Method, device, equipment and medium for disambiguating personal names

Abstract

The embodiment of the present invention discloses a method, device, equipment and medium for name disambiguation, wherein the method comprises: constructing a university teacher database, wherein the university teacher database comprises university teacher introduction information; extracting the text in the university teacher introduction information by a text extraction algorithm to obtain teacher text information, and disassembling the teacher text information to obtain disassembled text information; extracting key features in the disassembled text information by a large language model, and calculating an overall similarity score based on the key features; inputting the overall similarity score into a gradient boosting tree model for pairwise similarity estimation to obtain a pairwise similarity score; clustering by a clustering algorithm based on the pairwise similarity score to generate a disambiguation result after name disambiguation. The embodiment of the present application can improve the accuracy of name disambiguation.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and medium for disambiguating personal names. Background Art

[0002] The problem of duplicate names has a long history and has always been a hot topic in the fields of information science and computer science. Name disambiguation is an important and challenging task. With the increase in the number of university teachers and the continuous expansion of resume databases, it has become increasingly difficult to accurately identify and distinguish different teachers with the same or similar names. The solution to this problem is of great significance to the accuracy of university resume databases and the construction of scientific research cooperation networks. Existing name disambiguation methods mainly include the following: 1. Rule-based name disambiguation methods, which use predefined rules such as a person's name, contact information, work experience, and educational background to disambiguate names. However, this method relies on artificially set rules and thresholds, lacks flexibility, and has low disambiguation accuracy when processing large amounts of data or more data noise. 2. Name disambiguation methods based on supervised learning models, which are trained through a large amount of labeled data based on supervised learning models to learn how to distinguish different names. Although the accuracy of disambiguation has been significantly improved, it relies on a large amount of labeled data and has high labeling costs. 3. The name disambiguation method based on graph neural network uses graph structure data to disambiguate names. Although graph neural network can effectively process complex relational networks, the training and reasoning processes are computationally intensive and have high hardware requirements. Summary of the invention

[0003] Embodiments of the present invention provide a method, apparatus, device and medium for name disambiguation, aiming to solve the problem of low accuracy of existing name disambiguation.

[0004] In a first aspect, an embodiment of the present invention provides a method for disambiguating a person's name, which includes:

[0005] Constructing a university teacher database, wherein the university teacher database includes university teacher introduction information;

[0006] Extracting the text in the university teacher introduction information by a text extraction algorithm to obtain teacher text information, and disassembling the teacher text information to obtain disassembled text information;

[0007] Extracting key features from the disassembled text information through a large language model, and calculating an overall similarity score based on the key features;

[0008] Inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score;

[0009] Clustering is performed using a clustering algorithm according to the pairwise similarity scores to generate a disambiguation result after the person name is disambiguated.

[0010] In a second aspect, an embodiment of the present invention further provides a device for disambiguating a person's name, which includes:

[0011] A construction unit is used to construct a university teacher database, wherein the university teacher database includes university teacher introduction information;

[0012] An extraction and disassembly unit, used for extracting the text in the university teacher introduction information by using a text extraction algorithm to obtain the teacher text information, and disassembling the teacher text information to obtain disassembled text information;

[0013] An extraction calculation unit, used for extracting key features in the disassembled text information through a large language model, and calculating an overall similarity score based on the key features;

[0014] An estimation unit, used for inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score;

[0015] A clustering unit is used to perform clustering according to the pairwise similarity scores by using a clustering algorithm to generate a disambiguation result after the person name is disambiguated.

[0016] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the computer program.

[0017] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program can implement the above method when executed by a processor.

[0018] The embodiment of the present invention provides a method, device, equipment and medium for name disambiguation. The method includes: constructing a university teacher database, wherein the university teacher database includes university teacher introduction information; extracting the text in the university teacher introduction information through a text extraction algorithm to obtain teacher text information, and disassembling the teacher text information to obtain disassembled text information; extracting key features in the disassembled text information through a large language model, and calculating an overall similarity score based on the key features; inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score; clustering through a clustering algorithm based on the pairwise similarity score to generate a disambiguation result after name disambiguation. The technical solution of the embodiment of the present invention first extracts the text in the introduction information of college teachers through a text extraction algorithm to disassemble it to obtain disassembled text information; then extracts the key features in the disassembled text information through a large language model to calculate the overall similarity score; finally, based on the overall similarity score, pairwise similarity estimation and clustering are performed through a gradient boosting tree model and a clustering algorithm to generate a disambiguation result after name disambiguation, thereby improving the accuracy of name disambiguation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technical users in the field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0020] Figure 1 A flowchart of a method for disambiguating a person's name provided by an embodiment of the present invention;

[0021] Figure 2 A schematic diagram of the process of training and verifying a text segmentation and disassembly model provided in an embodiment of the present invention;

[0022] Figure 3 A schematic diagram of a sub-process of a method for disambiguating a person's name provided by an embodiment of the present invention;

[0023] Figure 4 A schematic diagram of a sub-process of a method for disambiguating a person's name provided by an embodiment of the present invention;

[0024] Figure 5 A schematic diagram of a sub-process of a method for disambiguating a person's name provided by an embodiment of the present invention;

[0025] Figure 6 A schematic block diagram of a device for disambiguating personal names provided by an embodiment of the present invention;

[0026] Figure 7 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technical users in the field without creative work are within the scope of protection of the present invention.

[0028] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0029] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0030] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0031] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0032] See also Figure 1 , Figure 1 is a flowchart of a method for disambiguating a person's name provided by an embodiment of the present invention. The method for disambiguating a person's name provided by an embodiment of the present invention can be applied to a computer device, for example, the method for disambiguating a person's name can be implemented by a software program configured on the computer device. Figure 1 As shown, the method includes the following steps S110-S150.

[0033] S110, constructing a university teacher database, wherein the university teacher database includes university teacher introduction information.

[0034] In an embodiment of the present invention, the web addresses of college teachers are first collected; then the college teacher introduction information in the web addresses of college teachers is captured by the developed crawler tool to construct the college teacher database. It should be noted that in this embodiment, when the college teacher introduction information is extracted by the crawler tool, the robots.txt protocol is followed. Robots.txt (also known as crawler protocol or robot protocol) is a convention used to inform web crawlers and search engines which web pages can be crawled and which web pages should not be crawled. Understandably, the college teacher database may also include thesis information of college teachers.

[0035] S120, extracting the text in the university teacher introduction information through a text extraction algorithm to obtain teacher text information, and disassembling the teacher text information to obtain disassembled text information.

[0036] In an embodiment of the present invention, the text extraction algorithm includes a first text extraction algorithm and a second text extraction algorithm; the extracting the text in the college teacher introduction information by the text extraction algorithm to obtain the teacher text information includes: extracting the text in the college teacher introduction information by the first text extraction algorithm to obtain the teacher text information; or extracting the text in the college teacher introduction information by the first text extraction algorithm to obtain the first teacher text information; extracting the text in the first teacher text information by the second text extraction algorithm to obtain the second teacher text information; using the union of the first teacher text information and the second teacher text information as the teacher text information. It should be noted that in this embodiment, when the text in the college teacher introduction information can be extracted by the first text extraction algorithm to obtain the teacher text information, it indicates that the college teacher introduction information is relatively simple; and when the text in the college teacher introduction information needs to be extracted by the first text extraction algorithm and the second text extraction algorithm to obtain the teacher text information, it indicates that the college teacher introduction information is relatively complex. It should also be noted that, in this embodiment, the first text extraction algorithm is a web page text extraction algorithm based on logical lines and maximum acceptance distance, that is, the content of the college teacher introduction information web page is divided by lines, and which logical lines are most likely to be part of the text according to the line spacing. The algorithm will calculate the acceptance distance between different logical lines to determine the text area, and the acceptance distance takes into account the relationship between lines, such as typesetting, spacing, etc., with the purpose of finding the area that is most likely to constitute the text. Specifically, it includes the following steps: 1. Page parsing: Parse the content of the college teacher introduction information web page into logical lines, which are usually the basic unit of each line of text in the web page. 2. Logical line feature extraction: Calculate the features of each logical line, including font size, line height, margin, etc. These features help to distinguish the text from other content (such as titles, advertisements). 3. Acceptance distance calculation: Calculate the acceptance distance between each logical line. The acceptance distance measures the arrangement and structure of the logical lines, and the acceptance distance takes into account factors such as line spacing and context relevance. 4. Maximum acceptance distance selection: Based on the acceptance distance, determine which logical lines may be part of the text. Generally speaking, teachers' personal homepages on university websites have certain regularities, that is, the content of university teacher introduction information web pages has certain regularities, so there are fewer noises or boundary conditions to consider. 5. Text area extraction: Comprehensively consider the characteristics of logical lines and the acceptance distance, select the logical line area that best meets the text characteristics, and finally form the text part of the web page, that is, obtain the teacher's text information. The second text extraction algorithm is a general web page text extraction algorithm based on block distribution function, which specifically includes the following steps when extracting the text: 1. Page parsing: Parse the web page content into different blocks, which can be different areas of the entire web page, such as title, text, advertisement, navigation bar, etc.2. Block feature extraction: Extract features for each block, including block size, location, content density, etc., to help identify the type of block. 3. Block distribution analysis: Use block distribution function to analyze the distribution pattern of different blocks. 4. Text block identification: Determine which blocks may be the text based on the block characteristics and distribution pattern. Text blocks usually have high content density and continuity, while other blocks may show different characteristics. 5. Text extraction: Extract the text content from the identified text blocks to form the text part of the web page.

[0037] Furthermore, after extracting the text from the university teacher introduction information to obtain the teacher text information, the teacher text information is disassembled to obtain disassembled text information. Specifically, the teacher text information is disassembled by a text segmentation disassembly model to obtain the disassembled text information, wherein the text segmentation disassembly model is a model obtained by training and verifying a large language model fine-tuned by LoRA using a sample teacher text information set. The steps of training and verifying the large language model fine-tuned by LoRA using the sample teacher text information set to obtain the text segmentation disassembly model are as follows: Figure 2As shown, it specifically includes steps S121-S126: S121, annotating the collected sample teacher text information set through a data annotation tool according to a custom information structure to obtain an annotated teacher text information set; S122, dividing the annotated teacher text information set into a teacher text training set and a teacher text verification set according to a preset ratio; S123, inputting the teacher text training set into the large language model after LoRA fine-tuning for training; S124, inputting the teacher text verification set into the trained LoRA fine-tuned large language model for verification to obtain verified disassembled text information; S125, calculating the disassembled text accuracy rate according to the verified disassembled text information and the annotated disassembled text information in the teacher text verification set; S126, if the disassembled text accuracy rate is greater than the preset disassembled text accuracy rate, the trained LoRA fine-tuned large language model is used as the text segmentation disassembly model. Understandably, if the accuracy of the disassembled text is not greater than the preset accuracy of the disassembled text, the large language model after LoRA fine-tuning continues to be trained. It should be noted that in this embodiment, the custom information structure is 1. Basic information: 1.1 Resume photo, 1.2 Name, 1.3 Gender, 1.4 Date of birth, 1.5 Mobile phone number, 1.6 Email, 1.7 Native place; 2. Educational experience: ...; The data annotation tool is the LabelStudio tool, that is, the content in the sample teacher text information set is annotated by the LabelStudio tool according to the fields in the custom information structure for NER (Named Entity Recognition). It should also be noted that in this embodiment, the reason why the large language model after LoRA fine-tuning is trained is that LoRA technology can significantly reduce the number of parameters that need to be stored and updated, while maintaining the performance of the model, so that the large language model can adapt to specific application scenarios or fields without the need to fully retrain the entire model. This method not only reduces the consumption of computing and storage resources, but also speeds up the training speed of the model, so that the model can adapt to new data or tasks more quickly. The disassembled text information is data in JSON format.

[0038] S130: extract key features from the disassembled text information through a large language model, and calculate an overall similarity score based on the key features.

[0039] In an embodiment of the present invention, after the teacher's text information is disassembled to obtain the disassembled text information, the key features in the disassembled text information are extracted through a large language model, wherein the large language model is a BERT (Bidirectional Encoder Representations from Transformers) model or a GPT (Generative Pre-trained Transformer) model. It should be noted that in this embodiment, the key features are features such as name, contact information, work experience, educational background, skills, and project experience.

[0040] Furthermore, after extracting the key features in the disassembled text information, the overall similarity score is calculated based on the key features, wherein the key features include numerical features and text features; Figure 3 As shown, it specifically includes steps S131-S134: S131, preprocessing the numerical features and the text features respectively to obtain target numerical features and target text features; S132, normalizing the target numerical features to obtain normalized numerical features; S133, vectorizing the target text features to obtain vector text features; S134, calculating feature similarity based on the normalized numerical features and the vector text features by a similarity method, and calculating an overall similarity score based on the feature similarity. It should be noted that, in the present embodiment, the preprocessing is to remove blank items in the numerical features and the text features and to unify the formats of the numerical features and the text features. It should also be noted that, in the present embodiment, the target text features are vectorized by a large language model to obtain the vector text features.

[0041] Furthermore, the feature similarity algorithm includes a cosine similarity algorithm and a Jaccard similarity algorithm; the feature similarity includes a numerical feature similarity and a text feature similarity; Figure 4As shown, step S134 specifically includes steps S1341-S1343: S1341, calculating the similarity between the normalized numerical features by the Jaccard similarity algorithm to obtain the numerical feature similarity; S1342, calculating the similarity between the vector text features by the cosine similarity algorithm to obtain the text feature similarity; S1343, performing weighted summation of the numerical feature similarity and the text feature similarity to obtain the overall similarity score. It should be noted that, in the present embodiment, the weight coefficients of the numerical feature similarity and the text feature similarity can be set manually or according to a decision tree model. It should also be noted that, in the present embodiment, in the process of determining the weight coefficient, the weight coefficient corresponding to the overall similarity score being greater than the overall similarity score threshold is used as the weight coefficient for weighted summation.

[0042] S140, inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score.

[0043] In an embodiment of the present invention, after obtaining the overall similarity score, the overall similarity score is input into a gradient boosting tree model for pairwise similarity estimation to obtain a pairwise similarity score, wherein the gradient boosting tree model is a pre-trained model, and during the training process, the training parameters are continuously adjusted so that the pairwise similarity score is greater than the pairwise similarity score threshold. It should be noted that in this embodiment, the training parameters are learning rate, number of iterations, maximum depth, subsampling rate, minimum sample splitting tree, minimum number of leaf node samples, maximum number of features and other parameters. During the model training process, the interval of the approximate parameters will be determined based on experience, and then automatically adjusted through cross-validation and grid search. Parameter design is mainly to prevent overfitting and underfitting.

[0044] S150, performing clustering by using a clustering algorithm according to the pairwise similarity scores to generate a disambiguation result after name disambiguation.

[0045] In an embodiment of the present invention, clustering is performed by a clustering algorithm according to the pairwise similarity score to generate a disambiguation result after name disambiguation. It can be understood that the introduction information of college teachers with high similarity is clustered together to form multiple resume sets. It should be noted that in this embodiment, the clustering algorithm is the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm, and the DBSCAN clustering algorithm is a density-based spatial clustering algorithm. It can find clusters of any shape and identify noise points. It should also be noted that in this embodiment, during the training of the DBSCAN clustering algorithm, it is necessary to continuously adjust eps (neighborhood radius) and min-samples (minimum number of samples) by detecting whether the clustering accuracy is greater than the clustering accuracy threshold.

[0046] Figure 5 A flowchart of a method for disambiguating a person's name provided by another embodiment of the present invention is shown in FIG. Figure 6 As shown, in this embodiment, the method includes steps S110-S180. That is, in this embodiment, after step S150 of the above embodiment, the method further includes steps S160-S180.

[0047] S160, saving the disambiguation result to the university teacher database;

[0048] S170, querying the disambiguation result in the university teacher database through a preset query interface;

[0049] S180. If it is detected that the university teacher introduction information in the university teacher database has changed, the step of extracting the text in the university teacher introduction information by a text extraction algorithm to obtain the teacher text information is executed.

[0050] In an embodiment of the present invention, after the disambiguation result is generated, the disambiguation result is saved in the university teacher database; an external system can query the disambiguation result in the university teacher database through a preset query interface; it can be understood that the university teacher database is monitored in real time, and if it is monitored that the university teacher introduction information in the university teacher database changes, the step of extracting the text in the university teacher introduction information through a text extraction algorithm to obtain the teacher text information is executed, and the disambiguation result is updated in time to ensure the real-time and accuracy of the data.

[0051] In summary, the name disambiguation method provided by this embodiment has the following advantages: 1. Improved accuracy of name disambiguation: By introducing multiple key features to calculate numerical feature similarity and text feature similarity, and by using a gradient boosting tree model to estimate pairwise similarity, and by clustering through a clustering algorithm, it is possible to more accurately identify and merge repeated or similar name entries; 2. Improved efficiency of name disambiguation: The gradient boosting tree model is used to improve the calculation speed of pairwise similarity estimation, and combined with a clustering algorithm, it is possible to complete the disambiguation processing of large-scale data in a relatively short time, significantly improving the efficiency of name disambiguation; 3. Enhanced flexibility and scalability: The system design adopts a modular architecture, which allows flexible adjustment of feature extraction and similarity calculation methods according to specific needs, and adapts to different data sets and application scenarios. At the same time, the system supports the processing of cross-industry and cross-regional data, and has good scalability; 4. Low cost. It should be noted that the implementation of the name disambiguation method provided in this embodiment also requires high-performance computer equipment and large-capacity storage devices. The close cooperation of hardware facilities and software methods ensures efficient operation when processing large amounts of data.

[0052] Figure 6 FIG. 2 is a schematic block diagram of a name disambiguation device 200 provided in an embodiment of the present invention. Figure 6 As shown, corresponding to the above name disambiguation method, the present invention also provides a name disambiguation device 200. The name disambiguation device 200 includes a unit for executing the above name disambiguation method. Figure 6 The name disambiguation device 200 includes a construction unit 201, an extraction and disassembly unit 202, an extraction and calculation unit 203, an estimation unit 204 and a clustering unit 205.

[0053] Among them, the construction unit 201 is used to construct a university teacher database, wherein the university teacher database includes university teacher introduction information; the extraction and disassembly unit 202 is used to extract the text in the university teacher introduction information through a text extraction algorithm to obtain the teacher text information, and disassemble the teacher text information to obtain the disassembled text information; the extraction and calculation unit 203 is used to extract the key features in the disassembled text information through a large language model, and calculate the overall similarity score based on the key features; the estimation unit 204 is used to input the overall similarity score into the gradient boosting tree model for pairwise similarity estimation to obtain a pairwise similarity score; the clustering unit 205 is used to perform clustering through a clustering algorithm according to the pairwise similarity score to generate a disambiguation result after name disambiguation.

[0054] In some embodiments, such as the present embodiment, the construction unit 201 includes a collection unit and a grabbing construction unit.

[0055] Among them, the collection unit is used to collect the web page addresses of university teachers; the crawling and construction unit is used to crawl the university teacher introduction information in the university teacher web page addresses through the developed crawler tool to construct the university teacher database.

[0056] In some embodiments, such as the present embodiment, the extraction and disassembly unit 202 includes a first extraction unit and a disassembly unit.

[0057] Among them, the first extraction unit is used to extract the text in the university teacher introduction information through the first text extraction algorithm to obtain the teacher text information; or extract the text in the university teacher introduction information through the first text extraction algorithm to obtain the first teacher text information; extract the text in the first teacher text information through the second text extraction algorithm to obtain the second teacher text information; take the union of the first teacher text information and the second teacher text information as the teacher text information; the disassembly unit is used to disassemble the teacher text information through the text segmentation disassembly model to obtain the disassembled text information, wherein the text segmentation disassembly model is a model obtained by training and verifying the large language model after LoRA fine-tuning with a sample teacher text information set.

[0058] In some embodiments, such as the present embodiment, the extraction calculation unit 203 includes a preprocessing unit, a normalization unit, a vectorization unit, and a calculation unit.

[0059] Among them, the preprocessing unit is used to preprocess the numerical features and the text features respectively to obtain target numerical features and target text features; the normalization unit is used to normalize the target numerical features to obtain normalized numerical features; the vectorization unit is used to vectorize the target text features to obtain vector text features; the calculation unit is used to calculate feature similarity based on the normalized numerical features and the vector text features through a similarity method, and calculate an overall similarity score based on the feature similarity.

[0060] In some embodiments, such as this embodiment, the calculation unit includes a first calculation subunit, a second calculation subunit, and a weighted summation unit.

[0061] Among them, the first calculation subunit is used to calculate the similarity between the normalized numerical features through the Jaccard similarity algorithm to obtain the numerical feature similarity; the second calculation subunit is used to calculate the similarity between the vector text features through the cosine similarity algorithm to obtain the text feature similarity; the weighted summation unit is used to perform weighted summation of the numerical feature similarity and the text feature similarity to obtain the overall similarity score.

[0062] In some embodiments, such as the present embodiment, the name disambiguation device 200 further includes a storage unit, a query unit, and a monitoring execution unit.

[0063] Among them, the saving unit is used to save the disambiguation result to the university teacher database; the query unit is used to query the disambiguation result in the university teacher database through a preset query interface; the detection execution unit is used to execute the step of extracting the text in the university teacher introduction information through a text extraction algorithm to obtain the teacher text information if it is monitored that the university teacher introduction information in the university teacher database has changed.

[0064] It should be noted that technical users in the relevant field can clearly understand that the specific implementation process of the above-mentioned name disambiguation device 200 and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, it will not be repeated here.

[0065] The above-mentioned name disambiguation device can be implemented in the form of a computer program. The computer program can be used in Figure 7 Runs on the computer device shown.

[0066] See also Figure 7 , Figure 7 900 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 900 is a device with a name disambiguation function.

[0067] See also Figure 7 The computer device 900 includes a processor 902 , a memory and an interface 905 connected via a system bus 901 , wherein the memory may include a storage medium 903 and an internal memory 904 .

[0068] The storage medium 903 may store an operating system 9031 and a computer program 9032. When the computer program 9032 is executed, the processor 902 may execute the above-mentioned method for disambiguating a person's name.

[0069] The processor 902 is used to provide computing and control capabilities to support the operation of the entire computer device 900 .

[0070] The internal memory 904 provides an environment for the operation of the computer program 9032 in the storage medium 903. When the computer program 9032 is executed by the processor 902, the processor 902 can execute a method for disambiguating a person's name.

[0071] The interface 905 is used to communicate with other devices. A user skilled in the art will appreciate that Figure 7The structure shown in the figure is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device 900 to which the solution of the present application is applied. The specific computer device 900 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0072] The processor 902 is used to run a computer program 9032 stored in the memory to implement any embodiment of the above-mentioned name disambiguation method.

[0073] It should be understood that in the embodiment of the present application, the processor 902 may be a central processing unit (CPU), and the processor 902 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0074] A user of ordinary skill in the art can understand that all or part of the processes in the method for implementing the above embodiment can be completed by instructing related hardware through a computer program. The computer program can be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the wireless communication system to implement the process steps of the embodiment of the above method.

[0075] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program. When the computer program is executed by a processor, the processor executes any embodiment of the above-mentioned method for disambiguating a person's name.

[0076] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0077] A user of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, wireless communication software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical users can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0078] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0079] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0080] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technical user familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present invention, and these modifications or substitutions should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for disambiguating a person's name, characterized in that: include: Constructing a university teacher database, wherein the university teacher database includes university teacher introduction information; Extracting the text in the university teacher introduction information by a text extraction algorithm to obtain teacher text information, and disassembling the teacher text information to obtain disassembled text information; Extracting key features from the disassembled text information through a large language model, and calculating an overall similarity score based on the key features; Inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score; Clustering is performed using a clustering algorithm according to the pairwise similarity scores to generate a disambiguation result after the person name is disambiguated.

2. The method for disambiguating personal names according to claim 1, characterized in that: The text extraction algorithm includes a first text extraction algorithm and a second text extraction algorithm; the text extraction algorithm is used to extract the text in the university teacher introduction information to obtain the teacher text information, including: Extracting the text in the university teacher introduction information by using the first text extraction algorithm to obtain the teacher text information; or Extracting the text in the university teacher introduction information by the first text extraction algorithm to obtain the first teacher text information; Extracting the text in the first teacher text information by the second text extraction algorithm to obtain the second teacher text information; The union of the first teacher text information and the second teacher text information is used as the teacher text information.

3. The method for disambiguating a person's name according to claim 1, characterized in that: The step of disassembling the teacher's text information to obtain disassembled text information includes: Decomposing the teacher text information by a text segmentation and decomposition model to obtain decomposed text information, wherein the text segmentation and decomposition model is a model obtained by training and verifying a large language model fine-tuned by LoRA using a sample teacher text information set; The steps of training and verifying the large language model fine-tuned by LoRA through the sample teacher text information set to obtain the text segmentation model include: Annotating the collected sample teacher text information set by using a data annotation tool according to a custom information structure to obtain an annotated teacher text information set; Dividing the annotated teacher text information set into a teacher text training set and a teacher text verification set according to a preset ratio; Inputting the teacher text training set into the large language model fine-tuned by LoRA for training; Input the teacher text verification set into the large language model after the training LoRA fine-tuning for verification to obtain verification and disassembly text information; Calculate the accuracy of the disassembled text according to the verified disassembled text information and the annotated disassembled text information in the teacher text verification set; If the accuracy of the disassembled text is greater than the preset accuracy of the disassembled text, the large language model after the trained LoRA fine-tuning is used as the text segmentation and disassembly model.

4. The method for disambiguating a person's name according to claim 1, characterized in that: The key features include numerical features and text features; and calculating the overall similarity score based on the key features includes: Preprocessing the numerical features and the text features respectively to obtain target numerical features and target text features; Normalizing the target numerical feature to obtain a normalized numerical feature; Performing vectorization processing on the target text feature to obtain a vector text feature; The feature similarity is calculated according to the normalized numerical feature and the vector text feature by a similarity method, and the overall similarity score is calculated according to the feature similarity.

5. The method for disambiguating a person's name according to claim 4, characterized in that: The feature similarity algorithm includes a cosine similarity algorithm and a Jaccard similarity algorithm; the feature similarity includes a numerical feature similarity and a text feature similarity; the feature similarity is calculated by a similarity method based on the normalized numerical feature and the vector text feature, and an overall similarity score is calculated based on the feature similarity, including: Calculating the similarity between the normalized numerical features by using the Jaccard similarity algorithm to obtain the numerical feature similarity; Calculate the similarity between the vector text features by using the cosine similarity algorithm to obtain the text feature similarity; The weighted sum of the numerical feature similarity and the text feature similarity is performed to obtain the overall similarity score.

6. The method for disambiguating a person's name according to claim 1, characterized in that: The construction of the university teacher database includes: Collect the web page addresses of university teachers; The university teacher introduction information in the university teacher webpage address is captured by the developed crawler tool to build the university teacher database.

7. The method for disambiguating a person's name according to any one of claims 1 to 6, characterized in that: The method further comprises: Saving the disambiguation result to the university teacher database; Query the disambiguation result in the university teacher database through a preset query interface; If it is monitored that the university teacher introduction information in the university teacher database changes, the step of extracting the text in the university teacher introduction information by using a text extraction algorithm to obtain the teacher text information is executed.

8. A device for disambiguating personal names, characterized in that: include: A construction unit is used to construct a university teacher database, wherein the university teacher database includes university teacher introduction information; An extraction and disassembly unit, used for extracting the text in the university teacher introduction information by using a text extraction algorithm to obtain the teacher text information, and disassembling the teacher text information to obtain disassembled text information; An extraction calculation unit, used for extracting key features in the disassembled text information through a large language model, and calculating an overall similarity score based on the key features; An estimation unit, used for inputting the overall similarity score into a gradient boosting tree model to perform pairwise similarity estimation to obtain a pairwise similarity score; A clustering unit is used to perform clustering according to the pairwise similarity scores by using a clustering algorithm to generate a disambiguation result after the person name is disambiguated.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 can be implemented.

Images