Enterprise employee data query method and device, equipment, medium and product

Abstract

The invention provides an enterprise employee data query method and device, equipment, a medium and a product, and relates to the field of enterprise employee information management. Comprising the steps of obtaining an original natural language query text corresponding to a to-be-queried enterprise employee; according to the trained intention analysis large model, performing first analysis processing on the original natural language query text to obtain a first analysis result; according to a trained semantic constraint large model, performing second analysis processing on the first analysis result to obtain a second analysis result; determining a target query vector according to the first analysis result and the second analysis result; and according to the target query vector, performing query in a preset employee information knowledge graph to obtain a query result of the to-be-queried enterprise employee. The technical problem that the query efficiency of enterprise employee data is low due to the fact that polysemy words cannot be processed and the query process is tedious in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of enterprise employee information management, and in particular to a method, apparatus, equipment, medium and product for querying enterprise employee data. Background Technology

[0002] With the rapid development of the internet, existing enterprise internal employee information management systems often have employee information scattered across multiple databases (such as contact list databases, project databases, training record databases, etc.). To meet the needs of enterprises and users, enterprise internal employee information management systems need to provide functions such as quickly finding specific information, performing multi-condition queries using natural language, and ensuring that query results are up-to-date.

[0003] In existing enterprise internal employee information management systems, the employee information search function is based on keyword matching to extract conditions, and then queries are performed in a scattered database according to the extracted conditions.

[0004] However, existing technologies cannot handle polysemous words, and the query process is cumbersome, resulting in low query efficiency for enterprise employee data. Summary of the Invention

[0005] This application provides a method, apparatus, equipment, medium, and product for querying enterprise employee data, in order to solve the problem that the existing technology cannot handle polysemous words and the query process is cumbersome, resulting in low query efficiency for enterprise employee data.

[0006] Firstly, this application provides a method for querying enterprise employee data, including:

[0007] Obtain the original natural language query text corresponding to the employees of the company to be queried;

[0008] Based on the trained intent parsing model, the original natural language query text is subjected to a first parsing process to obtain the first parsing result. The first parsing result includes the core intent, key entity list, inter-entity relationship and confidence score corresponding to the original natural language query text. The trained intent parsing model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationship and confidence score corresponding to the natural language query text samples.

[0009] Based on the trained semantic constraint model, the first parsing result is subjected to a second parsing process to obtain the second parsing result; wherein, the second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result;

[0010] Based on the first and second parsing results, the target query vector is determined; wherein, the target query vector includes the target entity and / or the target relationship;

[0011] Based on the target query vector, a query is performed in the preset employee information knowledge graph to obtain the query results for the employees of the enterprise to be queried.

[0012] In one possible design, the target query vector is determined based on the first and second parsing results, including:

[0013] Based on the confidence score, determine whether to perform the disambiguation process;

[0014] If it is determined that the disambiguation process will be executed, then the weighted score value corresponding to each first analysis result will be determined based on the first analysis result and the second analysis result.

[0015] Based on the weighted score, the target query vector is selected from the first analysis result.

[0016] In one possible design, based on the first and second analysis results, a weighted score value corresponding to each first analysis result is determined, including:

[0017] Based on the first analysis result, a matching is performed in the preset employee information knowledge graph to obtain the corresponding candidate entities;

[0018] Based on the second analysis result, the candidate entities are subjected to multi-dimensional scoring processing to obtain scores in multiple preset dimensions;

[0019] The scores of multiple preset dimensions are weighted to obtain the weighted score value corresponding to the first analysis result.

[0020] In one possible design, the preset dimensions include multiple of the following: literal matching degree, entity semantic similarity, context network similarity, and dynamic weight of association relationship.

[0021] In one possible design, before obtaining the raw natural language query text corresponding to the employees of the company to be queried, the following steps are also included:

[0022] Acquire employee data for all company employees; this data includes multiple items such as name, gender, job title, department, position, educational background, industry information, and certifications.

[0023] To obtain the relationships between enterprise employee data;

[0024] Based on enterprise employee data and relationships, a pre-defined employee information knowledge graph is established; where the entities in the pre-defined employee information knowledge graph are enterprise employee data, and the edges of the pre-defined employee information knowledge graph represent the relationships between entities.

[0025] In one possible design, the edges in the pre-defined employee information knowledge graph are configured with weights;

[0026] After establishing a pre-defined employee information knowledge graph based on enterprise employee data and relationships, the following is also included:

[0027] Update the weights of edges in the preset employee information knowledge graph according to the preset time rules.

[0028] In one possible design, based on the target query vector, a query is performed in a pre-defined employee information knowledge graph to obtain the query results for the employees of the enterprise to be queried, including:

[0029] Based on the target query vector, a query is performed in the preset employee information knowledge graph to obtain employee information from multiple candidate companies;

[0030] According to the preset sorting strategy, the employee information of multiple candidate companies is sorted.

[0031] Based on the sorting results, determine the query results for the employees of the company to be queried.

[0032] In one possible design, after determining the query results for the employees of the company to be queried based on the sorting results, the following steps are also included:

[0033] Generate and display information based on the query results for the employees of the company to be queried;

[0034] The information is displayed on the preset display interface.

[0035] Secondly, this application provides a device for querying enterprise employee data, comprising:

[0036] The acquisition module is used to acquire the original natural language query text corresponding to the employees of the enterprise to be queried;

[0037] The first parsing module is used to perform first parsing processing on the original natural language query text based on the trained intent parsing model to obtain the first parsing result. The first parsing result includes the core intent, key entity list, inter-entity relationship and confidence score corresponding to the original natural language query text. The trained intent parsing model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationship and confidence score corresponding to the natural language query text samples.

[0038] The second parsing module is used to perform a second parsing process on the first parsing result based on the trained semantic constraint model to obtain the second parsing result; wherein, the second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result;

[0039] The determination module is used to determine the target query vector based on the first parsing result and the second parsing result; wherein the target query vector includes the target entity and / or the target relationship;

[0040] The query module is used to perform a query in a preset employee information knowledge graph based on the target query vector to obtain the query results of the employees of the enterprise to be queried.

[0041] Thirdly, this application provides an electronic device, including: a memory and a processor;

[0042] The memory stores instructions that the computer executes;

[0043] The processor executes computer execution instructions stored in memory, causing the processor to perform a method for querying enterprise employee data as described in the first aspect of the invention.

[0044] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement a method for querying enterprise employee data as described in the first aspect of the invention.

[0045] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for querying enterprise employee data as described in the first aspect of the invention.

[0046] This application provides a method, apparatus, device, medium, and product for querying enterprise employee data. The method involves acquiring the original natural language query text corresponding to the enterprise employee to be queried; performing a first parsing process on the original natural language query text using a trained intent parsing model to obtain a first parsing result; performing a second parsing process on the first parsing result using a trained semantic constraint model to obtain a second parsing result; determining the target query vector based on the first and second parsing results; and querying the target query vector within a pre-defined employee information knowledge graph to obtain the query results for the enterprise employee to be queried. Compared to existing technologies that cannot handle polysemous words and have cumbersome query processes, resulting in low query efficiency for enterprise employee data, this application's intelligent enterprise employee search method, based on large-scale model and knowledge graph technology, combines the advantages of both to achieve fast, accurate, and flexible retrieval of enterprise employee information, thereby improving the query efficiency of enterprise employee data. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1A schematic flowchart of a prior art method for querying enterprise employee data provided in this application embodiment;

[0049] Figure 2 A schematic diagram of the system architecture for a method of querying enterprise employee data provided in this application embodiment;

[0050] Figure 3 A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 1 ;

[0051] Figure 4 A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 2 ;

[0052] Figure 5 A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 3 ;

[0053] Figure 6 A schematic diagram of the structure of the enterprise employee data query device provided in the embodiments of this application;

[0054] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0056] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with substantially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, nor do they necessarily imply difference. It should be noted that in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner. In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more.

[0057] It should be noted that in the embodiments of the present application, "when...", can be at the instant when a certain situation occurs, or within a period of time after a certain situation occurs. The embodiments of the present application do not make specific limitations on this. In addition, a query method for enterprise employee data provided by the embodiments of the present application is only an example, and the query method for enterprise employee data may also include more or less content.

[0058] With the rapid development of the Internet, in order to meet the needs of enterprises and users, the existing internal employee information management system of enterprises faces the following challenges:

[0059] Optionally, cross-departmental collaboration requirements: The information of enterprise employees is scattered in multiple databases (such as address book databases, project databases, training record databases, etc.), and it is necessary to quickly find employees with specific departments, positions or skills to support team formation or task assignment.

[0060] Optionally, complex condition queries: Users need to perform multi-condition combination queries through natural language (such as "find employees in department A who have participated in the cloud platform project and know Python"), and the existing system is difficult to handle fuzzy references (such as "Lao Wang"), polysemous words (such as "Java") or implicit relationships (such as "knowing Python" implies a skill relationship).

[0061] Optionally, dynamic information update: Information such as the positions and project participation of enterprise employees changes frequently, and it is necessary to ensure that the query results reflect the latest status to avoid making wrong decisions due to outdated data.

[0062] Optionally, decision-making support requirements: Management needs to optimize resource allocation through employee information analysis (such as skill distribution and project participation), and traditional systems lack in-depth analysis capabilities.

[0063] In the prior art, after the user inputs a query through the interface, the system first performs keyword parsing to clarify the query intention, and then concurrently executes the address book database query and the data analysis competition database query, realizes cross-data source collaboration through a modular data flow, and finally integrates the filtered results into a unified result set through a merging module and displays it.

[0064] Specifically, Figure 1 is a schematic flow chart of the query method for enterprise employee data in the prior art provided by the embodiments of the present application, as Figure 1 shown. Taking the example of finding employees in department A whose names contain the character "Wang" and who have participated in the data analysis competition, the specific operation process is as follows:

[0065] First, perform keyword parsing: After the user inputs a natural language query, the system extracts conditions (such as "Wang" and "data analysis competition") through pre-defined keyword matching rules, but cannot understand fuzzy references or complex logics (such as "Lao Wang" or "knowing Python").

[0066] Secondly, perform multi-database queries:

[0067] Optional, address book database: filter employees in Department A, and further filter out people whose names contain "Wang".

[0068] Optional, Data Analysis Competition Database: Query the list of participants in the competition separately.

[0069] Finally, the results are merged: the query results from the two databases are intersected to find the people who meet both conditions.

[0070] Therefore, in existing technologies, employee information search functions are mostly based on keyword matching, which cannot effectively handle fuzzy queries or queries with complex conditions. At the same time, employee information is usually scattered across different databases, lacking effective integration, resulting in low search efficiency.

[0071] Specifically, existing methods for querying employee data in network slicing companies have the following main shortcomings:

[0072] On the one hand, the query efficiency is low: it requires executing independent queries across multiple databases and then merging the results, which is a cumbersome and time-consuming process.

[0073] On the one hand, fuzzy search support is poor: keyword parsing of user queries is not intelligent enough, it cannot accurately understand compound query conditions, cannot handle polysemous words (such as "Java" referring to language / location / coffee), vague references (such as "Old Wang" and "cloud platform project"), omitted or implicit relationships (such as "find someone in Department A who knows Python" which implies a skill relationship), and it is difficult to accurately disambiguate, leading to misjudgment or missed detection.

[0074] On the one hand, there is a lack of contextual understanding: the understanding of entities in the query (such as person name, project name, certificate name) lacks the rich context provided by the enterprise knowledge graph (such as the employee's department, historical projects, certificate level), resulting in isolated parsing.

[0075] On the one hand, data integration is difficult: employee information is scattered across different databases, lacking a unified data structure, resulting in high integration costs.

[0076] On the one hand, the results are sorted irrationally: query results are usually sorted by simple matching degree, without considering key factors such as job level and certificate level, resulting in low practicality.

[0077] On the other hand, there is the issue of timeliness: static knowledge graphs do not consider the timeliness of relationships (such as the "freshness" of project cooperation relationships), leading to outdated information interfering with the results.

[0078] To address the aforementioned problems, the inventors, during their research on the low efficiency of querying enterprise employee data, discovered that existing technologies cannot handle polysemous words and have cumbersome query processes, resulting in low query efficiency. Therefore, the inventors considered an intelligent enterprise employee search method based on large-scale modeling and knowledge graph technologies. By combining the advantages of both, this method aims to achieve fast, accurate, and flexible retrieval of enterprise employee information. Based on this, embodiments of this application provide a method, apparatus, device, medium, and product for querying enterprise employee data, applicable to the field of enterprise employee information management, aiming to solve the problem of low query efficiency for enterprise employee data in existing technologies.

[0079] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0080] Figure 2 This is a schematic diagram of the system architecture for a method of querying enterprise employee data provided in an embodiment of this application. The enterprise employee data query system is a computer device. Figure 2 In the above architecture, at least one of data acquisition device 201, processing device 202 and display device 203 is included.

[0081] It is understood that the structure illustrated in the embodiments of this application does not constitute a specific limitation on the processing system architecture of the query method for enterprise employee data. In other feasible embodiments of this application, the above architecture may include more or fewer components than illustrated, or combine some components, or divide some components, or arrange different components, which can be determined according to the actual application scenario and is not limited here. Figure 2 The components shown can be implemented in hardware, software, or a combination of both.

[0082] In the specific implementation process, the data acquisition device 201 may include an input / output interface or a communication interface. The data acquisition device 201 can be connected to the processing device through the input / output interface or the communication interface to obtain the original natural language query text corresponding to the enterprise employee to be queried and the trained intent parsing big model.

[0083] The processing device 202 can process the original natural language query text of the enterprise employees to be queried according to the trained intent parsing big model, obtain the target query vector, and perform a query in the preset employee information knowledge graph to obtain the query results of the enterprise employees to be queried.

[0084] The display device 203 can also be a touch screen or the screen of a terminal device, used to receive user commands while displaying the above-mentioned content, so as to realize interaction with the user.

[0085] It should be understood that the aforementioned processing device can be implemented by a processor reading instructions from memory and executing those instructions, or it can be implemented by a chip circuit.

[0086] Furthermore, the network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0087] The technical solution of this application will be described in detail below with reference to specific embodiments:

[0088] Figure 3 A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 1 ,like Figure 3 As shown, the method includes:

[0089] S301. Obtain the original natural language query text corresponding to the employees of the enterprise to be queried.

[0090] S302. Based on the trained intent parsing model, perform the first parsing process on the original natural language query text to obtain the first parsing result.

[0091] The first parsing result includes the core intent, key entity list, inter-entity relationships, and confidence score corresponding to the original natural language query text; the trained intent parsing model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationships, and confidence scores corresponding to the natural language query text samples.

[0092] In this embodiment, a pre-trained intent parsing model is used as the parser. The user's original natural language query text is input for initial parsing. The pre-trained intent parsing model outputs a structured or semi-structured intent representation.

[0093] The intent is represented by the first parsing result.

[0094] Specifically, this representation includes at least:

[0095] Optional, core intent: such as "find employees", "screen employees with a certain skill", "find employees who participated in a certain project", etc. (predefined or model generated).

[0096] Optional, list of key entities: identified employee names, department names, project names, certificate names, skill names, etc.

[0097] Optional, relationships / constraints between entities: such as "belongs to Department A", "knows Java", "holds PMP certificate", "has participated in cloud platform projects".

[0098] Optional, confidence score: The confidence score of the model for each identified entity, relation and its association (usually available internally by the model).

[0099] For example, training intent parsing large models:

[0100] First, a large-scale annotated corpus was constructed, collecting natural language query text samples covering core intents such as "find employees" and "filter employees with skills". The core intent types, key entity lists (such as employee name / department / project / certificate / skill), entity relationship constraints (such as "belongs to department A" "knows Java"), and confidence score benchmark values ​​were manually annotated.

[0101] Secondly, a pre-trained language model is used as the basic architecture, and three sub-modules—intent classification, entity recognition, and relation extraction—are integrated through a multi-task learning framework. The intent classification module uses a softmax layer to output the core intent probability distribution, entity recognition uses the BIO annotation mode combined with a CRF layer to extract key entities, and the relation extraction module matches entity pairs through a dual-tower structure and outputs the relation type and confidence level.

[0102] Furthermore, during the training phase, a joint loss function is used for optimization, which integrates cross-entropy loss (intent classification), CRF loss (entity recognition), and contrastive learning loss (relationship similarity), and a dynamic learning rate adjustment strategy is set to prevent overfitting.

[0103] Then, a confidence calibration mechanism is introduced, with a temperature scaling module connected after the model output layer. The temperature parameters are adjusted through the validation set to make the prediction confidence positively correlated with the true accuracy.

[0104] Finally, the optimal model is determined by the early stopping method (monitoring the F1 score of the validation set). When outputting the structured intent representation, entity-level confidence scores are generated simultaneously to ensure that the first parsing result includes the core intent, key entities, entity relationships, and interpretable confidence assessment, supporting accurate decision-making in the subsequent disambiguation process.

[0105] In one possible implementation, a few-shot learning framework is coupled into the large model of intent parsing, and the large model is fine-tuned using a small amount of labeled data to enhance its ability to parse enterprise-specific terms (such as "cloud platform project").

[0106] Among these improvements, reducing reliance on large-scale labeled data enables the model to accurately parse internal enterprise terminology even in low-resource scenarios, further enhancing the accuracy of intent parsing.

[0107] S303. Based on the trained semantic constraint model, perform a second parsing process on the first parsing result to obtain the second parsing result.

[0108] The second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result.

[0109] In this embodiment, a pre-trained semantically constrained large model is used to perform a second parsing process on the first parsing result: all entities (nodes) and relationships (edges) in the constructed enterprise employee knowledge graph are mapped to a low-dimensional dense vector space. These vectors capture rich semantic and structural information (such as entity similarity and relationship patterns) in the graph. The second parsing result is then calculated: the semantic similarity between the entities / relationships identified in the query and the candidate entities / relationships in the graph.

[0110] For example, training a large model with semantic constraints:

[0111] First, a vector representation space for the enterprise employee knowledge graph is constructed. Knowledge graph embedding technology is used to map entities (employees / departments / projects / skills / certificates) and relationships ("belongs to", "holds", "participates") to low-dimensional dense vectors, through negative sampling. Comparative loss optimization of inter-entity relationship constraints (e.g., "Employee A") belong The vector distance to "department B" must be less than that of random entity pairs.

[0112] Secondly, a pre-trained language model is introduced to encode the natural language description in the first parsing result (such as "find employees who know Java and hold PMP certificates"), and the text features are aligned with the knowledge graph vector through a dual-tower structure.

[0113] Furthermore, a multi-task learning framework is adopted during the training phase, which integrates knowledge graph triple integrity loss (to ensure the preservation of graph structure), semantic similarity loss (to calculate the cosine similarity between query entities and graph entities), and relation consistency loss (to verify whether the relations between entities conform to graph constraints).

[0114] Then, a dynamic negative sampling strategy is introduced to generate difficult negative samples based on entity popularity, thereby improving the model's ability to distinguish low-frequency entities; and semantic similarity scores are calibrated by temperature scaling so that high-confidence query results correspond to higher vector similarity.

[0115] Finally, the optimal model parameters are determined through validation set evaluation. When outputting the second parsing result, a knowledge graph semantic similarity matrix is ​​generated simultaneously to support accurate matching and disambiguation decision-making of subsequent candidate entities, realizing lossless mapping and efficient similarity calculation from natural language query to knowledge graph semantic space.

[0116] In one possible embodiment, a graph neural network is coupled into a large semantically constrained model to enhance the semantic representation of entities / relationships through multi-hop neighbor information aggregation.

[0117] Among them, the context-aware capability of graph neural networks makes entity disambiguation more accurate and improves the resolution accuracy of fuzzy references (such as "Old Wang").

[0118] S304. Determine the target query vector based on the first and second parsing results.

[0119] The target query vector includes the target entity and / or target relationship.

[0120] S305. Based on the target query vector, perform a query in the preset employee information knowledge graph to obtain the query results of the employees of the enterprise to be queried.

[0121] It should be noted that the following steps are included before step S301:

[0122] First, obtain employee data for all company employees.

[0123] The data on company employees includes multiple items such as name, gender, job title, department, position, educational background, industry information, and certification.

[0124] In this embodiment, information on all employees within the enterprise is collected and integrated, including but not limited to name, gender, job level, department, position, educational background, industry information, and qualification certificates, thereby obtaining enterprise employee data for all employees of the enterprise.

[0125] Secondly, it involves understanding the relationships between data on company employees.

[0126] Specifically, the relationships between enterprise employee data include entities (such as employees, departments, and certificates) and the relationships between them.

[0127] Finally, based on the company's employee data and relationships, a pre-defined employee information knowledge graph is established.

[0128] The entities in the pre-defined employee information knowledge graph are enterprise employee data, and the edges of the pre-defined employee information knowledge graph represent the relationships between entities.

[0129] The edges in the pre-defined employee information knowledge graph are configured with weights.

[0130] Furthermore, based on preset time rules, the weights of the edges in the preset employee information knowledge graph are updated.

[0131] Specifically, initial weights are assigned to the relationships (connection edges) in the employee information knowledge graph, and an automatic decay rule is designed.

[0132] More specifically, the "newer" the relationship (e.g., a recent project collaboration), the higher the weight; the "older" the relationship (e.g., a temporary project team from a long time ago), the lower the weight will be over time.

[0133] It should be noted that the automatic decay rule solves the problem of outdated and irrelevant data interfering with query results in traditional static graphs (timeliness issues), ensuring that query results focus more on currently valid information.

[0134] In one possible implementation, an adaptive decay rule based on the business scenario is designed (e.g., the weight of project collaboration decays exponentially over time, but the weight of important projects decays more slowly).

[0135] Among them, dynamic weights are more in line with the actual business needs of enterprises, and high-value relationships (such as core project cooperation) are given higher priority in queries.

[0136] It should be noted that knowledge graphs, as a structured data management technology, can construct a semantic network representing real-world objects and their interrelationships through terms such as entities, relationships, and attributes. This effectively integrates multi-source heterogeneous data to form employee personal profiles (including name, gender, job level, department, position, etc.). The large model can effectively understand the user's query intent, so as to better search for corresponding content in the graph.

[0137] This embodiment provides a method for querying enterprise employee data, comprising: acquiring the original natural language query text corresponding to the enterprise employee to be queried; performing a first parsing process on the original natural language query text according to a trained intent parsing big data model to obtain a first parsing result; performing a second parsing process on the first parsing result according to a trained semantic constraint big data model to obtain a second parsing result; determining a target query vector based on the first and second parsing results; and querying in a preset employee information knowledge graph according to the target query vector to obtain the query result for the enterprise employee to be queried. Compared with existing technologies that cannot handle polysemous words and have cumbersome query processes, resulting in low query efficiency for enterprise employee data, this application's intelligent enterprise employee search method based on big data model and knowledge graph technology combines the advantages of both to achieve fast, accurate, and flexible retrieval of enterprise employee information, thereby improving the query efficiency of enterprise employee data.

[0138] Figure 4A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 2 ,like Figure 4 As shown, the train information includes train power source information, train power source power information, and train status information; therefore, step S304 specifically includes:

[0139] S401. Based on the confidence score, determine whether to execute the disambiguation process.

[0140] In this embodiment, the confidence score is used to determine whether the confidence score of the entity / relationship parsed by the large model is lower than the threshold or whether there is potential ambiguity.

[0141] Specifically, if the confidence level of the entity / relationship parsed by the large model is lower than the threshold or there is potential ambiguity (such as vague reference to "Wang"), the disambiguation process is activated.

[0142] S402. If it is determined that the disambiguation process will be executed, then the weighted score value corresponding to each first analysis result shall be determined based on the first analysis result and the second analysis result.

[0143] Specifically, based on the first analysis result, a matching is performed in the preset employee information knowledge graph to obtain the corresponding candidate entities.

[0144] For example, based on the fuzzy keywords (such as "Wang") in the first analysis result, all candidate entities corresponding to the candidate entities are retrieved in the preset employee information knowledge graph.

[0145] Furthermore, based on the second analysis result, the candidate entities are subjected to multi-dimensional scoring processing to obtain scores for multiple preset dimensions.

[0146] The preset dimensions include multiple factors such as literal matching degree, entity semantic similarity, context network similarity, and dynamic weight of association relationship.

[0147] For example, at the entity level, the cosine similarity between the query keyword vector and the candidate entity vector is calculated; at the context level, the local relationship network of the candidate entities (such as department / project association) is extracted, and its similarity to the overall semantics of the query is calculated.

[0148] Furthermore, the scores of multiple preset dimensions are weighted to obtain the weighted score value corresponding to the first analysis result.

[0149] Specifically, the weighting process uses an automatic decay rule: the "newer" the relationship (e.g., a recent project collaboration), the higher the weight; the "older" the relationship (e.g., a temporary project team from a long time ago), the weight will automatically decrease over time.

[0150] In one possible implementation, a context-aware module is introduced to optimize the interpretation of the intent of the current query by combining the user's historical query records.

[0151] Among them, the parsing accuracy for continuous queries (such as "I used to look for Java engineers, now I'm looking for Python engineers") has been improved.

[0152] S403. Based on the weighted score, select the target query vector from the first analysis result.

[0153] In this embodiment, based on the weighted score, the candidate with the highest score is selected as the final entity / relationship from the first parsing result, and the intent is calibrated to obtain the target query vector.

[0154] It should be noted that it has stronger parsing robustness for complex natural language queries that contain multiple fuzzy conditions and implicit relationships.

[0155] It should be noted that the joint calibration mechanism reduces reliance on the results of a single model, improving the system's fault tolerance and reliability. The intelligent parsing capabilities of the large model make user input more natural and flexible, eliminating the need to memorize complex query statements and improving the user experience.

[0156] It should be noted that automated querying and reasoning reduce human intervention and lower the error rate of information retrieval caused by human error.

[0157] In this embodiment, a disambiguation process decision mechanism based on confidence scores is used, combined with candidate entity matching from the employee information knowledge graph and multi-dimensional scoring (literal matching degree, entity semantic similarity, context network similarity, and dynamic weight of association relationship) weighting processing, to achieve accurate screening of the target query vector in the first parsing result. This not only improves the accuracy of query vector selection, but also ensures the comprehensiveness and reliability of entity matching through multi-dimensional scoring, ultimately optimizing the efficiency and quality of information parsing.

[0158] Figure 5 A flowchart illustrating a method for querying enterprise employee data provided in this application embodiment. Figure 3 ,like Figure 5 As shown, the specific implementation steps of S305 above include:

[0159] S501. Based on the target query vector, perform a query in the preset employee information knowledge graph to obtain employee information of multiple candidate companies.

[0160] In this embodiment, the target query vector (the intent and key information parsed by the large model) is converted into the query language of the knowledge graph; queries and reasoning are performed on the preset employee information knowledge graph to find employee information that meets the user's query conditions, i.e., candidate enterprise employee information.

[0161] S502. Sort the employee information of multiple candidate companies according to the preset sorting strategy.

[0162] The preset sorting strategy considers not only the basic matching degree, but also a comprehensive approach:

[0163] Optional, employee job level (higher job levels may appear earlier in the list).

[0164] Optional, the level or value of the relevant certificate.

[0165] Optionally, the semantic similarity between the employee information (node) and the user's original query intent is calculated using large models and knowledge graph embeddings.

[0166] In this embodiment, the employee information of multiple candidate companies is sorted based on factors such as the relevance of the query results, job level, and certificate level.

[0167] S503. Based on the sorting results, determine the query results for the employees of the enterprise to be queried.

[0168] Furthermore, based on the query results for the employees of the company to be queried, display information is generated.

[0169] Furthermore, information is displayed on the preset display interface.

[0170] For example, the sorted query results can be displayed to the user in the form of a list or cards.

[0171] Among them, the current weight of the association under the influence of the automatic decay rule will also affect the relevance of the result.

[0172] It's worth noting that this solution addresses the problem of traditional search results simply listing everything and lacking intelligent sorting. It ensures that the most relevant and valuable employees (such as those with higher job titles, those more semantically relevant to the query, and those with more up-to-date information) are presented to users first, improving the usability of the results and the user experience.

[0173] In this embodiment, by accurately locating the target query vector in the employee information knowledge graph, optimizing the sorting of multi-dimensional candidate information based on the sorting strategy, and generating and visually displaying the information in a structured manner, efficient information retrieval and user-friendly interaction throughout the entire process from query to presentation are achieved. This not only ensures the accuracy and comprehensiveness of employee information acquisition, but also improves information processing efficiency and user experience through ordered sorting and intuitive display, ultimately optimizing the level of intelligence in enterprise employee information query.

[0174] Figure 6 This is a schematic diagram of the structure of the enterprise employee data query device provided in the embodiments of this application, as shown below. Figure 6As shown, the device includes: an acquisition module 61, a first parsing module 62, a second parsing module 63, a determination module 64, and a query module 65.

[0175] The acquisition module 61 is used to acquire the original natural language query text corresponding to the employees of the enterprise to be queried;

[0176] The first parsing module 62 is used to perform a first parsing process on the original natural language query text based on the trained intent parsing big model to obtain a first parsing result. The first parsing result includes the core intent, key entity list, inter-entity relationship and confidence score corresponding to the original natural language query text. The trained intent parsing big model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationship and confidence score corresponding to the natural language query text samples.

[0177] The second parsing module 63 is used to perform a second parsing process on the first parsing result based on the trained semantic constraint large model to obtain a second parsing result; wherein, the second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result;

[0178] The determination module 64 is used to determine the target query vector based on the first parsing result and the second parsing result; wherein the target query vector includes the target entity and / or the target relationship;

[0179] The query module 65 is used to perform a query in a preset employee information knowledge graph based on the target query vector in order to obtain the query results of the employees of the enterprise to be queried.

[0180] In one possible design, the target query vector is determined based on the first and second parsing results, including:

[0181] Module 64 is also used to determine whether to execute the disambiguation process based on the confidence score;

[0182] If it is determined that the disambiguation process will be executed, then the weighted score value corresponding to each first analysis result will be determined based on the first analysis result and the second analysis result.

[0183] Based on the weighted score, the target query vector is selected from the first analysis result.

[0184] In one possible design, based on the first and second analysis results, a weighted score value corresponding to each first analysis result is determined, including:

[0185] The determination module 64 is also used to perform matching in the preset employee information knowledge graph based on the first parsing result to obtain the corresponding candidate entities;

[0186] Based on the second analysis result, the candidate entities are subjected to multi-dimensional scoring processing to obtain scores in multiple preset dimensions;

[0187] The scores of multiple preset dimensions are weighted to obtain the weighted score value corresponding to the first analysis result.

[0188] In one possible design, the preset dimensions include multiple of the following: literal matching degree, entity semantic similarity, context network similarity, and dynamic weight of association relationship.

[0189] In one possible design, before obtaining the raw natural language query text corresponding to the employees of the company to be queried, the following steps are also included:

[0190] Acquire employee data for all company employees; this data includes multiple items such as name, gender, job title, department, position, educational background, industry information, and certifications.

[0191] To obtain the relationships between enterprise employee data;

[0192] Based on enterprise employee data and relationships, a pre-defined employee information knowledge graph is established; where the entities in the pre-defined employee information knowledge graph are enterprise employee data, and the edges of the pre-defined employee information knowledge graph represent the relationships between entities.

[0193] In one possible design, the edges in the pre-defined employee information knowledge graph are configured with weights;

[0194] After establishing a pre-defined employee information knowledge graph based on enterprise employee data and relationships, the following is also included:

[0195] Update the weights of edges in the preset employee information knowledge graph according to the preset time rules.

[0196] In one possible design, based on the target query vector, a query is performed in a pre-defined employee information knowledge graph to obtain the query results for the employees of the enterprise to be queried, including:

[0197] The query module 65 is also used to perform a query in a preset employee information knowledge graph based on the target query vector to obtain employee information of multiple candidate companies;

[0198] According to the preset sorting strategy, the employee information of multiple candidate companies is sorted.

[0199] Based on the sorting results, determine the query results for the employees of the company to be queried.

[0200] In one possible design, after determining the query results for the employees of the company to be queried based on the sorting results, the following steps are also included:

[0201] Generate and display information based on the query results for the employees of the company to be queried;

[0202] The information is displayed on the preset display interface.

[0203] This embodiment provides a query device for enterprise employee data, which can execute a query method for enterprise employee data in the above embodiment. Its implementation principle and technical effect are similar, and will not be described again here.

[0204] In the specific implementation of the aforementioned method for querying enterprise employee data, each module can be implemented as a processor. The processor can execute computer execution instructions stored in the memory, thereby enabling the processor to execute the aforementioned method for querying enterprise employee data.

[0205] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 7 As shown, the electronic device 70 includes at least one processor 71 and a memory 72. The electronic device 70 also includes a communication component 73. The processor 71, memory 72, and communication component 73 are connected via a bus 74.

[0206] In the specific implementation process, at least one processor 71 executes computer execution instructions stored in memory 72, causing at least one processor 71 to execute a method in the field of enterprise employee information management as executed on the electronic device side.

[0207] The specific implementation process of processor 71 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0208] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0209] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage.

[0210] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0211] The above description of the functions implemented by electronic devices and main control devices has introduced the solutions provided by the embodiments of the present invention. It is understood that, in order to implement the above functions, the electronic device or main control device includes hardware structures and / or software modules corresponding to the execution of each function. By combining the units and algorithm steps of the various examples described in the embodiments of the present invention, the embodiments of the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the technical solutions of the embodiments of the present invention.

[0212] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the above-described method in the field of enterprise employee information management.

[0213] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0214] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in an electronic device or a host device.

[0215] This application also provides a computer program product, comprising: a computer program stored in a readable storage medium, wherein at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the scheme provided in any of the above embodiments.

[0216] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disk, or optical disk.

[0217] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it is readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. The above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for querying enterprise employee data, characterized in that, include: Obtain the original natural language query text corresponding to the employees of the company to be queried; Based on the trained intent parsing model, the original natural language query text is subjected to a first parsing process to obtain a first parsing result; wherein, the first parsing result includes the core intent, key entity list, inter-entity relationship and confidence score corresponding to the original natural language query text; the trained intent parsing model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationship and confidence score corresponding to the natural language query text samples; Based on the trained semantic constraint model, the first parsing result is subjected to a second parsing process to obtain a second parsing result; wherein, the second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result; Based on the first parsing result and the second parsing result, a target query vector is determined; wherein, the target query vector includes target entities and / or target relationships; Based on the target query vector, a query is performed in the preset employee information knowledge graph to obtain the query results of the enterprise employees to be queried.

2. The method according to claim 1, characterized in that, The step of determining the target query vector based on the first parsing result and the second parsing result includes: Based on the confidence score, determine whether to perform the disambiguation process; If it is determined that the disambiguation process will be executed, then a weighted score value corresponding to each of the first parsing results is determined based on the first parsing result and the second parsing result. Based on the weighted score, the target query vector is selected from the first parsing result.

3. The method according to claim 2, characterized in that, The step of determining the weighted score value corresponding to each of the first analysis results based on the first analysis result and the second analysis result includes: Based on the first parsing result, a match is made in the preset employee information knowledge graph to obtain the corresponding candidate entities; Based on the second analysis result, the candidate entity is subjected to multi-dimensional scoring processing to obtain scores in multiple preset dimensions; The scores of the multiple preset dimensions are weighted to obtain the weighted score value corresponding to the first analysis result.

4. The method according to claim 3, characterized in that, The preset dimensions include multiple factors such as literal matching degree, entity semantic similarity, context network similarity, and dynamic weight of association relationship.

5. The method according to any one of claims 1 to 4, characterized in that, Before obtaining the original natural language query text corresponding to the employees of the enterprise to be queried, the process also includes: Obtain employee data for all company employees; wherein, the employee data includes multiple items such as name, gender, job level, department, position, educational background, industry information, and qualification certificates; Obtain the relationships between the enterprise's employee data; Based on the enterprise employee data and the relationships, a preset employee information knowledge graph is established; wherein, the entities in the preset employee information knowledge graph are the enterprise employee data, and the edges of the preset employee information knowledge graph represent the relationships between the entities.

6. The method according to claim 5, characterized in that, The edges in the preset employee information knowledge graph are configured with weights. After establishing the preset employee information knowledge graph based on the enterprise employee data and the relationships, the method further includes: According to the preset time rules, update the weights of the edges in the preset employee information knowledge graph.

7. The method according to any one of claims 1 to 4, characterized in that, The step of querying the target query vector in a preset employee information knowledge graph to obtain the query results for the employees of the enterprise to be queried includes: Based on the target query vector, a query is performed in the preset employee information knowledge graph to obtain employee information from multiple candidate companies; The employee information of the multiple candidate companies is sorted according to a preset sorting strategy. Based on the sorting results, determine the query results for the employees of the enterprise to be queried.

8. The method according to claim 7, characterized in that, After determining the query results for the employees of the enterprise to be queried based on the sorting results, the method further includes: Based on the query results for the employees of the company to be queried, generate and display information; The information is displayed on the preset display interface.

9. A device for querying enterprise employee data, characterized in that, include: The acquisition module is used to acquire the original natural language query text corresponding to the employees of the enterprise to be queried; The first parsing module is used to perform a first parsing process on the original natural language query text based on the trained intent parsing model to obtain a first parsing result; wherein, the first parsing result includes the core intent, key entity list, inter-entity relationship and confidence score corresponding to the original natural language query text; the trained intent parsing model is trained from natural language query text samples and the core intent, key entity list, inter-entity relationship and confidence score corresponding to the natural language query text samples; The second parsing module is used to perform a second parsing process on the first parsing result based on the trained semantic constraint model to obtain a second parsing result; wherein, the second parsing result is the semantic similarity information of the knowledge graph corresponding to the first parsing result; The determination module is used to determine a target query vector based on the first parsing result and the second parsing result; wherein the target query vector includes target entities and / or target relationships; The query module is used to perform a query in a preset employee information knowledge graph based on the target query vector to obtain the query results of the enterprise employees to be queried.

10. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

Images