Automatic identification method for duplication inventors

Abstract

The invention provides an automatic identification method for a duplication inventor, belongs to the technical field of duplication inventor automatic identification, and solves the problem that a duplication inventor identification method capable of efficiently processing total data and ensuring the accuracy of a disambiguation result is lacked in the prior art. The method comprises the steps that the name of each to-be-recognized inventor serves as a retrieval keyword, retrieval is conducted one by one in a full-amount Chinese patent database, and description item information of all patent applications under the corresponding to-be-recognized inventor name is extracted; constructing a graph structure based on a co-occurrence relation according to the description item information of all patent applications under each to-be-identified inventor name; and according to the co-occurrence relation-based graph structure corresponding to each to-be-identified inventor, identifying the duplication inventor of the corresponding to-be-identified inventor.

Description

Technical Field

[0001] This invention relates to the field of automatic identification of inventors with the same name, and in particular to an automatic identification method for inventors with the same name. Background Technology

[0002] Currently, China has consistently ranked first globally in patent applications for many years. However, with such a massive volume of data, the phenomenon of inventors sharing the same name is becoming increasingly prominent. The resulting statistical distortion, search noise, talent evaluation bias, and operational decision-making risks have become unavoidable pain points in patent intelligence mining and patent asset management.

[0003] Currently, the mainstream approach to name discrepancy disambiguation both domestically and internationally can be summarized as "text similarity + clustering": First, extract fields such as patent title, abstract, claims, applicant, IPC, CPC, and citations to construct high-dimensional text vectors; then, use methods such as hierarchical clustering, DBSCAN, spectral clustering, or deep embedding clustering to merge patents "suspected to have the same inventor." However, this approach reveals a dual bottleneck of "scale and accuracy" when processing the full range of Chinese patent data, specifically as follows: (1) Computational bottleneck: Faced with pairwise similarity calculation of tens of millions of inventor nodes and hundreds of millions of patent texts, even with sparse vectors and GPU acceleration, it is still necessary to pay extremely high costs in terms of memory and time, making it difficult to complete the routine update of the full data.

[0004] (2) Semantic drift: Patent texts are highly templated and have domain synonyms. Different inventors may use the same technical terms, while the same inventor may also have significant differences in text distribution due to technological transformation, resulting in clustering failure phenomenon of "large distance between texts of the same person and small distance between texts of different people".

[0005] (3) Complex post-processing: Clustering results often result in extreme clusters of “over-merging” or “over-splitting”, which require the introduction of manual rules or crowdsourced annotations for secondary correction. This not only has high maintenance costs, but also makes it difficult to productize.

[0006] In summary, given the aforementioned shortcomings in existing technologies, there is an urgent need to research a method for identifying inventors with duplicate names that can balance overall processing efficiency with ensuring the accuracy of disambiguation results. Summary of the Invention

[0007] Based on the above analysis, the present invention aims to provide an automatic identification method for inventors with the same name, in order to solve the problem in the prior art of lacking an inventor identification method with the same name that can both efficiently process the full amount of data and ensure the accuracy of the disambiguation results.

[0008] This invention discloses an automatic identification method for inventors with the same name, the method comprising: Using the name of each inventor to be identified as a search keyword, a search was conducted in the full Chinese patent database to extract the bibliographic information of all patent applications under the name of the corresponding inventor. Based on the bibliographic information of all patent applications under each inventor to be identified, a graph structure based on co-occurrence relationships is constructed. Based on the graph structure of co-occurrence relationships corresponding to each inventor to be identified, the system identifies inventors with the same name as the corresponding inventor to be identified.

[0009] Based on the above solution, the present invention also makes the following improvements: Furthermore, the construction of the graph structure based on co-occurrence relationships is performed as follows: For each patent application under the name of the inventor to be identified, the application number of the patent application is used as a unique identifier to construct a node in the graph structure; Based on the different attributes of the patent applications under the name of the inventor to be identified, construct the edges of various types between the corresponding nodes in the graph structure; Based on the nodes in the graph structure and the edges of various types between the corresponding nodes, a graph structure based on co-occurrence relationships is constructed.

[0010] Furthermore, the process of identifying inventors with the same name as the inventor to be identified involves the following steps: Determine whether all nodes in the graph structure are isolated nodes. If so, treat the inventors of the patent applications corresponding to each node as independent inventors, and identify inventors with the same name among all independent inventors. The process of identifying inventors with the same name ends.

[0011] Furthermore, the process of identifying inventors with the same name as the corresponding inventor to be identified also includes: Determine whether the weights of the edges with applicant types between nodes in the graph structure are all greater than or equal to 1. If so, then the inventors of the patent applications corresponding to all nodes in the graph structure are considered to be the same independent inventors, and there are no inventors with the same name. The process of identifying inventors with the same name ends.

[0012] Furthermore, the process of identifying inventors with the same name as the corresponding inventor to be identified also includes: If not all nodes in the graph structure are isolated nodes, and the weights of the edges of applicant type between nodes in the graph structure are not all greater than or equal to 1, a combination of isolated node discrimination, applicant type edge weight discrimination, and graph structure clustering is used to identify inventors with the same name as the corresponding inventors to be identified.

[0013] Furthermore, the method employing a combination of isolated node discrimination, applicant type edge weight discrimination, and graph structure clustering is used to identify inventors with the same name as the corresponding inventor to be identified, and the following steps are performed: Step S331: Determine whether there are isolated nodes in the graph structure. If so, treat the inventors of the patent applications corresponding to each isolated node as independent inventors, remove the corresponding isolated nodes from the graph structure, update the graph structure, and jump to step S332; otherwise, jump directly to step S332. Step S332: Determine whether the weights of the edges of applicant types between each node in the graph structure are all greater than or equal to 1. If so, regard the inventors to be identified in the patent applications corresponding to all nodes in the graph structure as the same independent inventor. There is a name-duplicate relationship between the independent inventors determined in steps S331 and S332, and the name-duplicate inventor identification process ends; otherwise, jump to step S333. Step S333: Cluster the graph structure and regard the inventors of the patent applications corresponding to the nodes that are clustered into one class as the same independent inventor. There is a name duplication relationship among the independent inventors identified in steps S331, S332 and S333, and the process of identifying inventors with the same name ends.

[0014] Furthermore, the feature is that the dimensions include: co-signing inventors, same applicants, same application date, and same classification number.

[0015] Furthermore, for the dimension of co-authored inventors, the corresponding type of edge is constructed in the following way: If two patent applications have inventors who are jointly listed, then an edge is constructed between the corresponding nodes of the two patent applications in the graph structure, and the edge is marked as "jointly listed inventors" type. The number of co-authors is used as the weight of that edge.

[0016] Furthermore, for the same applicant dimension, edges of corresponding types are constructed in the following manner: If two patent applications have the same applicant, add an edge between the corresponding nodes and mark the type as "same applicant"; The number of identical applicants is used as the weight of that edge.

[0017] Furthermore, for the dimension of the same application date, edges of the corresponding type are constructed in the following way: When patent applications with the same application date are found, an edge is created between the nodes corresponding to the respective patent applications, and the edge type is defined as "same application date".

[0018] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: (1) Significantly improves processing efficiency 1) Full Data Processing Capability: By constructing a graph structure based on co-occurrence relationships, the high computational cost of pairwise similarity calculation for massive patent texts, as required by traditional methods, is avoided. This invention can efficiently process the entire set of Chinese patent data without relying on complex methods such as sparse vectors and GPU acceleration, enabling routine updates of large-scale data.

[0019] 2) Rapid retrieval and extraction: In step S1, through precise retrieval and standardized data processing, the patent application information of each inventor to be identified can be quickly extracted from the full patent database, which greatly improves the speed and accuracy of data preprocessing.

[0020] (2) Improve the accuracy of disambiguation results 1) Multi-dimensional correlation analysis: This invention not only considers the similarity of patent texts, but also constructs edges in a graph structure from multiple dimensions such as co-authored inventors, same applicants, same filing dates, and same classification numbers to comprehensively evaluate the correlation between patents. This multi-dimensional analysis method effectively avoids semantic drift problems and can more accurately identify the true same inventor.

[0021] 2) Advantages of graph structure clustering: In step S3, graph structure clustering algorithms (such as community detection algorithms or spectral clustering algorithms) are used to identify inventors with the same name. Based on the connection relationship between nodes and the weight of edges, inventors can be more reasonably divided, avoiding the problems of "over-merging" or "over-splitting" in traditional clustering methods, thereby improving the accuracy of disambiguation results.

[0022] (3) Reduce post-processing complexity and cost 1) Reduced manual intervention: Through the reasonable construction of graph structure and efficient clustering algorithm, this invention can automatically generate more accurate results for identifying inventors with the same name, reducing the reliance on manual rules or crowdsourced labeling, thereby reducing the complexity of post-processing and maintenance costs.

[0023] 2) Easy to productize: The method of the present invention has a high degree of automation and scalability, and can be easily integrated into existing patent management systems to achieve productized application, providing efficient and accurate data support and decision-making basis for patent management, technology research and development evaluation and other work.

[0024] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained through the details specifically pointed out in the description and drawings. Attached Figure Description

[0025] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 A flowchart of an automatic identification method for inventors with the same name provided in an embodiment of the present invention. Detailed Implementation

[0026] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0027] A specific embodiment of the present invention discloses an automatic identification method for inventors with the same name, the flowchart of which is shown below. Figure 1 As shown.

[0028] Step S1: Using the name of each inventor to be identified as a search keyword, search the entire Chinese patent database one by one to extract the bibliographic information of all patent applications under the name of the corresponding inventor to be identified.

[0029] Step S11: Obtain all Chinese patent data from the full Chinese patent database, extract the bibliographic information of each patent application, remove duplicates and invalid information, and obtain a standardized patent bibliographic data set.

[0030] In the specific implementation process, duplicate items and invalid information can be removed in the following way.

[0031] (1) Data acquisition Connect to the full Chinese patent database to ensure comprehensive access to all patent data stored within. Employ appropriate data acquisition methods (such as database query interfaces and data export functions) to completely extract all Chinese patent data and store it locally or in a designated data storage system for subsequent processing.

[0032] (2) Extraction of information from bibliographic records For each extracted patent application data, data parsing tools or corresponding data processing programs are used to extract the bibliographic information according to the format and specifications of patent bibliographic information. Bibliographic information typically includes, but is not limited to, patent application number, application date, invention title, applicant, inventor, address, patent agency, agent, priority, etc. During the extraction process, it is ensured that the content of each bibliographic item is obtained accurately and stored in a structured data format for easy subsequent processing and analysis.

[0033] (3) Remove duplicate items Because patent data may contain duplicate records, it is necessary to deduplicate the extracted bibliographic information. This can be done by sorting and comparing the data based on unique identifiers such as patent application numbers to identify duplicate records. For duplicate records, only one complete and accurate record is retained, while the remaining duplicates are deleted to ensure the uniqueness and accuracy of the data.

[0034] (4) Remove invalid information The extracted bibliographic information may contain invalid information, such as incomplete, incorrect, or non-compliant information. Each piece of bibliographic information is checked and filtered using data validation rules and logical judgments. For example, it checks whether the patent application number conforms to the prescribed format, whether the applicant and inventor information is complete, and whether the address information is valid. Invalid information is marked or deleted to ensure that the information in the final dataset is valid and usable.

[0035] (5) Data standardization The bibliographic data, after being deduplicated and having invalid information removed, undergoes further standardization processing.

[0036] Standardize data formats, such as unifying date formats to YYYY-MM-DD and standardizing address information.

[0037] For any potentially non-standard text information, such as extra spaces or special characters in the applicant's name, we cleaned and corrected it to ensure that the data conformed to the standard specifications for patent bibliographic information.

[0038] (6) Dataset generation and storage The standardized bibliographic information of the patent applications, after the above processing, is organized into a complete dataset according to a certain data structure and format.

[0039] Store the dataset in a suitable storage medium, such as a database, data warehouse, or file system, for subsequent use and analysis.

[0040] At the same time, corresponding indexes and query functions can be established for the dataset, making it convenient for users to quickly retrieve and obtain the required patent bibliographic information.

[0041] Through the specific implementation process of step S11 above, patent data can be obtained and processed from the full Chinese patent database, ultimately resulting in a standardized patent bibliographic data set, providing accurate, complete and standardized data support for patent application and other related work.

[0042] Step S12: Using the name of each inventor to be identified as a search keyword, accurately search the inventor information in the standardized patent bibliographic data set, and extract the bibliographic information of all patent applications under the name of each inventor to be identified.

[0043] This results in a list of patent information for each inventor to be identified.

[0044] The specific processing procedure is described below.

[0045] (1) Preparation of search keywords Extract the name of each inventor to be identified from the existing list of inventors, using it as search keywords. Ensure the accuracy of the names, including full names, possible aliases, or variant forms, to avoid omitting any relevant records.

[0046] For inventors with multiple names or nicknames, a name mapping table is created to associate all possible name forms with a unified identifier, ensuring comprehensive retrieval.

[0047] (2) Dataset retrieval Using database query tools or a specially developed search program, the inventor information field in a standardized patent bibliographic dataset is searched using the name of each inventor to be identified as a keyword. During the search, a combination of fuzzy and exact matching is employed to handle potential spelling errors, name abbreviations, or different language versions of the name. For each matching bibliographic item, a complete record is made, including the patent application number, application date, invention title, applicant information, inventor information, address, and other key information.

[0048] (3) Extraction and organization Information for each bibliographic entry retrieved is extracted to ensure that the extracted information is complete and accurate.

[0049] The bibliographic information of all patent applications under the name of each inventor to be identified is compiled into a separate data list. Each list is identified by the name of the inventor to be identified, which facilitates subsequent analysis and processing.

[0050] The extracted information is initially verified to ensure that no duplicate or erroneous records are mixed in, thus guaranteeing the accuracy and integrity of the data.

[0051] (4) Data storage and indexing Store the patent information list of each inventor to be identified in a dedicated data storage system, such as a relational database or file system.

[0052] Index the data list for each inventor to be identified to enable fast retrieval and querying, thereby improving data availability.

[0053] Data can be exported to standard formats (such as CSV, XML, etc.) as needed, facilitating data exchange with other systems or tools.

[0054] (5) Result verification The extracted results are sampled and verified to ensure the accuracy of the retrieval and extraction process.

[0055] During the verification process, the original dataset and the extracted results are compared to check for any missing or incorrect records, ensuring that the patent information of each inventor to be identified is complete and accurate.

[0056] Through the processing in step S12 above, patent application information under each inventor to be identified can be efficiently extracted from the standardized patent bibliographic data set, providing accurate data support for subsequent patent analysis, inventor contribution evaluation and other work.

[0057] Step S2: Construct a graph structure based on co-occurrence relationships based on the bibliographic information of all patent applications under the name of each inventor to be identified.

[0058] Step S21: For each patent application under the name of the inventor to be identified, use the application number of the patent application as a unique identifier to construct a node in the graph structure.

[0059] In the specific implementation process, corresponding nodes are created in the graph structure based on the extracted patent application numbers. Each node stores basic patent information associated with that patent application number, such as a list of inventors, applicant information, application date, classification number, etc., to facilitate subsequent edge construction and analysis.

[0060] Step S22: Based on the different dimensions of the attributes of the patent applications under the name of the inventor to be identified, construct the edges of various types between the corresponding nodes in the graph structure.

[0061] (1) Dimension of co-authored inventors Traverse patent applications: For each patent application, check whether there is any overlap between its inventor list and the inventor list of other patent applications, except for the inventor to be identified.

[0062] Edge Construction: If two patent applications have co-authored inventors, an edge is constructed between the corresponding nodes of the two patent applications in the graph structure, and this edge is labeled as "co-authored inventor". The number of co-authored inventors can be used as the weight of this edge for subsequent analysis and calculation.

[0063] (2) Same applicant dimension Compare applicant information: Compare the applicant information of each patent application one by one to determine whether there are the same applicants in different patent applications.

[0064] Edge Construction: If two patent applications have the same applicant, an edge is added between the corresponding nodes, and its type is marked as "same applicant". The number of identical applicants can be used as the weight of this edge for subsequent analysis and calculation. This edge can indicate that these patent applications are related in terms of the applicant, possibly reflecting the same applicant's continuous R&D and patent layout in the relevant technical field.

[0065] (3) Same application date dimension Check the application dates: Examine the application dates of different patent applications to determine if multiple patent applications were filed on the same day.

[0066] Edge construction: When patent applications with the same filing date are found, an edge is established between the nodes corresponding to these patent applications, and the edge type is defined as "same filing date". This may mean that these patent applications are patent layouts targeting the same or related technical solutions at the same time, or that they may have been filed simultaneously for some strategic purpose.

[0067] (4) Same classification number dimension Analysis of classification numbers: Analyze the classification numbers of each patent application and compare whether the classification numbers of different patent applications are the same.

[0068] Constructing an edge: If two patent applications have the same classification number, an edge is created between their corresponding nodes, with the edge type being "same classification number". This indicates that these patent applications have a certain similarity or connection in the technical field, and may involve the same or similar technical topics, which helps to explore potential connections between patents from a technical perspective.

[0069] That is, based on whether there are inventors with the same name, applicants, application dates, and classification numbers among different patent applications under the name of the inventor to be identified, these are used as attributes of different dimensions to construct edges of various types between corresponding nodes in the graph structure.

[0070] Step S23: Based on the nodes in the graph structure and the edges of various types between the corresponding nodes, construct a graph structure based on co-occurrence relationships.

[0071] Step S3: Based on the graph structure of co-occurrence relationships corresponding to each inventor to be identified, identify inventors with the same name as the corresponding inventor to be identified.

[0072] Step S31: Determine whether all nodes in the graph structure are isolated nodes. If so, then the inventors to be identified in the patent application corresponding to each node are all regarded as independent inventors, and there is a name-duplicate relationship among all independent inventors. The process of identifying inventors with duplicate names ends.

[0073] In the specific implementation, the graph structure based on co-occurrence relationships is traversed, and each node is checked to see if it has any edges connecting it to other nodes. If a node has no edges connecting it to other nodes, then that node is an isolated node. Isolated nodes are then processed: for each isolated node, the inventor to be identified in its corresponding patent application is identified as an independent inventor. If all nodes in the graph structure are isolated nodes, then the inventor to be identified in the patent application corresponding to each node is an independent inventor, and the process of identifying inventors with the same name ends.

[0074] Step S32: Determine whether the weights of the edges of applicant types between each node in the graph structure are all greater than or equal to 1. If so, then the inventors to be identified in the patent applications corresponding to all nodes in the graph structure are regarded as the same independent inventors. There are no inventors with the same name, and the process of identifying inventors with the same name ends.

[0075] In practice, if the weight of the applicant type edge between all nodes in the graph structure is greater than or equal to 1, it indicates that these patent applications have a strong correlation with the applicant. In this case, the inventors to be identified for all patent applications corresponding to all nodes in the graph structure are regarded as the same independent inventor. At this time, there are no inventors with the same name, and the process of identifying inventors with the same name ends.

[0076] Step S33: If steps S31 and S32 are not satisfied, a combination of isolated node discrimination, applicant type edge weight discrimination, and graph structure clustering is used to identify inventors with the same name as the corresponding inventors to be identified.

[0077] The specific implementation process is described below.

[0078] Step S331: First, determine whether there are isolated nodes in the graph structure. If so, treat the inventors of the patent applications corresponding to each isolated node as independent inventors, remove the corresponding isolated nodes from the graph structure, update the graph structure, and jump to step S332; otherwise, jump directly to step S332.

[0079] The inventor information corresponding to these isolated nodes is recorded independently, and these isolated nodes are removed from the graph structure for subsequent processing of the remaining nodes.

[0080] Step S332: Determine whether the weights of the edges of applicant types between each node in the graph structure are all greater than or equal to 1. If so, regard the inventors to be identified in the patent applications corresponding to all nodes in the graph structure as the same independent inventor. There is a name-duplicate relationship between the independent inventors determined in steps S331 and S332, and the name-duplicate inventor identification process ends; otherwise, jump to step S333.

[0081] Step S333: Cluster the graph structure and regard the inventors of the patent applications corresponding to the nodes that are clustered into one class as the same independent inventor. There is a name duplication relationship among the independent inventors identified in steps S331, S332 and S333, and the process of identifying inventors with the same name ends.

[0082] In the specific implementation process, graph structure clustering algorithms such as community detection algorithms or spectral clustering algorithms can be selected to implement graph structure clustering.

[0083] In practical applications, appropriate clustering algorithm parameters and settings should be selected based on specific circumstances. Taking community detection algorithms as an example, they can group closely related nodes into the same community based on the connections between nodes in a graph structure. During the clustering process, the weights of different types of edges in the graph structure are considered, such as the weights of edges with co-authored inventors, edges with the same applicant, edges with the same application date, and edges with the same classification number, to comprehensively evaluate the degree of association between nodes.

[0084] After clustering, the inventors of all patent applications corresponding to nodes in each cluster are considered to be the same independent inventor. The inventor information for each cluster is organized and recorded, and the clustering results can be visualized to intuitively present the relationships and groupings between different inventors.

[0085] Finally, the results of identifying duplicate inventors for all inventors are summarized and organized to form a complete report on duplicate inventor identification. The report may include relevant information for each inventor, identification results (whether they are independent inventors), and detailed information on cluster analysis, providing strong data support and decision-making basis for patent management, technology R&D evaluation, and other work. This graph-based automatic identification method for duplicate inventors can efficiently and accurately process large amounts of patent data, solve the problem of identifying duplicate inventors, and improve the efficiency and quality of patent management and analysis.

[0086] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0087] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatically identifying inventors with the same name, characterized in that, The method includes: Using the name of each inventor to be identified as a search keyword, a search was conducted in the full Chinese patent database to extract the bibliographic information of all patent applications under the name of the corresponding inventor. Based on the bibliographic information of all patent applications under each inventor to be identified, a graph structure based on co-occurrence relationships is constructed. Based on the graph structure of co-occurrence relationships corresponding to each inventor to be identified, the system identifies inventors with the same name as the corresponding inventor to be identified.

2. The automatic identification method for inventors with the same name according to claim 1, characterized in that, The construction of the graph structure based on co-occurrence relationships is performed as follows: For each patent application under the name of the inventor to be identified, the application number of the patent application is used as a unique identifier to construct a node in the graph structure; Based on the different attributes of the patent applications under the name of the inventor to be identified, construct the edges of various types between the corresponding nodes in the graph structure; Based on the nodes in the graph structure and the edges of various types between the corresponding nodes, a graph structure based on co-occurrence relationships is constructed.

3. The automatic identification method for inventors with the same name according to claim 2, characterized in that, The process of identifying inventors with the same name as the corresponding inventor to be identified is performed as follows: Determine whether all nodes in the graph structure are isolated nodes. If so, treat the inventors of the patent applications corresponding to each node as independent inventors, and identify inventors with the same name among all independent inventors. The process of identifying inventors with the same name ends.

4. The automatic identification method for inventors with the same name according to claim 2, characterized in that, The process of identifying inventors with the same name as the corresponding inventor to be identified also includes: Determine whether the weights of the edges with applicant types between nodes in the graph structure are all greater than or equal to 1. If so, then the inventors of the patent applications corresponding to all nodes in the graph structure are considered to be the same independent inventors, and there are no inventors with the same name. The process of identifying inventors with the same name ends.

5. The method for automatically identifying inventors with the same name according to claim 2, characterized in that, The process of identifying inventors with the same name as the corresponding inventor to be identified also includes: If not all nodes in the graph structure are isolated nodes, and the weights of the edges of applicant type between nodes in the graph structure are not all greater than or equal to 1, a combination of isolated node discrimination, applicant type edge weight discrimination, and graph structure clustering is used to identify inventors with the same name as the corresponding inventors to be identified.

6. The automatic identification method for inventors with the same name according to claim 5, characterized in that, The method employs a combination of isolated node discrimination, applicant type edge weight discrimination, and graph structure clustering to identify inventors with duplicate names, and performs the following: Step S331: Determine whether there are isolated nodes in the graph structure. If so, treat the inventors of the patent applications corresponding to each isolated node as independent inventors, remove the corresponding isolated nodes from the graph structure, update the graph structure, and jump to step S332; otherwise, jump directly to step S332. Step S332: Determine whether the weights of the edges of applicant types between each node in the graph structure are all greater than or equal to 1. If so, regard the inventors to be identified in the patent applications corresponding to all nodes in the graph structure as the same independent inventor. There is a name-duplicate relationship between the independent inventors determined in steps S331 and S332, and the name-duplicate inventor identification process ends; otherwise, jump to step S333. Step S333: Cluster the graph structure and regard the inventors of the patent applications corresponding to the nodes that are clustered into one class as the same independent inventor. There is a name duplication relationship among the independent inventors identified in steps S331, S332 and S333, and the process of identifying inventors with the same name ends.

7. The method for automatically identifying inventors with the same name according to any one of claims 2-6, characterized in that, The dimensions include: co-signing inventors, same applicants, same application date, and same classification number.

8. The method for automatically identifying inventors with the same name according to claim 7, characterized in that, For the dimension of co-authored inventors, the corresponding type of edges are constructed in the following manner: If two patent applications have inventors who are jointly listed, then an edge is constructed between the corresponding nodes of the two patent applications in the graph structure, and the edge is marked as "jointly listed inventors" type. The number of co-authors is used as the weight of that edge.

9. The method for automatically identifying inventors with the same name according to claim 7, characterized in that, For the same applicant dimension, edges of the corresponding type are constructed in the following manner: If two patent applications have the same applicant, add an edge between the corresponding nodes and mark the type as "same applicant"; The number of identical applicants is used as the weight of that edge.

10. The method for automatically identifying inventors with the same name according to claim 7, characterized in that, For the same application date dimension, edges of the corresponding type are constructed in the following manner: When patent applications with the same application date are found, an edge is created between the nodes corresponding to the respective patent applications, and the edge type is defined as "same application date".

Images