Project research content duplicate checking method and device based on semantic alignment and electronic equipment

Abstract

The invention discloses a semantic alignment-based project research content duplicate checking method and apparatus, and an electronic device, relates to the technical field of natural language processing, and can significantly improve the duplicate checking accuracy and comprehensively evaluate the overall and local similarities between a to-be-checked project and a plurality of historical science and technology projects. The method comprises the following steps: in response to a science and technology project duplicate checking request, determining a to-be-checked project; performing step-by-step semantic alignment on the project content of the to-be-checked project and the project content of the plurality of historical science and technology projects by utilizing a semantic duplicate checking portrait library to obtain semantic similarity between the to-be-checked project and each historical science and technology project; calculating the text similarity between the to-be-checked project and each historical science and technology project by using the text duplicate checking portrait library; according to the semantic similarity and the text similarity, calculating the comprehensive similarity between the to-be-checked project and each historical science and technology project, and generating a comprehensive duplicate checking result; and comparing and analyzing the to-be-checked items with similar items indicated by the comprehensive duplicate checking result to obtain a comparison report and outputting the comparison report.

Description

Technical Field

[0001] The present application relates to the field of natural language processing technology, and in particular to a method, device and electronic device for checking for duplicate content in project research based on semantic alignment. Background Art

[0002] In the field of science and technology project management, project planning and project establishment, as key links in new management work, play a vital role in promoting cutting-edge research, technological innovation and optimizing the allocation of innovation resources. It is not only related to the company's innovation input and output efficiency, but also an important guarantee for improving the overall innovation management level. However, in actual operation, due to the large number of projects submitted by business departments and the complex and cross-cutting research directions, how to scientifically and impartially evaluate and check the establishment of various science and technology projects within a limited time, so as to screen out high-value innovation projects, has become a major challenge facing current science and technology project management.

[0003] In the related technology, the traditional project duplication checking scheme mainly relies on the basic duplication checking database, and realizes the duplication checking function by comparing the text content similarity of basic metadata such as project name, introduction and research content. However, the applicant recognizes that the traditional natural language processing and text mining technology have obvious shortcomings in the accuracy of duplication checking. Due to the relatively backward algorithm, the duplication checking results are unsatisfactory, and often require a lot of manual secondary screening and review work; moreover, most of the existing duplication checking schemes only support overall duplication checking evaluation. For some duplication checking tasks that require accurate identification and discovery of whether there is repeated research on sub-topics between multiple texts, the existing duplication checking schemes cannot meet their needs. In summary, the traditional project duplication checking scheme not only has a high time cost and low duplication checking efficiency, but also has certain limitations, and it is difficult to adapt to the duplication checking needs of scientific and technological projects. Summary of the invention

[0004] In view of this, the present application provides a method, device and electronic device for checking for plagiarism in project research content based on semantic alignment. The main purpose is to solve the problem that the current traditional project plagiarism checking solutions not only have high time costs and low plagiarism checking efficiency, but also have certain limitations and are difficult to adapt to the plagiarism checking needs of scientific and technological projects.

[0005] According to the first aspect of the present application, a method for checking for duplicate content in project research based on semantic alignment is provided, the method comprising:

[0006] In response to a request for checking for duplicate technology projects, determining the project to be checked indicated by the request for checking for duplicate technology projects;

[0007] Using a semantic duplicate checking image library, semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects. The semantic duplicate checking image library is constructed after semantic element mining of the multiple historical science and technology projects.

[0008] Calculate the text similarity between the item to be checked and each of the historical science and technology items by using a text duplicate checking image library, wherein the text duplicate checking image library is constructed after metadata mining of the multiple historical science and technology items;

[0009] Calculate the comprehensive similarity between the item to be checked and each of the historical science and technology items based on the semantic similarity and the text similarity, and generate a comprehensive duplicate checking result;

[0010] The items to be checked are compared and analyzed with similar items indicated by the comprehensive duplicate checking results to obtain and output a comparison report.

[0011] According to a second aspect of the present application, a device for checking duplicate content of project research based on semantic alignment is provided, the device comprising:

[0012] A determination module, configured to respond to a request for checking for duplicate technology projects and determine the items to be checked indicated by the request for checking for duplicate technology projects;

[0013] A semantic similarity calculation module is used to use a semantic duplicate checking image library to semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects, wherein the semantic duplicate checking image library is constructed after semantic element mining of the multiple historical science and technology projects;

[0014] A text similarity calculation module, used to calculate the text similarity between the item to be checked and each of the historical science and technology items by using a text duplicate checking image library, wherein the text duplicate checking image library is constructed after metadata mining of the multiple historical science and technology items;

[0015] A comprehensive similarity calculation module, used to calculate the comprehensive similarity between the item to be checked and each of the historical science and technology items according to the semantic similarity and the text similarity, and to generate a comprehensive duplicate checking result;

[0016] The output module is used to compare and analyze the items to be checked with similar items indicated by the comprehensive duplication check results, obtain a comparison report and output it.

[0017] According to a third aspect of the present application, an electronic device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0018] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0019] Through the above technical scheme, the present application provides a method, device and electronic device for checking project research content for plagiarism based on semantic alignment. The present application responds to a request for checking for plagiarism in a scientific and technological project, determines the project to be checked indicated by the request for checking for plagiarism in a scientific and technological project, uses a semantic plagiarism checking portrait library to semantically align the project content of the project to be checked with the project content of multiple historical scientific and technological projects step by step, obtains the semantic similarity between the project to be checked and each historical scientific and technological project, uses a text plagiarism checking portrait library to calculate the text similarity between the project to be checked and each historical scientific and technological project, calculates the comprehensive similarity between the project to be checked and each historical scientific and technological project based on the semantic similarity and text similarity, and generates a comprehensive plagiarism checking result. , compare and analyze the project to be checked with similar projects indicated by the comprehensive duplication checking results, obtain and output a comparison report, and by introducing a semantic duplication checking portrait library and a text duplication checking portrait library, realize the step-by-step semantic alignment and text similarity calculation of the project content between the project to be checked and the historical scientific and technological projects, thereby significantly improving the accuracy of duplication checking, reducing the workload of manual secondary screening and review, reducing time costs, and improving the efficiency of duplication checking; at the same time, it can comprehensively evaluate the overall and local similarities between the project to be checked and multiple historical scientific and technological projects, accurately identify the repeated research of sub-topics, effectively overcome the limitations of traditional duplication checking schemes in duplication checking efficiency and accuracy, and better adapt to the duplication checking needs of scientific and technological projects.

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0022] Figure 1 A schematic diagram of a method for checking duplicate content of project research based on semantic alignment provided in an embodiment of the present application is shown;

[0023] Figure 2 A schematic diagram of a method for constructing a semantic duplicate checking image library provided in an embodiment of the present application is shown;

[0024] Figure 3 A schematic diagram of a method for constructing a text duplication detection image library provided in an embodiment of the present application is shown;

[0025] Figure 4 A schematic diagram of a method for calculating semantic similarity provided in an embodiment of the present application is shown;

[0026] Figure 5 A schematic diagram of a method for calculating text similarity provided in an embodiment of the present application is shown;

[0027] Figure 6 A schematic diagram of a method flow for calculating a comprehensive similarity and generating a comprehensive duplicate checking result provided in an embodiment of the present application is shown;

[0028] Figure 7 A schematic diagram of another method for checking duplicate content of project research based on semantic alignment provided in an embodiment of the present application is shown;

[0029] Figure 8 A schematic diagram of the structure of a device for checking duplicate content of project research based on semantic alignment provided in an embodiment of the present application is shown;

[0030] Fig. 9 A schematic diagram of the device structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0031] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0032] The present application embodiment provides a method for checking for duplicate content in project research based on semantic alignment, such as Figure 1 As shown, the method includes:

[0033] 101. In response to a request for checking for duplicate technology projects, determine the projects to be checked indicated by the request for checking for duplicate technology projects.

[0034] The embodiments of the present application can be applied to the duplicate checking system, which is a tool designed for checking duplicate content in scientific and technological projects. It integrates semantic analysis technology and text comparison algorithms to achieve in-depth mining and accurate comparison of project content. The workflow of the duplicate checking system is highly automated. Users only need to submit relevant information of the project to be checked, and the duplicate checking system can automatically parse, compare and analyze it to generate a comprehensive duplicate checking report. In addition, the duplicate checking system also has a high degree of flexibility and scalability, and can be applied to different fields and different types of scientific and technological projects, providing users with personalized duplicate checking services.

[0035] When a user has a need for duplicate checking, he or she can initiate a technology project duplicate checking request to the duplicate checking system. In response to the technology project duplicate checking request, the duplicate checking system will first parse the request, determine the items to be checked indicated by the technology project duplicate checking request, ensure that the items to be checked are accurately identified, and avoid duplicate checking failures due to incorrect or missing information.

[0036] 102. Using the semantic duplicate checking portrait library, the project content of the project to be checked is semantically aligned with the project content of multiple historical and technological projects step by step to obtain the semantic similarity between the project to be checked and each historical and technological project.

[0037] In the embodiment of the present application, after determining the project to be checked, the duplicate checking system will use the semantic duplicate checking portrait library to semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step, and obtain the semantic similarity between the project to be checked and each historical science and technology project. Among them, the semantic duplicate checking portrait library is constructed after semantic element mining of multiple historical science and technology projects. The semantic duplicate checking portrait library does not simply store the text content of historical science and technology projects, but through in-depth mining of the core metadata information of each project, extracting representative and discriminative semantic features, and then constructing a multi-dimensional and refined project portrait. These portraits not only contain basic information such as the theme and keywords of the project, but also cover the deep semantic content such as the research methods, technical routes, and innovations of the project, providing a rich information basis for subsequent semantic alignment and similarity calculation.

[0038] When using the semantic duplicate checking image library to calculate the semantic similarity between the item to be checked and each historical science and technology item, the duplicate checking system will input the project content of the item to be checked into the semantic duplicate checking image library, and perform semantic alignment with each historical science and technology item in the library step by step to accurately capture the semantic associations and differences between the items. Through alignment, the duplicate checking system can calculate the semantic similarity between the item to be checked and each historical item, and the semantic similarity reflects the degree of closeness between the two at the semantic level.

[0039] 103. Use the text duplication detection image library to calculate the text similarity between the project to be checked and each historical science and technology project.

[0040] In the embodiments of the present application, considering that the plagiarism check for scientific and technological projects is different from the plagiarism check for texts, pure text similarity comparison and short text similarity analysis cannot achieve accurate plagiarism check for projects, and it is necessary to explore and practice semantic plagiarism check for projects. Therefore, the plagiarism check system will continue to mine semantic features from the core metadata of the project, and combine the above-mentioned means of aligning the multi-dimensional semantic features of different projects to judge their similarity through the differences between the semantic characteristics.

[0041] Among them, in order to realize the mining of metadata, a text duplication detection portrait library is built in the duplicate detection system. The text duplication detection portrait library is constructed after metadata mining of multiple historical science and technology projects. The text duplication detection portrait library does not simply store the full text content of historical science and technology projects, but through in-depth mining of the metadata of each project, it constructs a portrait containing multi-dimensional content such as basic project information, abstracts, key paragraphs, etc. These portraits not only cover the text content of the project, but also pre-process and extract features of the text through specific algorithms and technologies to facilitate subsequent similarity calculations.

[0042] In the specific implementation process, the duplicate checking system will compare the text content of the project to be checked with each historical science and technology project in the text duplicate checking portrait library to evaluate the similarity between the project to be checked and the historical science and technology projects at the text level. Through calculation, the duplicate checking system can obtain the text similarity value between the project to be checked and each historical project. The text similarity value reflects the closeness of the two in terms of text content, and can evaluate the degree of duplication between projects from another dimension, providing more comprehensive and accurate support for the duplicate checking results; at the same time, the construction of the text duplicate checking portrait library also fully considers the characteristics of science and technology projects, and improves the accuracy and efficiency of text similarity calculation by deeply mining metadata information.

[0043] 104. Based on semantic similarity and text similarity, calculate the comprehensive similarity between the project to be checked and each historical science and technology project, and generate a comprehensive duplicate checking result.

[0044] In the embodiment of the present application, the duplicate checking system will calculate the comprehensive similarity between the project to be checked and each historical scientific and technological project based on the semantic similarity and text similarity, and realize the semantic level duplicate checking and text level duplicate checking for the project to be checked. The process of calculating the comprehensive similarity not only takes into account the project similarity at the semantic level, but also extracts metadata items for scientific and technological projects, and fully compares the project similarity at the text level, which has good adaptability to the similarity check of scientific and technological projects.

[0045] Among them, the duplicate checking system can calculate the comprehensive similarity through a weighted algorithm, and the comprehensive similarity can more comprehensively and accurately reflect the degree of similarity between the project to be checked and the historical project. Furthermore, the duplicate checking system will generate a comprehensive duplicate checking result for scientific and technological projects based on the calculated comprehensive similarity. The comprehensive duplicate checking result not only lists the historical projects with high similarity to the project to be checked, but also provides the specific numerical value and comparison details of the similarity, which helps users quickly identify potential duplicate or similar projects. In this way, through the integration of the dual-engine duplicate checking results, the accuracy and efficiency of the duplicate checking of scientific and technological projects can be greatly improved, avoiding the waste of resources and obstruction of scientific research innovation caused by duplicate or similar projects.

[0046] 105. Compare and analyze the items to be checked with similar items indicated by the comprehensive duplicate checking results, obtain a comparison report and output it.

[0047] In the embodiment of the present application, the duplicate checking system will further carry out detailed comparison and analysis at the full-text level based on the similar items determined by the comprehensive duplicate checking results, aiming to generate and output a detailed comparison report.

[0048] Among them, the above steps describe that a semantic duplicate checking portrait library and a text duplicate checking portrait library are constructed in the duplicate checking system. Therefore, the embodiment of the present application implements the semantic duplicate checking engine function based on the semantic duplicate checking portrait library, and implements the text duplicate checking engine function based on the text duplicate checking portrait library, and uses the semantic duplicate checking engine and the text duplicate checking engine to perform a comprehensive comparison between the items to be checked and similar items. The text duplicate checking engine is mainly responsible for identifying and marking the same or similar paragraphs, sentences and even words between the two, while the semantic duplicate checking engine goes a step further, by analyzing the semantic structure and meaning of the text, identifying potential similarities and differences, thereby more comprehensively and accurately evaluating the degree of similarity between the items to be checked and similar items.

[0049] Furthermore, the duplicate checking system will use the analysis results to generate a comparison report. The comparison report not only lists the similarities and differences between the project to be checked and similar projects, but also provides specific text comparison details and semantic analysis results, providing users with intuitive and clear comparison results, helping users to better understand the similarities and differences between the project to be checked and similar projects, assisting in evaluating the innovation and feasibility of the project, and providing strong support for the rational allocation of scientific research resources.

[0050] The method provided in the embodiment of the present application responds to a request for checking for duplicate technology projects, determines the project to be checked indicated by the request for checking for duplicate technology projects, uses a semantic checking for duplicate technology portrait library to semantically align the project content of the project to be checked with the project content of multiple historical technology projects step by step, obtains the semantic similarity between the project to be checked and each historical technology project, uses a text checking for duplicate technology portrait library to calculate the text similarity between the project to be checked and each historical technology project, calculates the comprehensive similarity between the project to be checked and each historical technology project based on the semantic similarity and the text similarity, and generates a comprehensive checking result, aligns the project to be checked with the project content indicated by the comprehensive checking for duplicate technology results, and obtains the semantic similarity between the project to be checked and each historical technology project. Similar projects are compared and analyzed, and a comparison report is obtained and output. By introducing the semantic duplicate detection portrait library and the text duplicate detection portrait library, the step-by-step semantic alignment and text similarity calculation of the project content between the project to be checked and the historical scientific and technological projects are achieved, thereby significantly improving the accuracy of duplicate detection, reducing the workload of manual secondary screening and review, reducing time costs, and improving duplicate detection efficiency; at the same time, it can comprehensively evaluate the overall and local similarities between the project to be checked and multiple historical scientific and technological projects, accurately identify the repeated research of sub-topics, effectively overcome the limitations of traditional duplicate detection schemes in duplicate detection efficiency and accuracy, and better adapt to the duplicate detection needs of scientific and technological projects.

[0051] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides a method for constructing a semantic duplicate detection image library, such as Figure 2 As shown, including:

[0052] 201. Acquire multiple historical science and technology projects, and extract multiple project semantic elements of each historical science and technology project based on a pre-trained semantic indexing model.

[0053] In the embodiment of the present application, the duplicate checking system collects multiple historical science and technology projects from a wide range of science and technology project databases, and uses a pre-trained semantic indexing model to deeply extract the project semantic elements of each historical science and technology project. Among them, the project semantic elements include but are not limited to the research object, research content, research attributes and limiting factors, which together constitute a comprehensive and in-depth disclosure of the project content.

[0054] Specifically, the duplicate checking system will obtain multiple historical science and technology projects, and then use the pre-trained semantic indexing model, which is built based on deep learning technology, such as BERT, Transformer, etc., to perform semantic analysis on the project text, and can automatically mine the implicit research objects, research content, research attributes and limiting factors. For example, a historical science and technology project on "new energy vehicle battery management system" may have its research object indexed as "new energy vehicle battery", the research content may be "battery management strategy", the research attribute may be "high efficiency and safety", and the limiting factor may be "specific temperature range". In this way, through semantic indexing, the duplicate checking system can more accurately capture the core information of the project, lay a solid foundation for subsequent comparison and duplicate checking, and compared with traditional keyword matching, it can more deeply understand the text content and improve the accuracy and comprehensiveness of duplicate checking.

[0055] 202. Based on the pre-trained retrieval indexing model, multiple project semantic elements of each historical science and technology project are labeled, and each labeled project semantic element is converted into a vector to obtain multiple semantic element vectors corresponding to each historical science and technology project.

[0056] In an embodiment of the present application, after extracting the semantic elements of the project, the duplicate checking system will use a pre-trained retrieval indexing model, such as an indexing model based on a neural vector space model (NVSM), to label each project semantic element and convert it into a vector form, so that the project semantic elements are converted into a numerical form that can be processed by a computer, facilitating subsequent rapid comparison and retrieval.

[0057] In the actual application process, since the retrieval indexing model can be an NVSM model, the duplicate checking system will first label each extracted semantic element of the project, and then use the NVSM model to convert each labeled semantic element into a high-dimensional vector, which can well represent the similarities and differences between semantic elements. Furthermore, in order to improve the efficiency of comparison, the embodiment of the present application can also introduce vector databases, such as Faiss, Elasticsearch, etc., to efficiently store and retrieve large-scale vector data through the vector database, thereby greatly improving the speed and efficiency of batch comparison; at the same time, the combination of label annotation and vector conversion enables the system to more flexibly handle various complex semantic information.

[0058] 203. Utilize the multiple semantic element vectors corresponding to each historical science and technology project to build a semantic duplicate detection portrait library.

[0059] In the embodiment of the present application, the duplicate checking system will integrate multiple semantic element vectors corresponding to each historical science and technology project to build a complete semantic duplicate checking image library. The semantic duplicate checking image library contains a large number of semantic features of historical science and technology projects, which is the basis for subsequent duplicate checking and comparison.

[0060] In the actual application process, when building a semantic duplicate checking image library, the duplicate checking system will store all semantic element vectors of each historical science and technology project persistently. Specifically, you can choose a high-performance storage engine, such as KBASE, to ensure the stability and accessibility of the data. Storage engines such as KBASE not only support efficient read and write operations, but also provide a wealth of query and optimization functions, so that historical science and technology projects similar to the project to be checked can be quickly retrieved from the semantic duplicate checking image library. In this way, by building a semantic duplicate checking image library, the duplicate checking system realizes the comprehensive recording and efficient management of the semantic features of historical projects, which not only provides rich data support for subsequent duplicate checking and comparison, but also enables the duplicate checking system to continuously optimize and improve the performance of duplicate checking as the number of historical science and technology projects continues to increase.

[0061] The method provided in the embodiment of the present application realizes in-depth disclosure of historical scientific and technological projects by constructing a semantic duplication detection portrait library, so as to facilitate the subsequent evaluation of the local similarities between scientific and technological projects, improve the accuracy and efficiency of duplication detection, and reduce the workload of manual review.

[0062] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides a method for constructing a text duplicate detection image library, such as Figure 3 As shown, including:

[0063] 301. Acquire multiple historical scientific and technological projects, and extract multiple project metadata of each historical scientific and technological project based on a pre-trained knowledge meta-indexing model.

[0064] In the embodiment of the present application, the duplicate checking system first collects a plurality of historical scientific and technological projects. In order to ensure that the extracted metadata is accurate and comprehensive, the duplicate checking system adopts a pre-trained knowledge meta-indexing model. The knowledge meta-indexing model is based on deep learning technology, and learns the extraction rules of key information in the text of scientific and technological projects through a large amount of training data. During the extraction process, the knowledge meta-indexing model can automatically identify and extract key metadata such as project titles, keywords, abstracts, technical routes, and sub-topics. These metadata are the basis for constructing a text duplicate checking portrait library. They not only reflect the core content of the project, but also provide an important basis for subsequent vector conversion and duplicate checking comparison.

[0065] For example, suppose the title of a scientific project is "Research on Image Processing Technology Based on Artificial Intelligence", the keywords include "artificial intelligence", "image processing", "deep learning", etc., the abstract outlines the research purpose, methods and preliminary results, and the technical route describes in detail the entire process from data preprocessing to model training to result evaluation. The sub-topics cover algorithm optimization, data set construction and other aspects. Through the knowledge meta-indexing model, the duplicate checking system can quickly and accurately extract these key information as the metadata of the scientific and technological project, laying the foundation for the subsequent construction of a text duplicate checking portrait library.

[0066] 302. Based on the pre-trained vector indexing model, multiple project metadata of each historical science and technology project are converted into vectors to obtain multiple project metadata vectors corresponding to each historical science and technology project.

[0067] In an embodiment of the present application, after extracting the project metadata, the duplicate checking system uses a pre-trained vector indexing model to perform vector conversion on multiple project metadata of each historical science and technology project, and maps the project metadata into a high-dimensional vector space, so that similar texts are closer in the vector space, fully capturing the semantic information in the text, and ensuring the accuracy of the vector conversion.

[0068] Specifically, the duplicate checking system will input multiple project metadata (such as title, keywords, abstract, etc.) of each project into the vector indexing model to obtain the corresponding vector representation.

[0069] 303. Utilize multiple project metadata vectors corresponding to each historical science and technology project to build a text duplication detection portrait library.

[0070] In the embodiment of the present application, after completing the vector conversion, in order to efficiently perform text duplication checking, the duplication checking system will store these vectors in a vector database to realize the construction of a text duplication checking portrait library. The text duplication checking portrait library supports efficient vector retrieval and comparison operations, and can complete the duplication checking task of a large amount of text in a short time. In this way, the constructed text duplication checking portrait library not only contains rich historical science and technology project information, but also has a powerful duplication checking ability, and can also convert the metadata of the project to be checked into a vector form, and compare it with the vector in the text duplication checking portrait library, so as to quickly determine whether there is repeated or similar content in the project to be checked at the text level, which not only improves the accuracy and efficiency of duplication checking, but also greatly reduces the cost of manual review.

[0071] The method provided in the embodiment of the present application realizes in-depth mining and efficient management of metadata of historical scientific and technological projects by constructing a text duplication detection portrait library based on knowledge element indexing and vector conversion, facilitates efficient retrieval and comparison in vector space, and improves the efficiency of duplication detection.

[0072] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides a method for calculating semantic similarity, such as Figure 4 As shown, using the semantic duplicate checking image library, the project content of the project to be checked is semantically aligned with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each historical science and technology project, including:

[0073] 401. Extract multiple target semantic elements of the item to be checked from the item content of the item to be checked.

[0074] In the embodiment of the present application, the duplicate checking system uses semantic mining technology to accurately identify and extract multiple key target semantic elements for the project content of the project to be checked. The target semantic elements comprehensively cover the target research object, target research content, target research attributes, and target limiting factors of the project, constituting a comprehensive and in-depth analysis of the project content. Through semantic mining, the duplicate checking system will be able to align semantic features at the semantic level in the future, thereby accurately discovering the similarities and differences between projects, laying a solid foundation for subsequent duplicate checking work.

[0075] 402. Based on the pre-trained retrieval indexing model, label multiple target semantic elements, and perform vector conversion on each labeled target semantic element to obtain a target semantic element vector corresponding to each target semantic element.

[0076] In the embodiment of the present application, after extracting the target semantic elements, the duplicate checking system uses the pre-trained retrieval indexing model to label each target semantic element and convert it into a vector form to obtain the target semantic element vector corresponding to each target semantic element, thereby fully capturing the semantic information in the text and ensuring the accuracy of the vector conversion. The target semantic element vector not only retains the key information of the original text, but also plays an important role in the subsequent similarity calculation, supporting the duplicate checking system to efficiently compare and analyze items.

[0077] 403. Extract a first target semantic element for indicating a target research object from multiple target semantic elements, and obtain a first target semantic element vector corresponding to the first target semantic element. Based on the semantic duplicate checking portrait library, calculate the similarity between the first target semantic element vector and the semantic element vector for indicating the research object in the semantic element vector corresponding to each historical science and technology project, and obtain the research object similarity between the project to be checked and each historical science and technology project.

[0078] In the embodiment of the present application, after obtaining multiple target semantic elements, the duplicate checking system starts to calculate the similarity between the item to be checked and other historical science and technology items. Among them, since the target semantic elements cover multiple different contents, the duplicate checking system will compare the different contents one by one to obtain the semantic similarity between the item to be checked and each historical science and technology item.

[0079] Specifically, the duplicate checking system will first make a semantic relevance judgment at the research object level, that is, extract the first target semantic element used to indicate the target research object from multiple target semantic elements, and obtain its corresponding first target semantic element vector. Then, based on the semantic duplicate checking portrait library, calculate the similarity between the first target semantic element vector and the semantic element vector used to indicate the research object corresponding to each historical science and technology project. Subsequently, by setting the first similarity threshold, the duplicate checking system can determine whether there is duplication between the project to be checked and each historical project at the research object level.

[0080] Among them, it should be noted that if the similarity of the research objects between the project to be checked and all historical scientific and technological projects is less than the first similarity threshold, it can be determined that there is no project in the current historical scientific and technological projects that repeats the research object of the project to be checked. Since the research object of the project to be checked is an innovative existence, it can also be determined that there is no project in the current historical scientific and technological projects that is highly similar to the project to be checked. Therefore, the duplicate checking system will generate corresponding duplicate checking feedback and output it, thereby effectively avoiding unnecessary resource consumption in subsequent steps and improving the efficiency of duplicate checking.

[0081] 404. When the similarity of the research object between the project to be checked and at least one historical and technological project is greater than or equal to a first similarity threshold, a second target semantic element for indicating the target research content is extracted from multiple target semantic elements, and a second target semantic element vector corresponding to the second target semantic element is obtained. Based on the semantic duplication detection portrait library, the similarity between the second target semantic element vector and the semantic element vector corresponding to each historical and technological project for indicating the research content is calculated to obtain the research content similarity between the project to be checked and each historical and technological project.

[0082] In the embodiment of the present application, if the similarity of the research object between the project to be checked and at least one historical project is greater than or equal to the first similarity threshold, the duplicate checking system will further extract the second target semantic element used to indicate the target research content, and calculate its similarity with the semantic element vector of the historical science and technology project at the research content level. Subsequently, by setting the second similarity threshold, the duplicate checking system can further determine whether there is duplication between the project to be checked and each historical project at the research content level.

[0083] Correspondingly, if the similarity in research content between the project to be checked and all historical scientific and technological projects is less than the second similarity threshold, the duplicate checking system can also determine that there are no projects in the current historical scientific and technological projects that duplicate the research content of the project to be checked. Similarly, since the research content of the project to be checked is innovative, it can also be determined that there are no projects in the current historical scientific and technological projects that are highly similar to the project to be checked. Therefore, the duplicate checking system will generate and output corresponding duplicate checking feedback, further narrowing the scope of the duplicate checking and improving the accuracy and efficiency of the duplicate checking.

[0084] 405. When the similarity of research contents between the project to be checked and at least one historical and technological project is greater than or equal to a second similarity threshold, a third target semantic element for indicating the target research attribute is extracted from multiple target semantic elements, and a third target semantic element vector corresponding to the third target semantic element is obtained; based on the semantic duplication checking portrait library, the similarity between the third target semantic element vector and the semantic element vector for indicating research attributes in the semantic element vector corresponding to each historical and technological project is calculated to obtain the research attribute similarity between the project to be checked and each historical and technological project.

[0085] In an embodiment of the present application, if the research content similarity between the project to be checked and at least one historical science and technology project is greater than or equal to a second similarity threshold, the duplication checking system will further extract a third target semantic element used to indicate the target research attribute, and calculate its similarity with the historical science and technology project at the research attribute level.

[0086] Among them, projects with aligned research attributes can basically be identified as highly semantically related projects, and can be determined as semantically duplicate research projects without any limiting factors. Therefore, a third similarity threshold is set in the duplicate checking system, and the third similarity threshold is used to judge the similarity of research attributes. The duplicate checking system can ultimately determine whether there are projects that are semantically duplicate with the project to be checked. In other words, if the research attribute similarity between the project to be checked and at least one designated historical project is greater than or equal to the third similarity threshold, the duplicate checking system will directly treat these designated historical projects as duplicate projects that are semantically duplicate with the project to be checked, and will not perform subsequent comparisons of limiting factors, and will directly generate abnormal feedback carrying these projects.

[0087] 406. When the research attribute similarities between the project to be checked and each historical and technological project are less than the third similarity threshold, a fourth target semantic element for indicating the target limiting factor is extracted from multiple target semantic elements, and a fourth target semantic element vector corresponding to the fourth target semantic element is obtained. Based on the semantic duplication detection portrait library, the similarity between the fourth target semantic element vector and the semantic element vector corresponding to each historical and technological project for indicating the semantic element vector of the limiting factor is calculated to obtain the limiting factor similarity between the project to be checked and each historical and technological project.

[0088] In the embodiment of the present application, if the research attribute similarity between the project to be checked and all historical projects is less than the third similarity threshold, the duplicate checking system will continue to extract the fourth target semantic element used to indicate the target limiting factor, and calculate its similarity with the historical science and technology project at the limiting factor level to obtain the limiting factor similarity. Among them, the calculation of similarity at the limiting factor level is particularly important for some fields with relatively in-depth research. Through further judgment by limiting factors, the time, space, state and other conditional attributes of the project research can be restricted, so as to more accurately judge the semantic relevance between projects.

[0089] 407. Calculate the semantic similarity between the project to be checked and each historical science and technology project based on the similarity of research objects, research contents, research attributes and limiting factors between the project to be checked and each historical science and technology project.

[0090] In the embodiment of the present application, the duplicate checking system will calculate a comprehensive semantic similarity based on the similarity of the research object, research content, research attribute and limiting factor between the project to be checked and each historical science and technology project. Specifically, the duplicate checking system will fuse the semantic feature items under each dimension with the similarity values ​​of the comparison project by weighted summation, so as to obtain a comprehensive similarity evaluation result, that is, the semantic similarity. Among them, the semantic similarity can be calculated using the following formula 1:

[0091] Formula 1:

[0092] Among them, Identical is used to represent the semantic similarity of P; P represents the item to be checked; sim(t i ) represents semantic feature items, including research object similarity, research content similarity, research attribute similarity and limiting factor similarity. Formula 1 indicates that when calculating the semantic similarity of a certain dimension, the semantic feature item ti under this dimension is weighted and summed with the similarity value of the comparison project. The specific weight ratio of each dimension can be customized at the system level according to actual needs. In this way, through the calculation of comprehensive semantic similarity, the duplicate checking system can more accurately judge the overlap correlation between the project to be checked and the historical scientific and technological projects, and generate corresponding duplicate checking results, providing strong support for the innovative management of scientific research projects.

[0093] The method provided in the embodiment of the present application realizes the analysis of the project to be checked at the semantic level through multi-dimensional semantic similarity calculation, so as to realize the duplication check operation of the project to be checked in combination with the analysis result in the subsequent process, thereby improving the accuracy and efficiency of the duplication check and providing a strong guarantee for the innovative management of scientific research projects.

[0094] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides a method for calculating text similarity, such as Figure 5 As shown, the text similarity between the project to be checked and each historical science and technology project is calculated by using the text duplicate checking image library, including:

[0095] 501. Based on the pre-trained knowledge meta-indexing model, multiple target metadata of the items to be searched are extracted.

[0096] In the embodiment of the present application, the duplicate checking system will use the pre-trained knowledge element indexing model to extract multiple key target metadata from the project to be checked. The multiple target metadata include but are not limited to the project's title, keywords, abstract, technical route, and sub-topics, etc., so as to effectively capture the core content and key features of the project, so as to generate text feature representation vectors in the future, and then calculate the text similarity through the distance between vectors, and perform text comparison and duplicate elimination.

[0097] 502. Based on a pre-trained vector indexing model, multiple target metadata are converted into vectors to obtain multiple target metadata vectors.

[0098] In an embodiment of the present application, after extracting multiple target metadata, the duplicate checking system will use a pre-trained vector indexing model to convert these target metadata into vector form, so that the target metadata is comparable in the vector space, thereby facilitating subsequent similarity calculations.

[0099] 503. Calculate the text similarity between the item to be checked and each historical science and technology item.

[0100] In the embodiment of the present application, after completing the vector conversion of the target metadata, the duplicate checking system will calculate the text similarity between the item to be checked and each historical science and technology item.

[0101] Specifically, for each historical science and technology project, the duplicate checking system will determine the corresponding designated project metadata vector for each target metadata vector based on the text duplicate checking portrait library, where the designated project metadata vector corresponding to the target metadata vector is also the project metadata vector with the same name as the target metadata vector. For example, the target metadata vector is a vector generated by the target metadata "title", and its corresponding designated project metadata vector is a designated project metadata vector generated by the project metadata "title" in multiple designated project metadata vectors corresponding to the historical science and technology project. Then, the duplicate checking system calculates the vector similarity between each target metadata vector and its corresponding designated project metadata vector to obtain multiple vector similarity values. Next, the duplicate checking system queries the vector weight associated with each target metadata vector, and uses these weights to perform a weighted average calculation on multiple vector similarities to obtain the final text similarity value.

[0102] The method provided in the embodiment of the present application can effectively extract the target metadata of the project to be checked and convert it into a comparable vector form, thereby accurately calculating the text similarity between the project to be checked and the historical scientific and technological projects, so as to subsequently combine the analysis results to implement the duplication check operation on the project to be checked, thereby improving the accuracy and efficiency of the duplication check and providing strong guarantees for the innovative management of scientific research projects.

[0103] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides a method for calculating the comprehensive similarity and generating a comprehensive duplicate checking result, such as Figure 6 As shown, the text similarity between the project to be checked and each historical science and technology project is calculated by using the text duplicate checking image library, including:

[0104] 601. Query a first similarity weight corresponding to semantic similarity and a second similarity weight corresponding to text similarity.

[0105] In an embodiment of the present application, the duplicate checking system will determine the weights of semantic similarity and text similarity in calculating the comprehensive similarity based on the actual needs of the item to be checked for duplicate checking. The weights reflect the importance of different levels of similarity in evaluating the originality of the project content.

[0106] Specifically, the duplicate checking system will determine the first similarity weight corresponding to semantic similarity and the second similarity weight corresponding to text similarity based on statistical analysis of experience and historical data. Among them, semantic similarity mainly measures the similarity of project content at abstract levels such as concepts and intentions, while text similarity focuses on the similarity of specific text expressions. Therefore, when setting weights, the duplicate checking system will comprehensively consider factors such as the project's professional field, research objectives, and innovations. For example, if a project to be checked focuses on theoretical research and conceptual innovation, the weight of semantic similarity may be higher; if a project to be checked focuses on experimental operations and specific implementations, the weight of text similarity may be more important. In this way, by setting weights reasonably and making personalized adjustments according to the characteristics of different projects, it can be ensured that the calculation of comprehensive similarity can fully reflect the originality of the project to be checked, thereby improving the accuracy and reliability of duplicate checking.

[0107] 602. Using the first similarity weight and the second similarity weight, perform weighted sum calculation on the semantic similarity and text similarity between the project to be checked and each historical science and technology project, and use the calculated value as the comprehensive similarity between the project to be checked and each historical science and technology project.

[0108] In the embodiment of the present application, after determining the weights of semantic similarity and textual similarity, the duplicate checking system will use these weights to perform a weighted sum calculation on the semantic similarity and textual similarity between the project to be checked and each historical science and technology project, thereby obtaining a comprehensive similarity.

[0109] Specifically, the duplicate checking system will multiply the semantic similarity and text similarity by the corresponding first similarity weight and second similarity weight respectively, and finally add the two weighted results to obtain the comprehensive similarity, thereby fully considering the similarity of project content at different levels, and realizing a comprehensive evaluation of project content at different levels and granularities. It can more comprehensively evaluate the originality of the project to be checked and improve the accuracy and reliability of duplicate checking.

[0110] 603. Extract at least one historical science and technology project from among multiple historical science and technology projects, whose comprehensive similarity with the project to be checked is greater than a comprehensive similarity threshold, as a similar project.

[0111] In the embodiment of the present application, after calculating the comprehensive similarity between the project to be checked and each historical science and technology project, the duplicate checking system will extract similar projects with a comprehensive similarity greater than a preset threshold as potential duplicate projects. For example, if the comprehensive similarity threshold is set to 0.8, all historical science and technology projects with a comprehensive similarity greater than 0.8 with the project to be checked will be regarded as similar projects, which will help to quickly screen out potential duplicate projects and facilitate subsequent comparison analysis.

[0112] 604. Generate a comprehensive duplicate checking result based on similar items and the comprehensive similarity of similar items.

[0113] In the embodiment of the present application, after similar items are extracted, the duplicate checking system will generate a comprehensive duplicate checking result based on the similar items and their comprehensive similarities. The comprehensive duplicate checking result will fully reflect the similarity between the item to be checked and the historical science and technology item, and support online viewing of the comparison results.

[0114] Specifically, the duplicate checking system will first generate a comprehensive duplicate checking report, which lists all similar items and their comprehensive similarities. Then, the duplicate checking system provides the function of viewing the comparison results online. Users can use the comparison tool to view the specific differences between the items to be checked and similar items. In addition, the duplicate checking system also supports batch comparison and mutual comparison functions to meet the needs of different users.

[0115] For example, a comprehensive duplicate checking report submitted for a scientific research project can show that the project has a high degree of comprehensive similarity with two historical science and technology projects. Users can view the specific differences between these projects through online comparison tools to further determine whether there is plagiarism or repeated application. It provides users with comprehensive and detailed duplicate checking results and comparison tools, which helps users quickly understand the originality of the project and make reasonable decisions; at the same time, the function of supporting online viewing and batch comparison also improves the efficiency and convenience of duplicate checking.

[0116] The method provided in the embodiment of the present application achieves a comprehensive evaluation of the originality of the content of the project to be checked and improves the accuracy and efficiency of the duplicate checking through the steps of reasonably setting weights, calculating comprehensive similarities, extracting similar projects and generating comprehensive duplicate checking results.

[0117] In summary, the logical process of the project research content duplication detection method based on semantic alignment proposed in this application is summarized as follows: Figure 7 First, construct a semantic plagiarism detection portrait library and a text plagiarism detection portrait library. When it is necessary to check the plagiarism of the project to be checked, the project to be checked is checked at the semantic level based on the semantic plagiarism detection portrait library to generate a semantic plagiarism detection result; at the same time, the project to be checked is checked at the text level based on the text plagiarism detection portrait library to generate a text plagiarism detection result. Next, according to the customized weight, the two plagiarism detection results are merged and a project plagiarism detection report is generated. Finally, a full-text comparison is performed between similar projects to further optimize the plagiarism detection results and generate corresponding comparison results.

[0118] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a project research content duplication detection device based on semantic alignment, such as Figure 8As shown, the device includes: a determination module 801, a semantic similarity calculation module 802, a text similarity calculation module 803, a comprehensive similarity calculation module 804 and an output module 805.

[0119] The determination module 801 is used to respond to a scientific and technological project duplicate checking request and determine the project to be checked indicated by the scientific and technological project duplicate checking request;

[0120] The semantic similarity calculation module 802 is used to use the semantic duplicate checking image library to semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects. The semantic duplicate checking image library is constructed after semantic element mining of the multiple historical science and technology projects.

[0121] The text similarity calculation module 803 is used to calculate the text similarity between the item to be checked and each of the historical science and technology items by using the text duplicate checking image library, wherein the text duplicate checking image library is constructed after metadata mining of the multiple historical science and technology items;

[0122] The comprehensive similarity calculation module 804 is used to calculate the comprehensive similarity between the item to be checked and each of the historical science and technology items according to the semantic similarity and the text similarity, and to generate a comprehensive duplicate checking result;

[0123] The output module 805 is used to compare and analyze the item to be checked with the similar items indicated by the comprehensive duplication check result, obtain a comparison report and output it.

[0124] In a specific application scenario, the device also includes:

[0125] The module for constructing a semantic duplicate checking portrait library is used to obtain the multiple historical scientific and technological projects, and based on a pre-trained semantic indexing model, extract multiple project semantic elements of each of the historical scientific and technological projects, wherein the multiple project semantic elements include but are not limited to the research object, research content, research attributes and limiting factors of the historical scientific and technological projects; based on a pre-trained retrieval indexing model, label the multiple project semantic elements of each of the historical scientific and technological projects, and perform vector conversion on each of the labeled project semantic elements to obtain multiple semantic element vectors corresponding to each of the historical scientific and technological projects; and use the multiple semantic element vectors corresponding to each of the historical scientific and technological projects to construct the semantic duplicate checking portrait library.

[0126] In a specific application scenario, the device also includes:

[0127] The module for constructing a text duplicate detection portrait library is used to obtain the multiple historical scientific and technological projects, and based on a pre-trained knowledge element indexing model, extract multiple project metadata of each of the historical scientific and technological projects, the project metadata includes but is not limited to the project title, keywords, abstract, technical route, and sub-topics of the historical scientific and technological projects; based on a pre-trained vector indexing model, perform vector conversion on the multiple project metadata of each of the historical scientific and technological projects to obtain multiple project metadata vectors corresponding to each of the historical scientific and technological projects; and use the multiple project metadata vectors corresponding to each of the historical scientific and technological projects to construct the text duplicate detection portrait library.

[0128] In a specific application scenario, the semantic similarity calculation module 802 is used to extract multiple target semantic elements of the project to be checked from the project content of the project to be checked, and the multiple target semantic elements include but are not limited to the target research object, target research content, target research attribute and target limiting factor of the project to be checked; based on the pre-trained retrieval indexing model, label the multiple target semantic elements, and perform vector conversion on each of the labeled target semantic elements to obtain the target semantic element vector corresponding to each target semantic element; extract the first target semantic element used to indicate the target research object from the multiple target semantic elements, and obtain the vector corresponding to the first target semantic element. A first target semantic element vector is used to calculate the similarity between the first target semantic element vector and the semantic element vector indicating the research object in the semantic element vector corresponding to each of the historical science and technology projects based on the semantic duplicate detection portrait library, so as to obtain the research object similarity between the project to be checked and each of the historical science and technology projects; when the research object similarity between the project to be checked and at least one historical science and technology project is greater than or equal to a first similarity threshold, a second target semantic element indicating the target research content is extracted from the multiple target semantic elements, and a second target semantic element vector corresponding to the second target semantic element is obtained, and the second target semantic element vector corresponding to the second target semantic element is calculated based on the semantic duplicate detection portrait library. The similarity between the semantic element vector indicating the research content in the semantic element vector corresponding to each of the historical science and technology projects is measured to obtain the research content similarity between the project to be checked and each of the historical science and technology projects; when the research content similarity between the project to be checked and at least one historical science and technology project is greater than or equal to a second similarity threshold, a third target semantic element indicating the target research attribute is extracted from the multiple target semantic elements, and a third target semantic element vector corresponding to the third target semantic element is obtained; based on the semantic duplicate checking portrait library, the semantic element vector indicating the research attribute in the semantic element vector corresponding to each of the historical science and technology projects is calculated. The similarity between the quantities is obtained to obtain the research attribute similarity between the item to be checked and each of the historical science and technology items; when the research attribute similarity between the item to be checked and each of the historical science and technology items is less than the third similarity threshold, a fourth target semantic element for indicating the target limiting factor is extracted from the multiple target semantic elements, and a fourth target semantic element vector corresponding to the fourth target semantic element is obtained, and based on the semantic duplicate checking portrait library, the similarity between the fourth target semantic element vector and the semantic element vector corresponding to each of the historical science and technology items for indicating the semantic element vector of the limiting factor is calculated to obtain the limiting factor similarity between the item to be checked and each of the historical science and technology items;Calculate the semantic similarity between the project to be checked and each of the historical science and technology projects based on the similarity of research objects, research contents, research attributes and limiting factors between the project to be checked and each of the historical science and technology projects. ;

[0129] In a specific application scenario, the semantic similarity 802 calculation module is also used to determine, when the research object similarity between the project to be checked and each of the historical science and technology projects is less than the first similarity threshold, that there is no project among the multiple historical science and technology projects that is semantically duplicated with the project to be checked, generate duplicate checking feedback indicating that there is no semantically duplicated project, and output the duplicate checking feedback; or, when the research content similarity between the project to be checked and each of the historical science and technology projects is less than the second similarity threshold, determine, when the multiple historical science and technology projects do not have projects that are semantically duplicated with the project to be checked, generate duplicate checking feedback indicating that there is no semantically duplicated project, and output the duplicate checking feedback; or, when the research attribute similarity between the project to be checked and at least one specified historical science and technology project is greater than or equal to the third similarity threshold, treat the at least one specified historical science and technology project as a duplicate project that is semantically duplicated with the project to be checked, generate abnormal feedback carrying the at least one specified historical science and technology project, and output the abnormal feedback.

[0130] In a specific application scenario, the text similarity calculation module 803 is used to extract multiple target metadata of the project to be checked based on the pre-trained knowledge element indexing model, and the multiple target metadata include but are not limited to the project title, keywords, abstracts, technical routes, and sub-topics of the project to be checked; based on the pre-trained vector indexing model, the multiple target metadata are vector-converted to obtain multiple target metadata vectors; the following processing is performed on each of the historical science and technology projects: based on the text duplication detection portrait library, among the multiple project metadata vectors corresponding to the historical science and technology projects, the corresponding designated project metadata vector is determined for each of the target metadata vectors, the vector similarity between each of the target metadata vectors and its corresponding designated project metadata vector is calculated, and multiple vector similarities corresponding to the multiple target metadata vectors are obtained; the vector weight associated with each of the target metadata vectors is queried, and the multiple vector similarities are calculated using the vector weights, and the calculated similarities are used as the text similarity between the historical science and technology project and the project to be checked.

[0131] In a specific application scenario, the comprehensive similarity calculation module 804 is used to query the first similarity weight corresponding to the semantic similarity and the second similarity weight corresponding to the text similarity; using the first similarity weight and the second similarity weight, a weighted sum calculation is performed on the semantic similarity and the text similarity between the item to be checked and each of the historical and technological items, and the calculated value is used as the comprehensive similarity between the item to be checked and each of the historical and technological items; at least one historical and technological project whose comprehensive similarity with the item to be checked is greater than a comprehensive similarity threshold is extracted from the multiple historical and technological projects as the similar project; based on the similar projects and the comprehensive similarities of the similar projects, the comprehensive duplicate checking result is generated.

[0132] The device provided in the embodiment of the present application responds to a request for checking for duplicate technology projects, determines the project to be checked indicated by the request for checking for duplicate technology projects, uses a semantic checking for duplicate technology portrait library to semantically align the project content of the project to be checked with the project content of multiple historical technology projects step by step, obtains the semantic similarity between the project to be checked and each historical technology project, uses a text checking for duplicate technology portrait library to calculate the text similarity between the project to be checked and each historical technology project, calculates the comprehensive similarity between the project to be checked and each historical technology project based on the semantic similarity and the text similarity, and generates a comprehensive checking result, aligning the project to be checked with the project content indicated by the comprehensive checking for duplicate technology results. Similar projects are compared and analyzed, and a comparison report is obtained and output. By introducing the semantic duplicate detection portrait library and the text duplicate detection portrait library, the step-by-step semantic alignment and text similarity calculation of the project content between the project to be checked and the historical scientific and technological projects are achieved, thereby significantly improving the accuracy of duplicate detection, reducing the workload of manual secondary screening and review, reducing time costs, and improving duplicate detection efficiency; at the same time, it can comprehensively evaluate the overall and local similarities between the project to be checked and multiple historical scientific and technological projects, accurately identify the repeated research of sub-topics, effectively overcome the limitations of traditional duplicate detection schemes in duplicate detection efficiency and accuracy, and better adapt to the duplicate detection needs of scientific and technological projects.

[0133] It should be noted that for other corresponding descriptions of the functional units involved in the project research content duplication detection device based on semantic alignment provided in the embodiment of the present application, please refer to Figures 1 to 7 The corresponding description in will not be repeated here.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0135] The above embodiments and the technical features in the embodiments can be arbitrarily combined with each other. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0136] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

[0137] In an exemplary embodiment, see Fig. 9 , and also provides a computer device, which includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device, wherein each functional unit can communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory and execute the project research content duplication detection method based on semantic alignment in the above embodiment.

[0138] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the project research content duplication checking method based on semantic alignment.

[0139] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0140] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present application.

[0141] Those skilled in the art will appreciate that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0142] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0143] The above disclosure only discloses several specific implementation scenarios of the present application. However, the present application is not limited thereto, and any changes that can be conceived by technicians in this field should fall within the scope of protection of the present application.

Claims

1. A method for checking duplicate content of project research based on semantic alignment, characterized in that: include: In response to a request for checking for duplicate technology projects, determining the project to be checked indicated by the request for checking for duplicate technology projects; Using a semantic duplicate checking image library, semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects. The semantic duplicate checking image library is constructed after semantic element mining of the multiple historical science and technology projects. Calculate the text similarity between the item to be checked and each of the historical science and technology items by using a text duplicate checking image library, wherein the text duplicate checking image library is constructed after metadata mining of the multiple historical science and technology items; Calculate the comprehensive similarity between the item to be checked and each of the historical science and technology items based on the semantic similarity and the text similarity, and generate a comprehensive duplicate checking result; The items to be checked are compared and analyzed with similar items indicated by the comprehensive duplicate checking results to obtain and output a comparison report.

2. The method according to claim 1, characterized in that Before responding to the scientific and technological project duplicate checking request and determining the project to be checked indicated by the scientific and technological project duplicate checking request, the method further includes: Acquire the plurality of historical science and technology projects, and extract a plurality of project semantic elements of each of the historical science and technology projects based on a pre-trained semantic indexing model, wherein the plurality of project semantic elements include but are not limited to the research object, research content, research attribute and limiting factor of the historical science and technology project; Based on the pre-trained retrieval indexing model, label multiple project semantic elements of each of the historical science and technology projects, and perform vector conversion on each of the labeled project semantic elements to obtain multiple semantic element vectors corresponding to each of the historical science and technology projects; The semantic duplicate checking image library is constructed by utilizing the multiple semantic element vectors corresponding to each of the historical science and technology projects.

3. The method according to claim 1, characterized in that Before responding to the scientific and technological project duplicate checking request and determining the project to be checked indicated by the scientific and technological project duplicate checking request, the method further includes: Acquire the plurality of historical science and technology projects, and extract a plurality of project metadata of each of the historical science and technology projects based on a pre-trained knowledge meta-indexing model, wherein the project metadata includes but is not limited to a project title, keywords, abstract, technical route, and sub-topic of the historical science and technology project; Based on a pre-trained vector indexing model, vector conversion is performed on multiple project metadata of each of the historical science and technology projects to obtain multiple project metadata vectors corresponding to each of the historical science and technology projects; The text duplication detection portrait library is constructed by utilizing multiple project metadata vectors corresponding to each of the historical science and technology projects.

4. The method according to claim 2, characterized in that: The semantic duplication checking image library is used to semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects, including: Extracting multiple target semantic elements of the project to be checked from the project content of the project to be checked, wherein the multiple target semantic elements include but are not limited to the target research object, target research content, target research attribute and target limiting factor of the project to be checked; Based on the pre-trained retrieval indexing model, label the multiple target semantic elements, and perform vector conversion on each of the labeled target semantic elements to obtain a target semantic element vector corresponding to each of the target semantic elements; Extracting a first target semantic element for indicating the target research object from the multiple target semantic elements, and obtaining a first target semantic element vector corresponding to the first target semantic element, and calculating the similarity between the first target semantic element vector and the semantic element vector for indicating the research object in the semantic element vector corresponding to each of the historical science and technology projects based on the semantic duplicate checking portrait library, to obtain the research object similarity between the project to be checked and each of the historical science and technology projects; In the case where the research object similarity between the project to be checked and at least one historical science and technology project is greater than or equal to a first similarity threshold, a second target semantic element for indicating the target research content is extracted from the multiple target semantic elements, and a second target semantic element vector corresponding to the second target semantic element is obtained, and based on the semantic duplicate checking portrait library, the similarity between the second target semantic element vector and the semantic element vector for indicating the research content in the semantic element vector corresponding to each of the historical science and technology projects is calculated to obtain the research content similarity between the project to be checked and each of the historical science and technology projects; In the case where the research content similarity between the project to be checked and at least one historical science and technology project is greater than or equal to a second similarity threshold, a third target semantic element for indicating the target research attribute is extracted from the multiple target semantic elements, and a third target semantic element vector corresponding to the third target semantic element is obtained, and based on the semantic duplicate checking portrait library, the similarity between the third target semantic element vector and the semantic element vector for indicating the research attribute in the semantic element vector corresponding to each of the historical science and technology projects is calculated to obtain the research attribute similarity between the project to be checked and each of the historical science and technology projects; In the case where the research attribute similarity between the project to be checked and each of the historical science and technology projects is less than the third similarity threshold, a fourth target semantic element for indicating the target limiting factor is extracted from the multiple target semantic elements, and a fourth target semantic element vector corresponding to the fourth target semantic element is obtained, and based on the semantic duplicate checking portrait library, the similarity between the fourth target semantic element vector and the semantic element vector corresponding to each of the historical science and technology projects for indicating the semantic element vector of the limiting factor is calculated to obtain the limiting factor similarity between the project to be checked and each of the historical science and technology projects; The semantic similarity between the project to be checked and each of the historical science and technology projects is calculated based on the research object similarity, research content similarity, research attribute similarity and limiting factor similarity between the project to be checked and each of the historical science and technology projects.

5. The method according to claim 4, characterized in that The method further comprises: In the case where the research object similarity between the project to be checked and each of the historical science and technology projects is less than the first similarity threshold, determining that there is no project in the multiple historical science and technology projects that is semantically duplicated with the project to be checked, generating duplicate checking feedback indicating that there is no semantically duplicated project, and outputting the duplicate checking feedback; or, In the case where the research content similarity between the project to be checked and each of the historical science and technology projects is less than the second similarity threshold, determining that there is no project in the multiple historical science and technology projects that is semantically duplicated with the project to be checked, generating duplicate checking feedback indicating that there is no semantically duplicated project, and outputting the duplicate checking feedback; or, When the research attribute similarity between the project to be checked and at least one designated historical scientific and technological project is greater than or equal to the third similarity threshold, the at least one designated historical scientific and technological project is treated as a duplicate project that is semantically duplicated with the project to be checked, and abnormal feedback carrying the at least one designated historical scientific and technological project is generated and output.

6. The method according to claim 3, characterized in that The method of using the text duplication checking image library to calculate the text similarity between the item to be checked and each of the historical science and technology items includes: Based on the pre-trained knowledge element indexing model, extract multiple target metadata of the project to be checked, the multiple target metadata include but are not limited to the project title, keywords, abstract, technical route, and sub-topic of the project to be checked; Based on the pre-trained vector indexing model, the plurality of target metadata are vector-converted to obtain a plurality of target metadata vectors; The following processing is performed on each of the historical and technological projects: based on the text duplication detection portrait library, among the multiple project metadata vectors corresponding to the historical and technological projects, a corresponding designated project metadata vector is determined for each of the target metadata vectors, the vector similarity between each of the target metadata vectors and its corresponding designated project metadata vector is calculated, and multiple vector similarities corresponding to the multiple target metadata vectors are obtained, and the vector weight associated with each of the target metadata vectors is queried, and the multiple vector similarities are calculated using the vector weights, and the calculated similarities are used as the text similarity between the historical and technological project and the project to be checked.

7. The method according to claim 1, characterized in that The step of calculating the comprehensive similarity between the item to be checked and each of the historical science and technology items based on the semantic similarity and the text similarity, and generating a comprehensive duplicate checking result, includes: Querying a first similarity weight corresponding to the semantic similarity and a second similarity weight corresponding to the text similarity; Using the first similarity weight and the second similarity weight, a weighted sum calculation is performed on the semantic similarity and the text similarity between the item to be checked and each of the historical science and technology items, and the calculated value is used as the comprehensive similarity between the item to be checked and each of the historical science and technology items; Extracting at least one historical science and technology project from the plurality of historical science and technology projects, whose comprehensive similarity with the project to be checked is greater than a comprehensive similarity threshold, as the similar project; Based on the similar items and the comprehensive similarities of the similar items, the comprehensive duplication checking result is generated.

8. A device for checking duplicate content of project research based on semantic alignment, characterized in that: include: A determination module, configured to respond to a request for checking for duplicate technology projects and determine the items to be checked indicated by the request for checking for duplicate technology projects; A semantic similarity calculation module is used to use a semantic duplicate checking image library to semantically align the project content of the project to be checked with the project content of multiple historical science and technology projects step by step to obtain the semantic similarity between the project to be checked and each of the historical science and technology projects, wherein the semantic duplicate checking image library is constructed after semantic element mining of the multiple historical science and technology projects; A text similarity calculation module, used to calculate the text similarity between the item to be checked and each of the historical science and technology items by using a text duplicate checking image library, wherein the text duplicate checking image library is constructed after metadata mining of the multiple historical science and technology items; A comprehensive similarity calculation module, used to calculate the comprehensive similarity between the item to be checked and each of the historical science and technology items according to the semantic similarity and the text similarity, and to generate a comprehensive duplicate checking result; The output module is used to compare and analyze the items to be checked with similar items indicated by the comprehensive duplication check results, obtain a comparison report and output it.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images