent sentence units is removed. Integrating multiple independent sentence units means merging sentences with similar or related meanings to avoid information dispersion.

[0058] For example, if two independent sentence units both revolve around the time schedule requirements of the bidding documents, but with slightly different expressions, they will be integrated into one sentence, thus obtaining a standardized sentence set.

[0059] Furthermore, each sentence in the standardized sentence set is broken down into individual words, and each word is mapped to a high-dimensional vector space. Vectors are used to represent the semantic information of words. For example, words such as "technology," "solution," and "implementation" are converted into vectors of a specific dimension. A sentence consists of multiple words, and arranging the vectors of multiple words in order forms the word vector sequence of the sentence. This allows the semantic information of the sentence to be expressed in vector form.

[0060] In detail, the importance of each word vector in the word vector sequence to the semantics of the sentence is determined based on the attention mechanism. Based on factors such as the word's position in the sentence, part of speech, and semantic association, a weight value is assigned to each word vector. The larger the weight, the greater the contribution of the word to the semantics of the sentence. Each word vector is multiplied by its corresponding weight value, and then the weighted word vectors are added together to obtain the sentence feature vector that can represent the core semantics of the entire sentence.

[0061] The aggregation calculation is used to integrate all sentence feature vectors and extract features that can represent the entire chapter of the tender document text. This can be done by summing, averaging, or other methods to average the corresponding dimension values ​​of all sentence feature vectors. For example, the first dimension values ​​of all sentence feature vectors are added together and then divided by the number of sentences to obtain the average vector, which is used as the value of the first dimension of the initial feature vector. Similar calculations are performed on each dimension in turn to obtain the initial feature vector of the chapter tender document text.

[0062] Specifically, vector normalization aims to eliminate the influence of differences in the dimensions and numerical ranges among the dimensions of the initial feature vector. It divides the value of each dimension of the initial feature vector by the magnitude of the initial feature vector, so that the magnitude of the normalized vector is a fixed value. This ensures that the values ​​of each dimension of the chapter feature vector are within a relatively uniform range, which facilitates subsequent similarity comparison and matching in the document material database, thereby improving the accuracy and stability of the matching.

[0063] In this embodiment of the invention, the step of performing multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector includes: Based on the feature vectors of the chapters, a first-level coarse matching is performed in a preset document material database to obtain a set of candidate materials that are initially related to the feature vectors of each chapter; For each material in the candidate material set, feature extraction is performed to generate a material feature vector; Calculate the cosine similarity between the feature vector of each chapter and the feature vector of the material; The candidate material set is sorted and filtered according to the cosine similarity to obtain a second filtered material set; Extract the text from the second set of filtered materials, and determine the text as the text content corresponding to the feature vector of the chapter.

[0064] In this embodiment of the invention, material tags for each material in the document material database are extracted, and coarse matching is performed between the material tags and the chapter feature vector to quickly filter out materials that are related to the chapter feature vector and form a candidate material set.

[0065] In detail, a method similar to that used for extracting chapter feature vectors is employed to perform text segmentation on each material in the candidate material set, breaking it down into independent sentence units. Each sentence unit is then transformed into a word vector sequence. Next, weights are assigned to the word vector sequence of each sentence based on a predefined attention mechanism to obtain sentence feature vectors. Finally, all sentence feature vectors are aggregated, calculated, and normalized to generate material feature vectors that represent the core features of the material, so as to perform more accurate cosine similarity comparisons with chapter feature vectors in the future.

[0066] Cosine similarity is an index that measures the difference in direction between two vectors. It treats the chapter feature vector and the material feature vector as two directed line segments in the vector space. The degree of similarity is judged by comparing the cosine value of the angle between the chapter feature vector and the material feature vector. The closer the cosine value, the more similar the chapter feature vector and the material feature vector are in direction, indicating that the chapter content represented by the chapter feature vector and the material content represented by the material feature vector are more similar in semantics.

[0067] Specifically, the materials in the candidate material set are sorted from high to low according to the calculated cosine similarity. The higher the similarity, the better the match between the material and the chapter content. Then, according to the preset screening threshold or screening quantity, the materials with higher similarity are selected from the sorted materials to form the second screening material set.

[0068] For example, if we set a filter to select a certain percentage of materials with the highest similarity, or materials with a similarity greater than a certain value, we can obtain a set of materials that better meet the needs of the chapter, further improving the accuracy and targeting of the matching.

[0069] Furthermore, the materials in the second selection set have a high degree of matching with the chapter feature vectors. The original text content is directly extracted from the second selection set and identified as the text content corresponding to the chapter feature vectors. The text content can be used to enrich and improve the chapter tender text, such as supplementing relevant cases, explaining technical points in detail, etc., so that the final tender text is more complete and accurate.

[0070] In this embodiment of the invention, extracting chapter feature vectors can accurately capture the core semantics and key information of chapter texts, while multi-level similarity matching can quickly and accurately locate text content that highly matches the feature vectors of each chapter, ensuring that the obtained text content closely matches the chapter requirements and effectively improving the quality and professionalism of bidding documents.

[0071] S5. Perform text fusion on the chapter tender document text and the text content to obtain the target tender document for the tender project.

[0072] In this embodiment of the invention, the target tender document refers to the generated chapter tender document text and corresponding text content that are integrated, redundant-free, logically connected, and formatted to form a target tender document that is structurally complete and semantically coherent.

[0073] In this embodiment of the invention, the step of text fusion of the chapter tender text and the text content to obtain the target tender document for the bidding project includes: Structural alignment analysis is performed on the chapter tender text and the text content to obtain the target tender chapter and the target tender text; Based on the preset bidding document template, the integration order and position mapping relationship of the target bid document chapters and the target bid document text are constructed; Based on the integration order and position mapping relationship, the target tender document chapters and the target tender document text are fused together to obtain the initial tender document; The initial tender document is optimized to obtain the target tender document for the tender project.

[0074] In this embodiment of the invention, text structure analysis technology is used to identify the text structure of the chapter proposal text and the text content, such as chapter titles, paragraph divisions, list formats, etc. For the chapter proposal text, its existing chapter framework is identified, such as different chapters like the introduction, project overview, and technical requirements. For the text content, its semantic logic and structural hierarchy are analyzed. Then, semantic similarity calculation and structure matching algorithms are used to align the text content with the structure of the chapter proposal text.

[0075] For example, if the text content is a detailed description of a bidding project, it will be matched with the "Technical Requirements" section in the chapter of the bid document to determine its corresponding position in the overall structure, thereby obtaining the target bid document chapter (i.e. the matched chapter) and the target bid document text (i.e. the matched specific content).

[0076] In detail, template matching and rule engine technology are used to determine the order of each target bid document section in the target bid document based on a pre-stored bid document specification template. This template defines the required chapter order, content elements of each chapter, and format requirements. Based on the target bid document chapters and text obtained from structural alignment analysis, and according to the rules in the bid document specification template, the order of each target bid document section within the target bid document is determined. For example, basic project information is written first, followed by the technical solution, and finally the commercial terms. Simultaneously, a positional mapping relationship is established between the target bid document text and its corresponding chapters, clarifying the position of each segment of the target bid document text within its respective chapter.

[0077] Furthermore, following the previously determined integration order, the target tender document chapters are arranged sequentially, and then, based on the position mapping relationship, the target tender document text is accurately inserted into the corresponding target tender document chapter positions. During the splicing process, the semantic coherence between texts is checked, and appropriate adjustments are made to unnatural transitions, such as adding transitional sentences or adjusting sentence order, to form the initial tender document.

[0078] Optionally, the initial tender document can be formatted and optimized. This involves adjusting the basic formatting such as font, font size, color, line spacing, and paragraph indentation to ensure it conforms to the specifications for tender documents. Specifically, the title of the initial tender document will be formatted uniformly for emphasis, and the tables and charts will be optimized for clear and easy-to-read layout. Additionally, the headers, footers, and page numbers will be checked for accuracy, and the overall layout will be aesthetically pleasing to make the target tender document appear more professional and standardized, thereby improving its readability and presentation.

[0079] In this embodiment of the invention, structural analysis of the chapter tender text and the matching text content can clearly grasp the semantic logic and framework hierarchy. Based on the integration order and position mapping relationship constructed according to the preset tender document specification template, the content integration is carried out in an orderly manner, avoiding logical confusion or information misalignment, which greatly improves the generation efficiency and quality of tender documents.

[0080] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0081] like Figure 4 The diagram shown is a functional block diagram of a plug-in-based tender document generation device provided in an embodiment of the present invention.

[0082] In this embodiment of the disclosure, a plug-in-based tender document generation device is provided, which corresponds one-to-one with the plug-in-based tender document generation method described in the above embodiments. For example... Figure 4 As shown, this plug-in-based tender document generation device 100 can be installed in an electronic device. According to its functions, the plug-in-based tender document generation device 100 includes a file semantic parsing module 101, a tender document plug-in configuration module 102, a chapter text generation module 103, a tender document content matching module 104, and a target tender document generation module 105. Detailed descriptions of each functional module are as follows: The document semantic parsing module 101 is used to obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata. The tender document plugin configuration module 102 is used to construct the tender document plugin for the bidding project based on the metadata of multiple bidding documents and the plugin configuration information. The chapter text generation module 103 is used to extract the requirement description statements of the bidding documents, generate chapters based on the requirement description statements of the bidding document plugin, and obtain multiple chapter bidding document texts. The tender document content matching module 104 is used to extract chapter feature vectors of multiple chapter tender document texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector; The target tender document generation module 105 is used to perform text fusion on the chapter tender document text and the text content to obtain the target tender document for the bidding project.

[0083] In one embodiment, when the document semantic parsing module 101 performs semantic parsing on the bidding documents to obtain multiple bidding document metadata, it is used to: The bidding documents are standardized to obtain standardized bidding documents. Extract the standardized text from the standardized bidding documents, and perform word segmentation on the standardized text to obtain multiple bidding text word segments; Semantic annotation is performed on multiple segments of the bidding text to obtain annotated bidding segments; Based on the labeled bidding and tendering word segmentation, the standardized text is context-associated to obtain contextual semantic information; The core metadata items in the labeled bidding and tendering word segment are extracted based on the contextual semantic information, and the core metadata items are stored in a structured manner to generate bidding and tendering document metadata.

[0084] In one embodiment, when the tender document plugin configuration module 102 executes the construction of the tender document plugin for the bidding project based on the metadata of the multiple bidding documents and the plugin configuration information, it is used to: Extract the plugin function elements from the plugin configuration information, and filter the bidding document metadata according to the plugin function elements to obtain the target metadata corresponding to the plugin function elements; The target metadata is converted into structured standard data that is readable by the plugin according to the preset data interface specification; The structured standard data is integrated with the preset plugin framework according to the format display requirements in the plugin configuration information to generate the initial plugin; The initial plugin is functionally verified and compatibility tested to generate the tender document plugin for the bidding project.

[0085] In one embodiment, when the chapter text generation module 103 generates chapters of the requirement description statement based on the tender document plugin to obtain multiple chapter tender documents, it is used to: The requirement description statement is cleaned to obtain the target requirement statement; The target requirement statements are classified to obtain the target tender document type corresponding to the target requirement statements; The target tender document type is matched using the tender document plugin to obtain the target chapter template. Semantically match the key requirement elements in the target requirement statement with the chapter fields of the target chapter template to establish a key chapter field correspondence table; Based on the key chapter field correspondence table, the key requirement elements are filled into the corresponding chapter fields of the target chapter template to generate initial chapter content; The initial chapter content is checked for coherence, resulting in multiple chapters of the tender document text.

[0086] In one embodiment, when the tender document content matching module 104 extracts the chapter feature vectors of the multiple chapter tender document texts, it is used to: Each chapter of the tender document is segmented to obtain multiple independent sentence units; The multiple independent sentence units are denoised and integrated to obtain a standardized sentence set. Each sentence in the standardized sentence set is transformed into a word vector to obtain a word vector sequence for each sentence; The word vector sequence of each sentence is weighted according to a predefined attention mechanism to obtain the sentence feature vector of each sentence; Aggregate and calculate the feature vectors of all the sentences to obtain the initial feature vector of the chapter title text; The initial feature vector is normalized to obtain the chapter feature vector.

[0087] In one embodiment, when the tender content matching module 104 performs multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector, it is used to: Based on the feature vectors of the chapters, a first-level coarse matching is performed in a preset document material database to obtain a set of candidate materials that are initially related to the feature vectors of each chapter; For each material in the candidate material set, feature extraction is performed to generate a material feature vector; Calculate the cosine similarity between the feature vector of each chapter and the feature vector of the material; The candidate material set is sorted and filtered according to the cosine similarity to obtain a second filtered material set; Extract the text from the second set of filtered materials, and determine the text as the text content corresponding to the feature vector of the chapter.

[0088] In one embodiment, when the target tender document generation module 105 performs text fusion on the chapter tender document text and the text content to obtain the target tender document for the bidding project, it is used to: Structural alignment analysis is performed on the chapter tender text and the text content to obtain the target tender chapter and the target tender text; Based on the preset bidding document template, the integration order and position mapping relationship of the target bid document chapters and the target bid document text are constructed; Based on the integration order and position mapping relationship, the target tender document chapters and the target tender document text are fused together to obtain the initial tender document; The initial tender document is optimized to obtain the target tender document for the tender project.

[0089] In this invention, the specific limitations of a plug-in-based tender document generation device can be found in the above-described limitations of a plug-in-based tender document generation method, and will not be repeated here. Each module in the aforementioned plug-in-based tender document generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0090] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a plug-in-based tender generation method on the server side.

[0091] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a plug-in-based tender generation method.

[0092] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata; The tender document plugin for the tender project is constructed based on the metadata of multiple tender documents and the plugin configuration information. Extract the requirement description statements from the bidding documents, and generate chapters from the requirement description statements according to the bidding document plugin to obtain multiple chapter bidding document texts; Extract chapter feature vectors from multiple chapter tender texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector; The text of the tender document and the text content of the aforementioned chapter are merged to obtain the target tender document for the tender project.

[0093] In the several embodiments provided by this invention, it should be understood that the disclosed devices and apparatuses can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0094] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0095] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0096] In some embodiments of this example, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the steps of the method described in the above embodiments.

[0097] The readable storage medium of the present invention stores a computer program, which, when executed by a processor of an electronic device, can perform the following: Obtain the bidding documents and plugin configuration information of the bidding project, perform semantic parsing on the bidding documents, and obtain multiple bidding document metadata; The tender document plugin for the tender project is constructed based on the metadata of multiple tender documents and the plugin configuration information. Extract the requirement description statements from the bidding documents, and generate chapters from the requirement description statements according to the bidding document plugin to obtain multiple chapter bidding document texts; Extract chapter feature vectors from multiple chapter tender texts, and perform multi-level similarity matching in a preset document material database based on the chapter feature vectors to obtain the text content corresponding to each chapter feature vector; The text of the tender document and the text content of the aforementioned chapter are merged to obtain the target tender document for the tender project.

[0098] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0099] Computer-readable storage media may also store at least one computer-executable program / instruction, such as computer-readable instructions. Computer-readable storage media include, but are not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Computer-readable storage media may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions stored on the computer-readable storage medium, the various methods described above can be performed.

[0100] In addition, the computer device may include (but is not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (e.g., keyboard, mouse, speakers, etc.).

[0101] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product / computer program product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.

[0102] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0103] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0104] In the embodiments provided in this disclosure, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0105] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

[0106] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

Claims

1. A plug-in based tender generation method, characterized by, The method comprises: obtaining the bidding document and plug-in configuration information of the bidding project, performing semantic analysis on the bidding document to obtain a plurality of bidding document metadata; constructing the bidding document plug-in of the bidding project according to the plurality of bidding document metadata and the plug-in configuration information; extracting the requirement description sentence of the bidding document, generating chapters according to the bidding document plug-in, and obtaining a plurality of chapter bidding text; extracting the chapter feature vector of the plurality of chapter bidding text, and performing multi-level similarity matching in a preset document material database according to the chapter feature vector to obtain text content corresponding to each chapter feature vector; performing text fusion on the chapter bidding text and the text content to obtain the target bidding document of the bidding project.

2. The plug-in based proposal generation method of claim 1, wherein, The semantic analysis on the bidding document to obtain a plurality of bidding document metadata comprises: performing format standardization processing on the bidding document to obtain a standardized bidding document; extracting the standardized text in the standardized bidding document, performing word segmentation processing on the standardized text to obtain a plurality of bidding text word segmentation; performing semantic annotation on the plurality of bidding text word segmentation to obtain annotated bidding word segmentation; performing context association on the standardized text based on the annotated bidding word segmentation to obtain context semantic information; extracting the core metadata item in the annotated bidding word segmentation according to the context semantic information, and storing the core metadata item in a structured manner to generate bidding document metadata.

3. The plug-in based proposal generation method of claim 1, wherein, The construction of the bidding document plug-in of the bidding project according to the plurality of bidding document metadata and the plug-in configuration information comprises: extracting the plug-in function element in the plug-in configuration information, and performing data filtering on the bidding document metadata according to the plug-in function element to obtain target metadata corresponding to the plug-in function element; convert the target metadata into structured standard data readable by the plug-in according to a preset data interface specification; integrate the structured standard data with a preset plug-in framework according to the format display requirement in the plug-in configuration information to generate an initial plug-in; perform function verification and compatibility test on the initial plug-in to generate the bidding document plug-in of the bidding project.

4. The plug-in based proposal generation method of claim 1, wherein, The text fusion on the chapter bidding text and the text content to obtain the target bidding document of the bidding project comprises: performing structure alignment analysis on the chapter bidding text and the text content to obtain target chapter and target text of the bidding document; constructing the integration order and position mapping relationship of the target chapter and the target text of the bidding document based on a preset bidding document specification template; performing content fusion processing on the target chapter and the target text of the bidding document according to the integration order and position mapping relationship to obtain an initial bidding document; performing format optimization on the initial bidding document to obtain the target bidding document of the bidding project.

5. The plug-in based proposal generation method of claim 1, wherein, The chapter generation according to the requirement description sentence of the bidding document plug-in to obtain a plurality of chapter bidding text comprises: Text cleaning is performed on the requirement description sentence to obtain a target requirement sentence; Requirement classification is performed on the target requirement sentence to obtain a target bid document type corresponding to the target requirement sentence; Template matching is performed on the target bid document type by using the bid document plug-in to obtain a target chapter template; Semantic matching is performed between a key requirement element in the target requirement sentence and each chapter field of the target chapter template to establish a key chapter field correspondence table; According to the key chapter field correspondence table, the key requirement element is filled into a corresponding chapter field of the target chapter template to generate initial chapter content; Conciseness verification is performed on the initial chapter content to obtain a plurality of chapter bid document texts.

6. The plug-in based proposal generation method of claim 1, wherein, The extraction of the chapter feature vectors of the plurality of chapter bid document texts includes: Text segmentation processing is performed on each of the chapter bid document texts to obtain a plurality of independent sentence units; Denoising integration processing is performed on the plurality of independent sentence units to obtain a standardized sentence set; Word vector conversion is performed on each sentence in the standardized sentence set to obtain a word vector sequence of each sentence; According to a predefined attention mechanism, weight distribution is performed on the word vector sequence of each sentence to obtain a sentence feature vector of each sentence; Aggregation calculation is performed on all the sentence feature vectors to obtain an initial feature vector of the chapter bid document text; Vector normalization processing is performed on the initial feature vector to obtain a chapter feature vector.

7. The plug-in based proposal generation method of claim 1, wherein, The multi-level similarity matching in the preset document material database according to the chapter feature vector to obtain text content corresponding to each chapter feature vector includes: First-level rough matching is performed in the preset document material database according to the chapter feature vector to obtain a candidate material set preliminarily related to each chapter feature vector; Feature extraction is performed on each material in the candidate material set to generate a material feature vector; Cosine similarity is calculated between each chapter feature vector and the material feature vector; According to the cosine similarity, the candidate material set is sorted and filtered to obtain a second filtered material set; Material text in the second filtered material set is extracted, and the material text is determined as the text content corresponding to the chapter feature vector.

8. A plug-in based tender generation apparatus, characterized by comprising: The apparatus includes: A file semantic analysis module configured to obtain a bidding document of a bidding project and plug-in configuration information, perform semantic analysis on the bidding document, and obtain a plurality of bidding document metadata; A bid document plug-in configuration module configured to construct a bid document plug-in of the bidding project according to the plurality of bidding document metadata and the plug-in configuration information; A chapter text generation module configured to extract a requirement description sentence of the bidding document, perform chapter generation on the requirement description sentence according to the bid document plug-in, and obtain a plurality of chapter bid document texts; A bid document content matching module configured to extract chapter feature vectors of the plurality of chapter bid document texts, and perform multi-level similarity matching in a preset document material database according to the chapter feature vectors to obtain text content corresponding to each chapter feature vector. The target tender document generation module is configured to perform text fusion on the chapter tender document text and the text content to obtain a target tender document of the tender project.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the plug-in based tender document generation method according to any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the plug-in based tender document generation method according to any one of claims 1 to 7.

Images