Scientific and technological novelty retrieval method and device based on semantic elements, equipment and medium

Abstract

The invention discloses a scientific and technological novelty retrieval method, device and equipment based on semantic elements and a medium, relates to the technical field of text retrieval, combines novelty retrieval semantic vectors and semantic retrieval formulas, captures deep semantics of target novelty retrieval points, ensures that literatures highly related to the target novelty retrieval points are screened out, and meanwhile, through the semantic analysis and retrieval technology, the target novelty retrieval points are obtained. The operation difficulty and the time cost are reduced. The method comprises the steps that when a novelty search application form is received, a comprehensive novelty search point is generated according to content to be subjected to novelty search and retrieval indicated by the novelty search application form, and a novelty search semantic vector of the comprehensive novelty search point is obtained; based on the novelty search application form and the comprehensive novelty search points, extracting target semantic elements; expanding the target semantic elements to obtain a semantic set corresponding to the target semantic elements, and constructing a semantic retrieval formula by utilizing the semantic set; and based on the novelty search semantic vector and the semantic retrieval formula, executing novelty search operation to obtain a plurality of target literatures, sorting the plurality of target literatures, and outputting the sorted target literatures.

Description

Technical Field

[0001] The present application relates to the field of text retrieval technology, and in particular to a technology novelty retrieval method, device, equipment and medium based on semantic elements. Background Art

[0002] In today's era, scientific and technological information is growing at an explosive rate. Scientific and technological novelty search has become an indispensable key link in the process of scientific research workers in the process of scientific and technological project establishment, achievement evaluation, award application and scientific research management. The core purpose of scientific and technological novelty search is to ensure the innovation of scientific research projects, avoid duplication of research, and provide objective basis for the evaluation of scientific research results through comprehensive retrieval and analysis of relevant scientific and technological literature.

[0003] In the related technologies, scientific and technological novelty retrieval mainly relies on manually constructed search formulas. Constructing search formulas requires users to not only have profound professional knowledge, but also to accurately select search terms in the hope of fully capturing all relevant scientific and technological literature; alternatively, some platforms provide natural language retrieval functions, which can parse the user's natural language, try to understand its semantics and perform scientific and technological novelty retrieval based on the semantics.

[0004] However, the applicants realize that due to the diversity and high complexity of scientific and technological literature, it is often difficult for users to comprehensively and accurately list all relevant search terms, and it is difficult to construct search formulas and the constructed search formulas are not perfect. The natural language search function usually lacks special optimization for the field of scientific and technological novelty search. Many relevant documents are omitted due to inaccurate search formulas or inadequate natural language understanding, resulting in low recall and accuracy of search results and low retrieval efficiency. Summary of the invention

[0005] In view of this, the present application provides a method, device, equipment and medium for scientific and technological novelty retrieval based on semantic elements. The main purpose is to solve the current difficulties in constructing retrieval formulas and the problems that the constructed retrieval formulas are not perfect, and the natural language retrieval function usually lacks special optimization for the field of scientific and technological novelty retrieval. Many relevant documents are omitted due to inaccurate retrieval formulas or insufficient natural language understanding, resulting in low recall and accuracy of retrieval results and low retrieval efficiency.

[0006] According to the first aspect of the present application, a technology novelty search method based on semantic elements is provided, the method comprising:

[0007] When a novelty search application form is received, a comprehensive novelty search point is generated according to the content to be searched for novelty as indicated in the novelty search application form, and a novelty search semantic vector of the comprehensive novelty search point is obtained;

[0008] Extracting target semantic elements based on the novelty search application form and the comprehensive novelty search points;

[0009] Expanding the target semantic element to obtain a semantic set corresponding to the target semantic element, and constructing a semantic search formula using the semantic set;

[0010] Based on the novelty search semantic vector and the semantic search formula, a novelty search operation is performed to obtain a plurality of target documents, and the plurality of target documents are sorted and then output.

[0011] Optionally, generating a comprehensive novelty search point according to the content to be searched for novelty indicated by the novelty search application form, and obtaining a novelty search semantic vector of the comprehensive novelty search point includes:

[0012] Read the novelty search application form, and extract the target novelty search point, title and technical points to be retrieved from the novelty search application form;

[0013] Using a pre-trained scientific and technological literature search model, extract a designated information segment from the title and the technical points, wherein the similarity between the designated information segment and the target search point is greater than a first similarity threshold and the designated information segment has the same technical theme as the target search point;

[0014] Integrate the designated information fragment with the target novelty search point to obtain the comprehensive novelty search point;

[0015] Extracting at least one novelty search keyword set by a user from the novelty search application form, and splicing the comprehensive novelty search point with the at least one novelty search keyword;

[0016] The pre-trained technology novelty search vector model is used to perform vector conversion on the spliced ​​comprehensive novelty search point and the at least one novelty search keyword to obtain the novelty search semantic vector.

[0017] Optionally, extracting target semantic elements based on the novelty search application form and the comprehensive novelty search points includes:

[0018] Extracting target novelty points to be searched from the novelty search application form, and extracting research object information and technical subject information of the novelty search application form according to the target novelty search points;

[0019] Extracting technical feature information based on the comprehensive novelty search points;

[0020] The extracted research object information, technical subject information and technical feature information are used as the target semantic elements.

[0021] Optionally, the expanding the target semantic element to obtain a semantic set corresponding to the target semantic element, and constructing a semantic search formula using the semantic set, includes:

[0022] Based on the pre-built novelty search semantic library, the target semantic element is searched, and the target semantic element is expanded according to the search result to obtain the semantic set;

[0023] The semantic set is converted into the semantic search formula through the pre-constructed three-segment rule.

[0024] Optionally, the target semantic element is searched based on the pre-built novelty search semantic library, and the target semantic element is expanded according to the search result to obtain the semantic set, including:

[0025] Based on the novelty search semantic library, searching for related words associated with the target semantic element, wherein the related words include one or more of standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words associated with the target semantic element;

[0026] If related words of the target semantic element are retrieved, the retrieved related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set;

[0027] If the related words of the target semantic element are not retrieved, the related words are extracted from the research object, technical theme and technical characteristics indicated by the target semantic element based on the pre-trained scientific and technological literature retrieval model, and the extracted related words are used as extended content of the target semantic element. The retrieved related words are integrated with the target semantic element to obtain the semantic set.

[0028] Optionally, the novelty search operation is performed based on the novelty search semantic vector and the semantic search formula to obtain a plurality of target documents, including:

[0029] Constructing a structured query statement based on the novelty search semantic vector and the semantic search formula;

[0030] Executing the structured query statement to obtain multiple candidate documents;

[0031] The plurality of candidate documents are filtered according to a preset filtering condition, and the plurality of candidate documents remaining after filtering are used as the plurality of target documents.

[0032] Optionally, the outputting the plurality of target documents after sorting includes:

[0033] Read the novelty search application form, extract the target novelty search point to be retrieved from the novelty search application form, and use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the target novelty search point to obtain a target semantic vector;

[0034] Segmenting each of the target documents to obtain multiple text segments;

[0035] Using the pre-trained technology search vector model, the plurality of text segments are vectorized to obtain a plurality of segment semantic vectors;

[0036] Batch calculating the vector similarity between the target semantic vector and each of the multiple segment semantic vectors;

[0037] The segment semantic vectors belonging to the same target document are divided into the same vector group to obtain a vector group corresponding to each target document, and the specified segment semantic vector with the highest vector similarity is extracted from the vector group corresponding to each target document to obtain a specified semantic segment vector corresponding to each target document;

[0038] According to the vector similarity of the specified semantic segment vector corresponding to each of the target documents, the multiple specified semantic segment vectors currently obtained are sorted to obtain a vector sorting result, and according to the vector sorting result, the multiple target documents are sorted, and the sorted multiple target documents are output.

[0039] According to the second aspect of the present application, a technology novelty search device based on semantic elements is provided, the device comprising:

[0040] A generating module, for generating a comprehensive novelty search point according to the content to be searched indicated by the novelty search application form when receiving the novelty search application form, and obtaining the novelty search semantic vector of the comprehensive novelty search point;

[0041] An extraction module, used for extracting target semantic elements based on the novelty search application form and the comprehensive novelty search point;

[0042] A construction module, used to expand the target semantic element, obtain a semantic set corresponding to the target semantic element, and construct a semantic search formula using the semantic set;

[0043] The retrieval module is used to perform a novelty search operation based on the novelty search semantic vector and the semantic search formula to obtain multiple target documents, and to sort the multiple target documents and then output them.

[0044] Optionally, the generation module is used to read the novelty search application form, extract the target novelty search point, title and technical points to be retrieved from the novelty search application form; use a pre-trained scientific and technological literature novelty search large model to extract a specified information segment from the title and the technical points, the similarity between the specified information segment and the target novelty search point is greater than a first similarity threshold and has the same technical subject as the target novelty search point; integrate the specified information segment with the target novelty search point to obtain the comprehensive novelty search point; extract at least one novelty search keyword set by the user from the novelty search application form, and splice the comprehensive novelty search point with the at least one novelty search keyword; use a pre-trained scientific and technological novelty search vector model to perform vector conversion on the spliced ​​comprehensive novelty search point and the at least one novelty search keyword to obtain the novelty search semantic vector.

[0045] Optionally, the extraction module is used to extract the target novelty search points to be retrieved in the novelty search application form, and extract the research object information and technical subject information of the novelty search application form according to the target novelty search points; extract the technical feature information according to the comprehensive novelty search points; and use the extracted research object information, the technical subject information and the technical feature information as the target semantic elements.

[0046] Optionally, the construction module is used to retrieve the target semantic elements based on a pre-constructed novelty search semantic library, and to expand the target semantic elements according to the search results to obtain the semantic set; and to convert the semantic set into the semantic search formula through a pre-constructed three-segment rule.

[0047] Optionally, the construction module is used to retrieve related words that are associated with the target semantic element based on the novelty search semantic library, wherein the related words include one or more of standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words that are associated with the target semantic element; if the related words of the target semantic element are retrieved, the retrieved related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set; if the related words of the target semantic element are not retrieved, the related words are extracted from the research objects, technical topics, and technical features indicated by the target semantic element based on a pre-trained scientific and technological literature novelty search model, the extracted related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set.

[0048] Optionally, the retrieval module is used to construct a structured query statement based on the novelty search semantic vector and the semantic retrieval formula; execute the structured query statement to obtain multiple candidate documents; filter the multiple candidate documents according to preset filtering conditions, and use the multiple candidate documents remaining after filtering as the multiple target documents.

[0049] Optionally, the retrieval module is used to read the novelty search application form, extract the target novelty search points to be searched from the novelty search application form, and use a pre-trained scientific and technological novelty search vector model to perform vector conversion on the target novelty search points to obtain a target semantic vector; segment each of the target documents to obtain multiple text segments; use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the multiple text segments to obtain multiple segment semantic vectors; batch calculate the vector similarity between the target semantic vector and each segment semantic vector in the multiple segment semantic vectors; divide the segment semantic vectors belonging to the same target document into the same vector group to obtain a vector group corresponding to each target document, and extract the specified segment semantic vector with the highest vector similarity from the vector group corresponding to each target document to obtain a specified semantic segment vector corresponding to each target document; sort the multiple specified semantic segment vectors currently obtained according to the vector similarity of the specified semantic segment vector corresponding to each target document to obtain a vector sorting result, and sort the multiple target documents according to the vector sorting result, and output the sorted multiple target documents.

[0050] According to a third aspect of the present application, a device is provided, including 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.

[0051] According to a fourth aspect of the present application, a medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0052] By means of the above technical scheme, the present application provides a technology novelty search and retrieval method, device, equipment and medium based on semantic elements. When receiving a novelty search application form, the present application generates a comprehensive novelty search point according to the content to be searched indicated in the novelty search application form, and obtains the novelty search semantic vector of the comprehensive novelty search point. Based on the novelty search application form and the comprehensive novelty search point, the target semantic element is extracted, the target semantic element is expanded, and the semantic set corresponding to the target semantic element is obtained. The semantic set is used to construct a semantic search formula. Based on the novelty search semantic vector and the semantic search formula, a novelty search operation is performed to obtain multiple target documents, and the multiple target documents are sorted and output. By intelligently processing the information in the novelty search application form, obtaining the novelty search semantic vector, and automatically generating the novelty search retrieval formula, the novelty search semantic vector and the semantic retrieval formula are combined for retrieval to capture the deep semantics of the target novelty search point, ensuring that the documents highly relevant to the target novelty search point are quickly screened out from the massive document resources. At the same time, through intelligent semantic analysis and retrieval technology, the user's operation difficulty and time cost are reduced. Users do not need to have excessive professional knowledge to easily conduct scientific and technological novelty searches and obtain satisfactory search results, which significantly improves the accuracy, recall rate and retrieval efficiency of the retrieval results, allowing users to obtain comprehensive and accurate novelty search results more quickly.

[0053] 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

[0054] 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:

[0055] Figure 1 A schematic diagram of a technology novelty search method based on semantic elements provided in an embodiment of the present application is shown;

[0056] Figure 2 A schematic diagram of a method flow for generating a comprehensive novelty search point and obtaining a novelty search semantic vector provided by an embodiment of the present application is shown;

[0057] Figure 3 A schematic diagram of a method for extracting target semantic elements provided in an embodiment of the present application is shown;

[0058] Figure 4A schematic diagram of a method flow for determining a semantic set and constructing a semantic search formula provided in an embodiment of the present application is shown;

[0059] Figure 5 A schematic diagram of a method for searching a target document provided in an embodiment of the present application is shown;

[0060] Figure 6 A schematic diagram of a method for sorting target documents provided in an embodiment of the present application is shown;

[0061] Figure 7 A schematic diagram of the process flow of another technology novelty search method based on semantic elements provided in an embodiment of the present application is shown;

[0062] Figure 8 A schematic diagram of the structure of a technology novelty search device based on semantic elements provided in an embodiment of the present application is shown;

[0063] 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

[0064] 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.

[0065] The present application embodiment provides a technology novelty search method based on semantic elements, such as Figure 1 As shown, the method includes:

[0066] 101. When a novelty search application form is received, a comprehensive novelty search point is generated according to the content to be searched indicated in the novelty search application form, and a novelty search semantic vector of the comprehensive novelty search point is obtained.

[0067] The embodiments of the present application can be applied to a scientific and technological novelty search and retrieval system, which aims to improve the efficiency and accuracy of scientific and technological novelty search work. It intelligently processes the information in the novelty search application form, obtains the novelty search semantic vector, and automatically generates the novelty search retrieval formula. It combines the novelty search semantic vector and the semantic retrieval formula to perform retrieval, quickly screens out documents that are highly relevant to the target novelty search point from massive document resources, provides users with a sorted list of target documents, and assists users in novelty evaluation and decision-making for scientific and technological projects.

[0068] When receiving a novelty search application form, the science and technology novelty search retrieval system will generate a comprehensive novelty search point and obtain the novelty search semantic vector of the comprehensive novelty search point according to the content to be searched indicated in the novelty search application form. Among them, the science and technology novelty search retrieval system first receives the novelty search application form and parses out the target novelty search point, title and technical points. Subsequently, the science and technology novelty search retrieval system expands this information, enriches the information contained in the novelty search point, makes it more comprehensive and in-depth, thereby minimizing the user's retrieval cost and improving retrieval efficiency. Based on the information expansion, the science and technology novelty search retrieval system further constructs a novelty search semantic vector, and converts the expanded novelty search point information into a high-dimensional numerical vector, namely, a novelty search semantic vector. The novelty search semantic vector not only contains the core information of the novelty search point, but also incorporates the extended information related to it, making the subsequent semantic search more accurate and efficient.

[0069] Through the above process, for the case where the target search points filled in by some users have incomplete information or are missing some important information, the science and technology search and retrieval system can automatically extract the corresponding information from the title and technical points to supplement it, and then vectorize it. In this way, in the subsequent search process, the science and technology search and retrieval system can more accurately identify the documents that are highly relevant to these search points, and provide users with more accurate search results.

[0070] 102. Extract target semantic elements based on the novelty search application form and comprehensive novelty search points.

[0071] In an embodiment of the present application, the science and technology novelty search retrieval system will extract the target semantic elements of the comprehensive novelty search point based on the novelty search application form and the comprehensive novelty search point. Among them, the science and technology novelty search retrieval system will extract the research object, technical theme, technical characteristics and other contents to constitute the target semantic elements. For example, if the novelty search application form indicates to conduct a novelty search on "artificial intelligence diabetes prediction model", and the comprehensive novelty search point involves "artificial intelligence-based diabetes prediction model", then the semantic elements that may be extracted by the science and technology novelty search retrieval system in the novelty search application form include "diabetes" (research object) and "artificial intelligence algorithm" (technical theme), and the semantic elements that may be extracted in the comprehensive novelty search point include "improved prediction accuracy" (technical characteristics), etc.

[0072] In this way, based on semantic analysis technology, it is possible not only to understand the literal meaning of the text, but also to capture deep semantic connections, thereby ensuring that the extracted semantic elements can accurately reflect the core content and intention of the novelty search application form and the comprehensive novelty search point; further, the extracted target semantic elements lay a solid foundation for generating accurate and efficient search formulas. By organizing the target semantic elements in a structured manner, the science and technology novelty search retrieval system can construct more specific and targeted search formulas, thereby significantly improving the accuracy and efficiency of novelty searches and providing users with more valuable novelty search results.

[0073] 103. Expand the target semantic element to obtain the semantic set corresponding to the target semantic element, and use the semantic set to construct a semantic search formula.

[0074] In the embodiment of the present application, the science and technology novelty search system takes the target semantic element as input, and based on the pre-constructed novelty search semantic library, efficiently and accurately searches for the standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, related words, etc. corresponding to the research object, technical theme, and technical characteristics in the target semantic element, and uses the retrieved words to expand the target semantic element to obtain the semantic set corresponding to the target semantic element, and uses the semantic set to construct a semantic search formula according to the pre-trained search formula data model. In this way, by expanding the target semantic element, not only the connotation of the target semantic element is enriched, but also the semantic understanding ability of the science and technology novelty search system for the search request is enhanced. By constructing a semantic set containing multiple types of vocabulary, the science and technology novelty search system can more accurately grasp the user's search intention and generate a more comprehensive and flexible semantic search formula, so that the semantic search formula not only contains the semantic elements directly input by the user, but also covers various vocabulary closely related to these elements, thereby greatly improving the accuracy and comprehensiveness of the search.

[0075] For example, if the target semantic element includes "deep learning algorithm", the science and technology novelty search retrieval system will retrieve and expand the standard words, synonyms and hypernyms such as "deep learning", "neural network", "machine learning algorithm", "artificial intelligence algorithm", and hyponyms such as "convolutional neural network" and "recurrent neural network". These expanded vocabulary sets will together constitute a semantic set and serve as an important basis for constructing semantic retrieval formulas. By using the semantic retrieval formula generated by such a semantic set, the science and technology novelty search retrieval system can more accurately locate scientific and technological literature that is highly relevant to the target novelty search point, thereby providing users with more accurate and comprehensive novelty search results.

[0076] 104. Based on the novelty search semantic vector and the semantic search formula, a novelty search operation is performed to obtain multiple target documents, and the multiple target documents are sorted and then output.

[0077] In the embodiment of the present application, the science and technology novelty search and retrieval system adopts a standard and vector fusion retrieval / recall strategy, which cleverly combines the novelty search semantic vector generated in the above process with the constructed semantic retrieval formula. Specifically, the science and technology novelty search and retrieval system will construct a structured query statement based on the novelty search semantic vector and the semantic retrieval formula, execute the structured query statement, and realize the recall of documents. Through the combination of these two methods, the science and technology novelty search and retrieval system can accurately and efficiently retrieve data that is highly relevant to the query requirements from the huge novelty search library, greatly enhancing the comprehensiveness and efficiency of the retrieval, overcoming the possible information omissions and false detection problems of traditional retrieval methods, and realizing comprehensive coverage and precise matching of data in the novelty search library by deeply mining the semantic characteristics and relevance of the data.

[0078] In addition, after retrieving relevant documents, the science and technology novelty search retrieval system introduces a re-ranking mechanism to conduct a secondary screening of the retrieved documents, obtain multiple target documents, sort the multiple target documents, and further optimize the priority order of the search results and novelty points to ensure that users can see the most relevant and valuable information first.

[0079] Through the above analysis, it can be seen that the method of the present application can be divided into five stages. First, in the first stage, the target search point is expanded based on the search application form to obtain a comprehensive search point, and a search semantic vector that accurately reflects the characteristics of the comprehensive search point is generated. Then, in the second stage, the target semantic elements are extracted based on the search application form and the expanded comprehensive search point. In the third stage, based on the constructed search semantic library, the target semantic elements are semantically expanded to form a semantic set, and a semantic search formula is constructed. Subsequently, in the fourth stage, multi-field label and vector fusion retrieval and filtering are performed based on the search semantic vector and the semantic search formula to obtain highly relevant document data to form a candidate list, which includes multiple target documents. Finally, in the fifth stage, based on the search vector model, the semantic relevance between the target search point and the candidate list is calculated, and the multiple target documents are sorted from high to low according to the relevance, and the final search list is generated. In this way, by introducing scalar and vector fusion retrieval technology, and combining it with a sorting method based on novelty semantic vectors, this application significantly improves the accuracy, recall rate and retrieval efficiency of retrieval results, allowing users to obtain comprehensive and accurate novelty results more quickly; at the same time, through intelligent semantic analysis and retrieval technology, it also reduces the user's operating difficulty and time cost. Users do not need to have excessive professional knowledge to easily perform scientific and technological novelty searches and obtain satisfactory retrieval results, which greatly optimizes the user experience and improves user satisfaction and loyalty.

[0080] In summary, compared with traditional novelty search and retrieval methods, the present application has significant advantages in the field of scientific and technological novelty search, including enhanced semantic understanding capabilities, improved retrieval accuracy and recall rate, and optimized user experience. These advantages make the present application have broad application prospects and market value in the field of scientific and technological novelty search. In addition, the technical solutions protected by the present application include methods for expanding novelty search point information and constructing semantic vectors, methods for extracting and expanding semantic elements of novelty search points to form semantic sets, methods for constructing semantic retrieval formulas based on semantic sets, retrieval / recall and filtering methods based on the fusion of semantic retrieval formulas and semantic vectors, and re-ranking methods based on novelty search points. In addition, the present application also protects scientific and technological novelty search and retrieval systems and devices that include the above-mentioned technical solutions, as well as computer program products for implementing the above-mentioned technical solutions.

[0081] The method provided in the embodiment of the present application generates a comprehensive novelty search point according to the content to be searched indicated in the novelty search application form, and obtains the novelty search semantic vector of the comprehensive novelty search point when receiving a novelty search application form; extracts the target semantic element based on the novelty search application form and the comprehensive novelty search point; expands the target semantic element to obtain the semantic set corresponding to the target semantic element; constructs a semantic search formula using the semantic set; performs a novelty search operation based on the novelty search semantic vector and the semantic search formula to obtain multiple target documents; and outputs the multiple target documents after sorting them, thereby intelligently processing the information in the novelty search application form. , obtain the novelty search semantic vector, and automatically generate the novelty search retrieval formula, combine the novelty search semantic vector and the semantic retrieval formula to perform retrieval, capture the deep semantics of the target novelty search point, and ensure that the documents highly relevant to the target novelty search point are quickly screened out from the massive document resources. At the same time, through intelligent semantic analysis and retrieval technology, the user's operation difficulty and time cost are reduced. Users do not need to have too much professional knowledge to easily perform scientific and technological novelty searches and obtain satisfactory search results, which significantly improves the accuracy, recall rate and retrieval efficiency of the retrieval results, allowing users to obtain comprehensive and accurate novelty search results more quickly.

[0082] 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 generating a comprehensive novelty search point and obtaining a novelty search semantic vector, such as Figure 2 As shown, that is, according to the content to be searched for novelty indicated in the novelty search application form, a comprehensive novelty search point is generated, and the novelty search semantic vector of the comprehensive novelty search point is obtained, including:

[0083] 1011. Read the novelty search application form, and extract the target novelty search points, titles and technical points to be searched from the novelty search application form.

[0084] In the process of scientific and technological novelty search, the novelty search application form is the starting point and key input, which describes in detail the specific content that the user wants to search for. In the embodiment of this application, the scientific and technological novelty search retrieval system will first read the novelty search application form, which contains the target novelty search point clearly indicated by the user, the title of the project, and detailed technical points. Among them, the target novelty search point is the point where the user wants to confirm whether the same or similar technology has been disclosed; the title is a brief summary of the entire project; and the technical points are a detailed description of the core technology of the project.

[0085] For example, assuming that a new type of solar cell is described in the novelty search application form, the target novelty search points include necessary technical topics such as "photoelectric conversion efficiency improvement technology" and other content that can reflect novelty, and the title is "Research and Development of High-efficiency Solar Cells". The technical points include "nanostructured light absorption layer", "multi-layer thin film structure design" and other content related to the background, purpose, effect, and technical solution. The science and technology novelty search system will accurately extract this information. Through automated extraction, it can reduce the error rate and time cost of manual operations, and ensure the accuracy and efficiency of subsequent processing steps.

[0086] 1012. Use the pre-trained scientific literature search model to extract specified information fragments from titles and technical points.

[0087] In the embodiment of the present application, the scientific and technological novelty search system will use the pre-trained scientific and technological literature novelty search big model to extract the specified information fragment from the title and technical key points. Among them, the similarity between the specified information fragment and the target novelty search point is greater than the first similarity threshold and has the same technical theme as the target novelty search point. When extracting the specified information fragment, the scientific and technological literature novelty search big model will identify the information fragment with similarity to the target novelty search point, and require this similarity to reach the preset first similarity threshold to ensure that the extracted information fragment is highly consistent with the content of the target novelty search point; at the same time, these information fragments must also have the same technical theme as the target novelty search point, so that the specified information fragment and the target novelty search point are consistent in terms of technical scope, application field or core problem solved.

[0088] Specifically, the scientific and technological novelty search and retrieval system will conduct an in-depth analysis of the title and technical points through the scientific and technological literature novelty search model to extract the key information and technical features therein; then, the extracted information will be compared with the target novelty search point one by one to calculate the similarity between them; finally, those information fragments whose similarity exceeds the first similarity threshold and have the same technical subject will be screened out as designated information fragments that are highly consistent with the target novelty search point.

[0089] 1013. Integrate the designated information fragment with the target novelty search point to obtain a comprehensive novelty search point.

[0090] In the embodiment of the present application, the technology novelty search system will integrate the specified information fragments with the target novelty search point to form a comprehensive novelty search point with rich content and strong pertinence. The comprehensive novelty search point not only includes the novelty search point explicitly specified by the user, but also integrates the key information fragments extracted from the title and technical points, thereby more accurately describing the specific content that the user wants to search for.

[0091] Specifically, the science and technology novelty search and retrieval system combines the specified information fragments with the target novelty search points through text splicing, semantic fusion and other technologies. Continuing with the above example, the target novelty search point is "photoelectric conversion efficiency improvement technology". Assuming that the specified information fragments "high efficiency", "solar cell", and "nanostructure light absorption layer" are extracted from the title, the integrated comprehensive novelty search point can be described as "research and development of high-efficiency solar cells that use nanostructure light absorption layers to improve photoelectric conversion efficiency." In this way, by integrating the specified information fragments and the target novelty search point, a more accurate and comprehensive comprehensive novelty search point is formed, providing a more valuable data foundation for the subsequent steps of vector conversion and novelty search and retrieval.

[0092] 1014. Extract at least one novelty search keyword set by the user in the novelty search application form, splice the comprehensive novelty search point with the at least one novelty search keyword, and use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the spliced ​​comprehensive novelty search point and the at least one novelty search keyword to obtain a novelty search semantic vector.

[0093] In the embodiment of the present application, the scientific and technological novelty search system will extract at least one novelty search keyword set by the user in the novelty search application form, splice the comprehensive novelty search point with at least one novelty search keyword, and use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the spliced ​​comprehensive novelty search point and at least one novelty search keyword to obtain a novelty search semantic vector, that is, based on the scientific and technological novelty search vector model, the semantic features of the spliced ​​comprehensive novelty search point and at least one novelty search keyword are characterized, and the novelty search semantic vector is generated. Through vectorization processing, the foundation for subsequent label and vector fusion retrieval / recall is laid. Among them, the scientific and technological novelty search vector model is a pre-trained neural network model that can convert the semantic information of scientific and technological documents or novelty search points into a high-dimensional vector form so as to be efficiently processed in a computer. Through vectorization processing, the spliced ​​comprehensive novelty search point and at least one novelty search keyword can be stored and retrieved in a computer in a more accurate and efficient manner, and the novelty search semantic vector provides a possibility for the subsequent fusion recall mechanism, that is, labels (such as keywords, themes, etc.) and vectors (such as novelty search semantic vectors) can be used simultaneously to match and recall documents, thereby greatly improving the accuracy and recall rate of novelty search.

[0094] The method provided in the embodiment of the present application significantly improves the efficiency and accuracy of scientific and technological novelty searches by automatically reading novelty search application forms, extracting key information, integrating information fragments, and vectorizing them. At the same time, it uses pre-trained models and vector conversion technology to quickly generate novelty search semantic vectors that reflect the core semantics of the novelty search point, providing a more accurate and efficient benchmark for subsequent document retrieval and matching, thereby helping researchers quickly locate scientific and technological literature that is highly relevant to the target novelty search point and improving novelty search efficiency.

[0095] 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 extracting target semantic elements, such as Figure 3 As shown, that is, according to the preset data extraction dimensions, the target semantic elements of the comprehensive novelty search point are extracted, including:

[0096] 1021. Extract the target novelty points to be searched in the novelty search application form, and extract the research object information and technical subject information of the novelty search application form based on the target novelty search points.

[0097] In an embodiment of the present application, in order to extract the target semantic elements, the scientific and technological novelty search and retrieval system will first understand the content of the novelty search application form, determine the target novelty search points to be searched, and extract the research object information and technical subject information of the novelty search application form based on the target novelty search points, so as to accurately grasp the novelty search intention and lay the foundation for subsequent novelty search and retrieval.

[0098] It should be noted that in the actual application process, users can customize the element types extracted from the target novelty search points according to their own business needs to adapt to novelty search tasks in different fields.

[0099] 1022. Extract technical feature information based on comprehensive novelty search points.

[0100] In the embodiments of this application, the technology search and retrieval system will clarify the core research content and expected goals of the comprehensive search point, carefully sort out the technical details involved in the searched project, and extract the technical characteristics information of the searched project. The technical characteristics information may be reflected in new theoretical applications, method improvements, process optimization, equipment innovation or system integration, and this application does not limit the specific content of the technical characteristics.

[0101] 1023. The extracted research object information, technical subject information and technical feature information are used as target semantic elements.

[0102] In the embodiment of the present application, the technology search system will use the extracted research object information, technical theme information and technical feature information as target semantic elements. The target semantic elements are the basic units that constitute the core meaning of the novelty search application form and the comprehensive novelty search point, and can include keywords, phrases, sentences or paragraphs.

[0103] The method provided in the embodiment of the present application, through the extraction of semantic elements, not only improves the intelligent level of information extraction, but also makes the extracted target semantic elements more accurate and comprehensive, providing a more precise and reliable basis for subsequent document retrieval and matching, thereby improving the accuracy and efficiency of new search.

[0104] 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 determining a semantic set and constructing a semantic search formula, such as Figure 4 As shown, the target semantic element is expanded to obtain the semantic set corresponding to the target semantic element, and the semantic set is used to construct a semantic search formula, including:

[0105] 1031. Based on the pre-built novelty search semantic library, the target semantic elements are searched, and the target semantic elements are expanded according to the search results to obtain a semantic set.

[0106] In order to comprehensively and accurately cover the literature information related to the target semantic elements, the embodiment of the present application pre-constructs a novelty search semantic library. The novelty search semantic library is a database that integrates a large number of professional vocabulary in the field of science and technology and their associations, including multiple vocabulary types such as standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words. These vocabulary types together constitute a comprehensive description of the content of scientific and technological documents, providing a solid foundation for subsequent semantic retrieval.

[0107] In the specific implementation process, the science and technology novelty search and retrieval system will search for related words that have an association relationship with the target semantic elements based on the novelty search semantic library. The related words include one or more of the standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words that have an association relationship with the target semantic elements, so as to achieve a full-scale expansion of the target semantic elements.

[0108] If the related words of the target semantic element are retrieved, the science and technology novelty search system will use the retrieved related words as the extended content of the target semantic element, and integrate the retrieved related words with the target semantic element to obtain a semantic set. The semantic set not only covers the key information in the original query, but also increases the breadth and depth of the query through the expansion of related words, thereby improving the accuracy and comprehensiveness of the novelty search.

[0109] If the relevant words of the target semantic element are not retrieved, the scientific and technological novelty search and retrieval system will conduct an in-depth analysis of the research object, technical subject and technical characteristics indicated by the target semantic element based on the pre-trained scientific and technological literature novelty search model, extract relevant standard words, synonyms and other related words from the research object, technical subject and technical characteristics indicated by the target semantic element, use the extracted related words as the extended content of the target semantic element, and integrate the retrieved related words with the target semantic element to obtain a semantic set, thereby realizing the intelligent extension of the target semantic element. Even when faced with words not covered in the novelty search semantic library, the analysis and reasoning capabilities of the model can be used to find related words, thereby constructing a semantic set.

[0110] For example, in the novelty search task of "intelligent manufacturing system based on artificial intelligence", the target semantic elements received by the science and technology novelty search retrieval system include "intelligent manufacturing system", "artificial intelligence algorithm", etc. Through the search of the novelty search semantic library, synonyms (such as "intelligent production system", "AI algorithm"), hypernyms (such as "automated production system", "machine learning algorithm"), related words (such as "intelligent manufacturing technology", "Industry 4.0"), etc. of these elements can be found, and they can be integrated with the target semantic elements into a semantic set. If some target semantic elements do not find related words in the novelty search semantic library, the science and technology novelty search retrieval system will use the scientific and technological literature novelty search large model to extract relevant words from the research object, technical theme, and technical characteristics, such as "deep learning", "real-time data analysis", etc., as an extension of the target semantic elements. In this way, by constructing a novelty search semantic library and using it to search and expand semantic elements, the semantic understanding ability of the science and technology novelty search retrieval system is significantly enhanced, so that the science and technology novelty search retrieval system can more accurately understand the user's query intention and construct a semantic set containing multiple vocabulary types, providing rich query resources for subsequent semantic retrieval.

[0111] 1032. The semantic set is converted into a semantic retrieval formula through the pre-built three-segment rules.

[0112] In the embodiment of the present application, after constructing a semantic set containing multiple vocabulary types, the science and technology novelty search retrieval system will convert the semantic set into a semantic search formula that can be used for actual retrieval. In order to convert the semantic search formula, a three-segment rule is pre-constructed in the science and technology novelty search retrieval system. The science and technology novelty search retrieval system will integrate the key information in the semantic set according to the three-segment rule to generate a search formula that meets the retrieval requirements.

[0113] Specifically, the technology novelty search system will classify the words in the semantic set according to the dimensions of research object, technical theme, technical characteristics, etc., and sort them according to the importance of the words in each dimension (such as word frequency, semantic weight, etc.). Then, the technology novelty search system will combine these words into a structured semantic search formula according to the three-segment rule. The semantic search formula not only contains the key information in the original query, but also realizes the accurate expression of the query intent through the correlation and importance ranking between words.

[0114] In addition, when generating search formulas, the science and technology search system will also take into account the logical relationship between words, such as using "+" to represent "or" relationships and "*" to represent "and" relationships, to ensure the accuracy and comprehensiveness of the search results. At the same time, the science and technology search system will further adjust and optimize the search formula according to the optimization strategy of the three-stage rule to improve the search efficiency and accuracy.

[0115] For example, the three-segment rule can be (research object + synonyms + hypernyms + standard words)*(technical topics + synonyms + hypernyms + standard words + related words)*(technical features + standard words + synonyms). In this way, in the "intelligent manufacturing system based on artificial intelligence" novelty search task, the words in the semantic concentration may be combined into a semantic retrieval formula as follows: "(intelligent manufacturing system + intelligent production system + AI algorithm + machine learning algorithm)*(automated production system + real-time data analysis + deep learning)*(Industry 4.0 + manufacturing technology)". This retrieval formula not only includes the key information in the original query, but also realizes the precise expression of the query intent through the association and logical relationship between words, so that the literature information related to the target semantic elements can be retrieved more accurately.

[0116] The method provided in the embodiment of the present application significantly enhances the semantic understanding ability by using a pre-constructed novelty search semantic library and using it to retrieve and expand semantic elements, more accurately understands the user's query intention, and constructs a semantic set containing multiple vocabulary types. At the same time, the semantic set is converted into a semantic retrieval formula through a pre-constructed three-segment rule, thereby achieving accurate expression of query intention and accurate control of retrieval results, which not only improves the accuracy and efficiency of novelty search, but also provides users with more comprehensive and in-depth document retrieval services.

[0117] Further, as a refinement and expansion 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 retrieving a target document, such as Figure 5 As shown, based on the novelty search semantic vector and the semantic search formula, a novelty search operation is performed to obtain multiple target documents, including:

[0118] 1041. Construct structured query statements based on novelty semantic vectors and semantic retrieval formulas.

[0119] In an embodiment of the present application, in order to realize the fusion of vector and scalar retrieval, the scientific and technological novelty search and retrieval system will construct a structured query statement based on the novelty search semantic vector and semantic retrieval formula, so that the structured query statement can accurately locate the literature resources related to the project to be searched.

[0120] 1042. Execute the structured query statement to obtain multiple candidate documents.

[0121] In the embodiment of the present application, after constructing the structured query statement, the science and technology novelty search system submits the structured query statement to the document database, and the database matches and screens the document resources according to the semantics and structure of the query statement. The database returns the document resources that highly match the query statement to form a candidate document set, which has a high correlation with the core technology and research content of the project to be searched.

[0122] In this way, through the above steps 1041 and 1042, the embodiment of the present application realizes the comprehensive recall of relevant data items based on semantic retrieval formula and novelty semantic vector, realizing the combination of comprehensiveness and precision. On the one hand, it relies on the precise positioning ability of the semantic retrieval formula to ensure that data items that strictly match the query retrieval formula can be directly captured; on the other hand, it incorporates the deep matching ability of the semantic vector, which is based on natural language processing technology and machine learning algorithms, and can capture more subtle and deep semantic associations between data items and query requirements. The embodiment of the present application takes these two types of information into consideration at the same time in a retrieval action, realizes the comprehensive and accurate recall of relevant data, and greatly improves the accuracy and efficiency of retrieval.

[0123] 1043. Filter the multiple candidate documents according to the preset filtering conditions, and use the multiple candidate documents remaining after filtering as the multiple target documents.

[0124] In the embodiment of the present application, although the candidate documents are all related to the query requirements to a certain extent, they may contain some redundant, irrelevant or outdated information. Therefore, in order to further improve the quality of data recall, the candidate documents will be carefully screened and filtered in the embodiment of the present application.

[0125] Among them, in the embodiment of the present application, filtering conditions are preset, and the filtering conditions are set based on professional knowledge and actual needs in the patent field, aiming to eliminate irrelevant or redundant information and retain high-quality candidate data that best meets the query needs, including but not limited to international patent classification numbers, time ranges, etc. The international patent classification number is a standardized number used to identify the technical field to which the patent document belongs. The international patent classification number can be used to quickly screen out documents related to the technical field of the query requirement. The time range is used to limit the time span of the search results to ensure that users obtain the latest research results.

[0126] Specifically, the science and technology novelty search system will check whether the candidate documents meet the preset filtering conditions one by one. For documents that meet the filtering conditions, they will be retained as target documents; for documents that do not meet the conditions, they will be eliminated from the candidate documents. Through fine filtering, the accuracy and practicality of the search results can be effectively improved, so that the target document set finally obtained by the user is not only highly relevant to the query needs, but also all are high-quality documents that have been strictly screened, which greatly saves the user's time and energy and improves the search efficiency. In the actual application process, a minimum of 5 and a maximum of 15 documents can be selected as target documents.

[0127] The method provided in the embodiment of the present application realizes efficient and accurate retrieval of patent documents by combining multiple retrieval methods and sophisticated filtering operations. It can not only provide users with comprehensive and high-quality retrieval results, but also can be flexibly adjusted and optimized according to user needs, thereby further improving the retrieval effect and user experience.

[0128] Further, as a refinement and expansion 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 sorting target documents, such as Figure 6 As shown, multiple target documents are sorted and output, including:

[0129] 1044. Read the novelty search application form, extract the target novelty search point to be retrieved from the novelty search application form, and use the pre-trained scientific novelty search vector model to perform vector conversion on the target novelty search point to obtain the target semantic vector.

[0130] In the embodiment of the present application, in order to enable users to preferentially see the documents that best match the target novelty search point, the science and technology novelty search retrieval system will semantically reorder multiple target documents. The basis of the semantic reordering is the degree of semantic matching between the retrieval results and the original target novelty search point. The semantic similarity between the retrieval results and the novelty search point is calculated using semantic analysis technology, and priority is sorted according to the similarity.

[0131] Specifically, the science and technology novelty search retrieval system will read the novelty search application form, extract the target novelty search point to be searched from the novelty search application form, and use the pre-trained science and technology novelty search vector model to perform vector conversion on the target novelty search point to obtain the target semantic vector. Among them, the target novelty search point refers to the specific technical point or innovation point that the user clearly pointed out in the novelty search application form and hopes to search. The science and technology novelty search retrieval system will use the science and technology novelty search vector model to analyze the semantic features of the target novelty search point and convert it into a point in the high-dimensional vector space, so as to focus on the semantic features of the novelty search point itself, which can more accurately capture the part directly related to the novelty search point and further improve the accuracy of the target document sorting.

[0132] 1045. Each target document is segmented to obtain multiple text segments.

[0133] In the embodiment of the present application, the science and technology novelty search system will segment each target document to generate multiple text segments. Among them, since the science and technology novelty search vector model for vector conversion has a specific length requirement, the science and technology novelty search system will make a preliminary judgment and segment the target document according to this requirement, aiming to ensure that each text segment obtained by segmentation meets the input requirements of the science and technology novelty search vector model. At the same time, in order to ensure that the segmented text segments can retain the semantic information of the original data as much as possible, a certain overlap between the previous and next segments is allowed during segmentation. This overlap not only helps to maintain the integrity of the semantics, but also provides more context information for subsequent analysis and processing.

[0134] 1046. Using the pre-trained technology novelty search vector model, multiple text fragments are converted into vectors to obtain multiple fragment semantic vectors.

[0135] In the embodiment of the present application, for each text segment, the technology novelty search retrieval system will use the pre-trained technology novelty search vector model to perform vector conversion on each text segment to obtain the segment semantic vector of each text segment, thus obtaining multiple segment semantic vectors. Through the vector conversion and set construction of text segments, the technology novelty search retrieval system can express the semantic features of the target document in the form of a vector, providing a basis for subsequent semantic similarity calculations, and helping to improve the accuracy and efficiency of retrieval.

[0136] 1047. Batch calculate the vector similarity between the target semantic vector and each of the multiple fragment semantic vectors, divide the fragment semantic vectors belonging to the same target document into the same vector group to obtain the vector group corresponding to each target document, and extract the specified fragment semantic vector with the highest vector similarity from the vector group corresponding to each target document to obtain the specified semantic fragment vector corresponding to each target document.

[0137] In an embodiment of the present application, the science and technology novelty search and retrieval system will batch calculate the similarity between the target semantic vector and multiple fragment semantic vectors. Similarity calculation can adopt methods such as cosine similarity and Euclidean distance. Then, the science and technology novelty search and retrieval system will divide the fragment semantic vectors belonging to the same target document into the same vector group, and extract the specified fragment semantic vector with the highest vector similarity in each vector group as the representative of the document. In this way, through similarity calculation and vector group division, the document fragments most relevant to the target novelty search point can be screened out, providing an important basis for subsequent document sorting.

[0138] 1048. According to the vector similarity of the specified semantic fragment vector corresponding to each target document, the multiple specified semantic fragment vectors currently obtained are sorted to obtain a vector sorting result, and according to the vector sorting result, the multiple target documents are sorted, and the sorted multiple target documents are output.

[0139] In an embodiment of the present application, the scientific and technological novelty search and retrieval system will sort the vector similarities of the specified semantic fragment vectors corresponding to each target document to obtain a vector sorting result. Then, multiple target documents are sorted according to the vector sorting result, and the sorted document list is output to the user. The sorting result can be displayed in a list format, including key information such as document title, author, and publication time, which is convenient for users to quickly browse and select. Through document sorting, users can quickly find the documents most relevant to the target novelty search point, thereby improving the efficiency of novelty search; at the same time, the sorting results are displayed to users in an intuitive manner, which is convenient for users to understand and use.

[0140] The method provided in the embodiment of the present application performs recall based on the expanded novelty search points and semantic sets, effectively expands the search information, and preliminarily screens the content related to the novelty search points; further, in order to further improve the user experience and search accuracy, the embodiment of the present application performs secondary screening and priority sorting on the retrieved documents based on the scientific and technological novelty search vector model, ensuring that the most relevant and valuable content to the novelty search points can be ranked first, so that users can find the information they need more quickly and accurately.

[0141] In summary, the logical process of a technology novelty search method based on semantic elements provided by this application is summarized as follows: Figure 7As shown, a novelty search application form is received, and a comprehensive novelty search point is generated according to the content to be searched indicated in the novelty search application form, and a novelty search semantic vector of the comprehensive novelty search point is obtained. The target semantic element is extracted, and the target semantic element is expanded to obtain the semantic set corresponding to the target semantic element. It is detected whether manual intervention on the semantic set is requested. When it is detected that no manual intervention is performed, the semantic set is used to construct a semantic search formula. Otherwise, manual intervention on the semantic set is allowed, and the semantic search formula is constructed using the intervened semantic set. Next, a fusion recall operation is performed using the novelty search semantic vector and the semantic search formula, and the recalled documents are filtered using preset filtering conditions to obtain multiple target documents. After the multiple target documents are comprehensively sorted, a candidate list is obtained and output.

[0142] Further, as Figure 1 In the specific implementation of the method, the present application embodiment provides a technology novelty search device based on semantic elements, such as Figure 8 As shown, the device includes: a generation module 801, an extraction module 802, a construction module 803 and a retrieval module 804.

[0143] The generating module 801 is used to extract the target semantic elements based on the novelty search application form and the comprehensive novelty search points;

[0144] The extraction module 802 is used to extract the target semantic elements of the comprehensive novelty search point according to the preset data extraction dimension;

[0145] The construction module 803 is used to expand the target semantic element to obtain a semantic set corresponding to the target semantic element, and to construct a semantic search formula using the semantic set;

[0146] The retrieval module 804 is used to perform a novelty search operation based on the novelty search semantic vector and the semantic search formula to obtain multiple target documents, and to sort and output the multiple target documents.

[0147] In a specific application scenario, the generation module 801 is used to read the novelty search application form, extract the target novelty search point, title and technical points to be retrieved from the novelty search application form; use the pre-trained scientific and technological literature novelty search large model to extract the specified information segment from the title and the technical points, the similarity between the specified information segment and the target novelty search point is greater than the first similarity threshold and has the same technical theme as the target novelty search point; integrate the specified information segment with the target novelty search point to obtain the comprehensive novelty search point; extract at least one novelty search keyword set by the user from the novelty search application form, and splice the comprehensive novelty search point with the at least one novelty search keyword; use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the spliced ​​comprehensive novelty search point and the at least one novelty search keyword to obtain the novelty search semantic vector.

[0148] In a specific application scenario, the extraction module 802 is used to extract the target novelty search points to be retrieved in the novelty search application form, and extract the research object information and technical subject information of the novelty search application form based on the target novelty search points; extract the technical feature information based on the comprehensive novelty search points; and use the extracted research object information, the technical subject information and the technical feature information as the target semantic elements.

[0149] In a specific application scenario, the construction module 803 is used to retrieve the target semantic elements based on a pre-constructed novelty search semantic library, and to expand the target semantic elements according to the search results to obtain the semantic set; and to convert the semantic set into the semantic search formula through a pre-constructed three-segment rule.

[0150] In a specific application scenario, the construction module 803 is used to retrieve related words that are associated with the target semantic element based on the novelty search semantic library, wherein the related words include one or more of standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words that are associated with the target semantic element; if the related words of the target semantic element are retrieved, the retrieved related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set; if the related words of the target semantic element are not retrieved, the related words are extracted from the research objects, technical topics, and technical features indicated by the target semantic element based on the pre-trained scientific and technological literature novelty search model, the extracted related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set.

[0151] In a specific application scenario, the retrieval module 804 is used to construct a structured query statement based on the novelty search semantic vector and the semantic retrieval formula; execute the structured query statement to obtain multiple candidate documents; filter the multiple candidate documents according to preset filtering conditions, and use the multiple candidate documents remaining after filtering as the multiple target documents.

[0152] In a specific application scenario, the retrieval module 804 is used to read the novelty search application form, extract the target novelty search point to be searched from the novelty search application form, and use a pre-trained scientific and technological novelty search vector model to perform vector conversion on the target novelty search point to obtain a target semantic vector; segment each of the target documents to obtain multiple text segments; use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the multiple text segments to obtain multiple segment semantic vectors; batch calculate the vector similarity between the target semantic vector and each segment semantic vector in the multiple segment semantic vectors; divide the segment semantic vectors belonging to the same target document into the same vector group to obtain a vector group corresponding to each target document, and extract the specified segment semantic vector with the highest vector similarity from the vector group corresponding to each target document to obtain a specified semantic segment vector corresponding to each target document; sort the multiple specified semantic segment vectors currently obtained according to the vector similarity of the specified semantic segment vector corresponding to each target document to obtain a vector sorting result, and sort the multiple target documents according to the vector sorting result, and output the sorted multiple target documents.

[0153] The device provided in the embodiment of the present application, when receiving a novelty search application form, generates a comprehensive novelty search point according to the content to be searched indicated in the novelty search application form, and obtains the novelty search semantic vector of the comprehensive novelty search point, extracts the target semantic element based on the novelty search application form and the comprehensive novelty search point, expands the target semantic element to obtain the semantic set corresponding to the target semantic element, and constructs a semantic search formula using the semantic set, performs a novelty search operation based on the novelty search semantic vector and the semantic search formula, obtains multiple target documents, and outputs the multiple target documents after sorting, and intelligently processes the information in the novelty search application form. , obtain the novelty search semantic vector, and automatically generate the novelty search retrieval formula, combine the novelty search semantic vector and the semantic retrieval formula to perform retrieval, capture the deep semantics of the target novelty search point, and ensure that the documents highly relevant to the target novelty search point are quickly screened out from the massive document resources. At the same time, through intelligent semantic analysis and retrieval technology, the user's operation difficulty and time cost are reduced. Users do not need to have too much professional knowledge to easily perform scientific and technological novelty searches and obtain satisfactory search results, which significantly improves the accuracy, recall rate and retrieval efficiency of the retrieval results, allowing users to obtain comprehensive and accurate novelty search results more quickly.

[0154] It should be noted that for other corresponding descriptions of the functional units involved in the technology novelty search device based on semantic elements provided in the embodiment of the present application, reference can be made to Figures 1 to 7 The corresponding description in will not be repeated here.

[0155] 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.

[0156] 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.

[0157] 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 understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations 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.

[0158] 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 technology novelty search method based on semantic elements in the above embodiment.

[0159] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the technology novelty search method based on semantic elements.

[0160] 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.

[0161] 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.

[0162] 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.

[0163] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0164] 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 technology novelty search method based on semantic elements, characterized in that: include: When a novelty search application form is received, a comprehensive novelty search point is generated according to the content to be searched for novelty as indicated in the novelty search application form, and a novelty search semantic vector of the comprehensive novelty search point is obtained; Extracting target semantic elements based on the novelty search application form and the comprehensive novelty search points; Expanding the target semantic element to obtain a semantic set corresponding to the target semantic element, and constructing a semantic search formula using the semantic set; Based on the novelty search semantic vector and the semantic search formula, a novelty search operation is performed to obtain a plurality of target documents, and the plurality of target documents are sorted and then output.

2. The method according to claim 1, characterized in that The step of generating a comprehensive novelty search point according to the content to be searched for novelty indicated by the novelty search application form, and obtaining a novelty search semantic vector of the comprehensive novelty search point comprises: Read the novelty search application form, and extract the target novelty search point, title and technical points to be retrieved from the novelty search application form; Using a pre-trained scientific and technological literature search model, extract a designated information segment from the title and the technical points, wherein the similarity between the designated information segment and the target search point is greater than a first similarity threshold and the designated information segment has the same technical theme as the target search point; Integrate the designated information fragment with the target novelty search point to obtain the comprehensive novelty search point; Extracting at least one novelty search keyword set by a user from the novelty search application form, and splicing the comprehensive novelty search point with the at least one novelty search keyword; The pre-trained technology novelty search vector model is used to perform vector conversion on the spliced ​​comprehensive novelty search point and the at least one novelty search keyword to obtain the novelty search semantic vector.

3. The method according to claim 1, characterized in that The step of extracting target semantic elements based on the novelty search application form and the comprehensive novelty search point includes: Extracting target novelty points to be searched from the novelty search application form, and extracting research object information and technical subject information of the novelty search application form according to the target novelty search points; Extracting technical feature information based on the comprehensive novelty search points; The extracted research object information, technical subject information and technical feature information are used as the target semantic elements.

4. The method according to claim 1, characterized in that: The step of expanding the target semantic element to obtain a semantic set corresponding to the target semantic element, and constructing a semantic search formula using the semantic set, includes: Based on the pre-built novelty search semantic library, the target semantic element is searched, and the target semantic element is expanded according to the search result to obtain the semantic set; The semantic set is converted into the semantic search formula through the pre-constructed three-segment rule.

5. The method according to claim 4, characterized in that The method of searching the target semantic element based on the pre-built novelty search semantic library and expanding the target semantic element according to the search result to obtain the semantic set includes: Based on the novelty search semantic library, searching for related words associated with the target semantic element, wherein the related words include one or more of standard words, synonyms, keywords, free words, abbreviations, hypernyms, hyponyms, and related words associated with the target semantic element; If related words of the target semantic element are retrieved, the retrieved related words are used as extended content of the target semantic element, and the retrieved related words are integrated with the target semantic element to obtain the semantic set; If the related words of the target semantic element are not retrieved, the related words are extracted from the research object, technical theme and technical characteristics indicated by the target semantic element based on the pre-trained scientific and technological literature retrieval model, and the extracted related words are used as extended content of the target semantic element. The retrieved related words are integrated with the target semantic element to obtain the semantic set.

6. The method according to claim 1, characterized in that The novelty search operation is performed based on the novelty search semantic vector and the semantic search formula to obtain multiple target documents, including: Constructing a structured query statement based on the novelty search semantic vector and the semantic search formula; Executing the structured query statement to obtain multiple candidate documents; The plurality of candidate documents are filtered according to a preset filtering condition, and the plurality of candidate documents remaining after filtering are used as the plurality of target documents.

7. The method according to claim 1, characterized in that The step of sorting and outputting the plurality of target documents comprises: Read the novelty search application form, extract the target novelty search point to be retrieved from the novelty search application form, and use the pre-trained scientific and technological novelty search vector model to perform vector conversion on the target novelty search point to obtain a target semantic vector; Segmenting each of the target documents to obtain multiple text segments; Using the pre-trained technology search vector model, the plurality of text segments are vectorized to obtain a plurality of segment semantic vectors; Batch calculating the vector similarity between the target semantic vector and each of the multiple segment semantic vectors; The segment semantic vectors belonging to the same target document are divided into the same vector group to obtain a vector group corresponding to each target document, and the specified segment semantic vector with the highest vector similarity is extracted from the vector group corresponding to each target document to obtain a specified semantic segment vector corresponding to each target document; According to the vector similarity of the specified semantic segment vector corresponding to each of the target documents, the multiple specified semantic segment vectors currently obtained are sorted to obtain a vector sorting result, and according to the vector sorting result, the multiple target documents are sorted, and the sorted multiple target documents are output.

8. A technology novelty search device based on semantic elements, characterized in that: include: A generating module, for generating a comprehensive novelty search point according to the content to be searched indicated by the novelty search application form when receiving the novelty search application form, and obtaining the novelty search semantic vector of the comprehensive novelty search point; An extraction module, used for extracting target semantic elements based on the novelty search application form and the comprehensive novelty search point; A construction module, used to expand the target semantic element, obtain a semantic set corresponding to the target semantic element, and construct a semantic search formula using the semantic set; The retrieval module is used to perform a novelty search operation based on the novelty search semantic vector and the semantic search formula to obtain multiple target documents, and to sort the multiple target documents and then output them.

9. A device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: 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 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