Resource retrieval method and device and electronic equipment
By combining semantic parsing of user search needs with knowledge graphs and using large models to match resource description information, the problem of insufficient understanding of user intent in existing technologies is solved, and more accurate government information resource retrieval is achieved.
Patent Information
- Application Number
- CN202511052210.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-29
- Publication Date
- 2025-11-18
AI Technical Summary
In existing technologies, the retrieval methods for government information resources rely on literal matching and manual classification rules, which lack semantic understanding of user intent, resulting in a discrepancy between the retrieval results and the user's actual goals.
By semantically parsing the retrieval request information using a pre-trained large model and combining it with the target user's knowledge graph, resource identifiers and resource descriptions are determined. The pre-trained large model is then used for matching, and the retrieval results are adjusted based on user feedback.
It achieves a precise understanding of user needs, provides resources that match users' search habits and preferences, and improves the accuracy of search results.
Smart Images
Figure CN120973918A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of data query technology, and in particular to a resource retrieval method, apparatus and electronic device. Background Technology
[0002] Within the national integrated government management system, information resources are characterized by diverse sources, numerous types, and a large quantity, while the search needs of different users vary significantly.
[0003] Currently, government information resources are typically retrieved through resource catalog categorization browsing and keyword matching based on fields such as name and attributes. However, this retrieval method relies on literal matching and manual classification rules, lacks semantic understanding of user intent, and is difficult to accurately understand user needs, resulting in a discrepancy between search results and the user's actual goals. Summary of the Invention
[0004] This disclosure provides a resource retrieval method, apparatus, and electronic device to address the technical problem in related technologies where the retrieval of government information resources relies on literal matching and manual classification rules, lacks semantic understanding of user intent, and is difficult to accurately understand user needs, resulting in retrieval results that deviate from the user's actual goals.
[0005] In a first aspect, embodiments of this disclosure provide a resource retrieval method, the method comprising: Obtain the search request information input by the target user; The retrieval request information is semantically parsed using a pre-trained large model to obtain the semantic parsing results; Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier is determined; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records. The first resource corresponding to the first resource identifier is determined from the resource library, and a second resource is determined from the first resource based on the resource description information of the first resource and the retrieval requirement information using the pre-trained large model; wherein the matching degree between the resource description information of the second resource and the retrieval requirement information is higher than a preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the knowledge graph of the target user; Based on the target user's feedback on the second resource, the resource retrieval results corresponding to the retrieval request information are determined.
[0006] Secondly, embodiments of this disclosure provide a resource retrieval device, the device comprising: The acquisition module is used to acquire the search request information input by the target user; The parsing module is used to perform semantic parsing on the retrieval request information using a pre-trained large model to obtain semantic parsing results; The first determining module is used to determine a first resource identifier based on the semantic parsing result and the target user's knowledge graph; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records. The matching module is used to determine the first resource corresponding to the first resource identifier from the resource library, and to determine the second resource from the first resource based on the resource description information of the first resource and the retrieval requirement information using the pre-trained large model; wherein the matching degree between the resource description information of the second resource and the retrieval requirement information is higher than a preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the knowledge graph of the target user; The second determining module is used to determine the resource retrieval results corresponding to the retrieval request information based on the feedback from the target user to the second resource.
[0007] Thirdly, embodiments of this disclosure provide an electronic device, including: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the method described in the first aspect above.
[0008] Fourthly, embodiments of this disclosure provide a computer-readable storage medium for storing computer-executable instructions that, when executed by a processor, implement the steps of the method described in the first aspect above.
[0009] Fifthly, embodiments of this disclosure provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect above.
[0010] The at least one technical solution provided by the embodiments of the present invention can achieve the following technical effects: In this embodiment of the invention, after the target user inputs search query information, a pre-trained large model is used to perform semantic parsing on the search query information to obtain semantic parsing results. Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier is determined. Since the first resource identifier can be determined by combining the knowledge graph constructed from the user's historical resource search records, the first resource corresponding to the determined first resource identifier can be more in line with the target user's search habits. Then, the first resource corresponding to the first resource identifier can be determined from the resource library. Using the pre-trained large model, a second resource is determined from the first resource based on the resource description information of the first resource and the search query information. Since the target user's knowledge graph can reflect the target user's description preferences for resource retrieval to a certain extent, when describing resources in conjunction with the target user's knowledge graph, resource description information that matches the target user's resource retrieval description preferences can be obtained. When matching this resource description information with the search query information input by the target user, the second resource that the target user wants to retrieve can be determined more accurately from the resource library based on the matching results. Finally, the resource retrieval results corresponding to the search query information can be finally determined based on the target user's feedback on the second resource. Because it can combine the search habits and preferences of target users to accurately understand their needs and provide them with accurate search results, it can effectively solve the technical problem in related technologies that rely on literal matching and manual classification rules to retrieve government information resources, lack semantic understanding of user intent, and make it difficult to accurately understand user needs, resulting in search results that deviate from the user's actual goals. Attached Figure Description
[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is one of the flowcharts illustrating a resource retrieval method provided in an embodiment of the present invention; Figure 2 This is a second schematic flowchart of a resource retrieval method provided in one embodiment of the present invention; Figure 3 A schematic diagram of the module composition of a resource retrieval device 300 provided in one embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation
[0012] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0013] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0014] Please see Figure 1 , Figure 1 This is a flowchart illustrating a resource retrieval method provided in one embodiment of the present invention, as shown below. Figure 1 As shown, the method includes the following steps: Step 102: Obtain the search requirements information input by the target user.
[0015] Step 104: Use a pre-trained large model to perform semantic parsing on the retrieval requirement information to obtain the semantic parsing results.
[0016] Step 106: Determine the first resource identifier based on the semantic parsing results and the target user's knowledge graph; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records.
[0017] Step 108: Determine the first resource corresponding to the first resource identifier from the resource library, and determine the second resource from the first resource based on the resource description information and retrieval requirement information of the first resource using a pre-trained large model; wherein, the matching degree between the resource description information and the retrieval requirement information of the second resource is higher than the preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the target user's knowledge graph.
[0018] Step 110: Based on the target user's feedback on the second resource, determine the resource retrieval results corresponding to the retrieval request information.
[0019] In embodiments of the present invention, a target user can input search requirements to retrieve desired resources based on those requirements.
[0020] In one example, target users can input their search requirements through a human-computer interaction interface, using various formats such as text, voice, or graphics. This interface provides a unified access point for all users and can subsequently generate resource search responses in text, audio, or graphical formats based on the retrieved resources.
[0021] After the target user inputs their search requirements, the system can obtain these requirements and use a pre-trained large model to perform semantic parsing on the search requirements to obtain the semantic parsing results.
[0022] In one embodiment, after obtaining the search request information input by the target user, the user identifier of the target user can also be obtained, and the target user's identity and at least one of the target user's access permissions to the resource library to be searched can be authenticated based on the target user's user identifier.
[0023] In one embodiment, the pre-trained large model can be a multimodal large model (MLM), specifically a large model capable of processing multiple modalities of data such as text, audio, and video. After obtaining the retrieval requirement information, the pre-trained large model can be invoked to input the retrieval requirement information. The pre-trained large model will perform semantic parsing on the input retrieval requirement information and output the semantic parsing results.
[0024] After obtaining the semantic parsing results from the large model, the first resource identifier can be determined based on these results and the target user's knowledge graph. The target user's knowledge graph can be constructed based on the target user's historical resource retrieval records.
[0025] In one embodiment, the historical resource retrieval records of the target user can be obtained, and a knowledge graph corresponding to the retrieval behavior can be generated based on the obtained historical resource retrieval records using the pre-trained large model mentioned above.
[0026] In one example, historical resource retrieval records may include at least one of historical resource retrieval directories and historical resource retrieval frequencies. The knowledge graph constructed based on these historical resource retrieval records and the large model can then be G = {id, r, f}, where: id can be the user's registration code; r can be the resource directories searched by the user; and f can be the retrieval frequency for each resource, measured in months. It should be noted that the content of the knowledge graph shown here is merely an example and is not intended to limit the invention.
[0027] After obtaining the target user's knowledge graph, a first resource identifier can be determined based on the obtained knowledge graph and the semantic analysis results of the retrieval request information. The first resource identifier can be a pre-set number, letter, field, etc., for each resource, or it can be a summary, directory, name, etc. This embodiment of the invention does not impose any limitations, as long as each first resource identifier can uniquely identify each first resource. Since the target user's knowledge graph is constructed based on the target user's historical resource retrieval records, it can effectively represent the target user's retrieval habits and preferences. Therefore, combining the target user's knowledge graph and the semantic analysis results of the retrieval request information to determine the first resource identifier can, to a certain extent, ensure that the determined first resource is one that the target user is interested in and wants to retrieve.
[0028] In one example, the constructed user knowledge graph can include the user's browsing order of historical resources. The constructed knowledge graph can be G={id, r, f, s}, where id is the user's registration code; r is the resource directories searched by the user; f is the search frequency of each resource in monthly units; and s is the search order of different resources. When determining the first resource identifier based on the semantic parsing results and the target user's knowledge graph, the display order corresponding to the first resource identifier can also be determined based on the semantic parsing results and the target user's knowledge graph. The display order corresponding to the first resource identifier can be used to determine the display order of the first resource corresponding to the first resource identifier. Since the constructed user knowledge graph includes the user's browsing order of historical resources, the display order corresponding to the first resource identifier determined by combining the knowledge graph can well match the user's resource browsing order preference when searching for resources. That is, the first resource can be displayed according to the user's preferred browsing order, thereby effectively improving the user's search experience.
[0029] After obtaining the first resource identifier, the first resource corresponding to the first resource identifier can be determined from the resource library, and the pre-trained large model mentioned above can be invoked. Then, the pre-trained large model can be used to determine the second resource from the first resource based on the resource description information of the first resource and the retrieval requirement information. The matching degree between the resource description information and the retrieval requirement information of the second resource is higher than a preset matching degree threshold.
[0030] After obtaining the first resource identifier, the first resource can be identified from the resource repository, which is a database used by the target user to search for resources based on their search requirements. Once the first resource is identified, its resource description information can be obtained. This description information is derived by describing the resources in the resource repository based on the target user's knowledge graph.
[0031] In one example, after constructing a user's knowledge graph based on their historical resource retrieval records, the resources in the resource repository can be described using this knowledge graph to obtain resource description information for each resource. Different resources will have different resource descriptions. Because the constructed knowledge graph effectively reflects the user's search preferences and habits, describing resources using the user's knowledge graph yields resource description information that aligns with the user's search preferences, thus effectively avoiding inaccurate or missed resource searches due to incorrect user descriptions.
[0032] After obtaining the resource description information of the first resource, the pre-trained large model can be invoked. The retrieval requirement information and the resource description information of the first resource are input into this large model. The model performs semantic matching on the retrieval requirement information and the resource description information of the first resource, outputting the semantic matching result. Then, based on the semantic matching result, the first resource corresponding to a semantic matching degree higher than a preset matching degree threshold can be identified as the second resource. Since the second resource is determined based on the target user's retrieval description preferences and retrieval requirement information, the second resource can better meet the target user's retrieval needs; that is, it further identifies the second resource that the target user is likely to retrieve from the first resource that the target user may need to retrieve.
[0033] In one example, after determining the second resource, if the display order corresponding to the first resource identifier has already been determined, the display order of the second resource can be determined based on the display order of the first resource identifier. Specifically, the display order of the first resource can be determined first based on the display order corresponding to the first resource identifier, and then the display order of the second resource can be directly determined according to the display order of the first resource. The display order of the first resource is consistent with the display order of the second resource. Since the display order of the first resource can well match the user's resource browsing order preference when searching for resources, the display order of the second resource determined according to the display order of the first resource can also well match the user's resource browsing order preference when searching for resources, thereby improving the user's search experience.
[0034] After obtaining the second resource, the search results corresponding to the search requirements can be determined based on the target user's feedback on the second resource.
[0035] In one embodiment of the present invention, before determining the resource retrieval results corresponding to the retrieval request information based on the target user's feedback on the second resource, users can be clustered. Specifically, users can be clustered based on the knowledge graph after constructing a knowledge graph based on the user's historical resource retrieval records.
[0036] In one example, the knowledge graph constructed based on a user's historical resource retrieval records can be G={id, r,f, s}, where id is the user's registration code; r is the resource directories the user has searched; f is the search frequency for each resource in monthly units; and s is the search order for different resources. Then, clustering can be performed based on the searched resource directories, search frequency, and search order, i.e., based on r, f, and s, to generate user demand and behavior preference clusters Ci, where i can be the index of different clusters.
[0037] After clustering users, the user cluster corresponding to the target user can be determined. After determining the second resource, it can be updated using a collaborative filtering algorithm based on the knowledge graph of the target user's co-cluster users. This allows for more accurate determination of the second resource even when historical user data is limited. Specifically, after determining the target user's co-cluster users, the preferred resources of that user cluster and their display order can be determined. The target user's co-cluster users are those within the user cluster corresponding to the target user. When determining the preferred resources of co-cluster users, resources with browsing frequency and usage times exceeding a preset threshold can be identified as preferred resources. When determining the display order of the preferred resources, the browsing order of resources with usage frequency and usage times exceeding a preset threshold can be used as the preferred display order. Then, using a collaborative filtering algorithm, the preferred resources are added to the second resource based on their display order, resulting in the added second resource.
[0038] In one embodiment of the present invention, when determining the resource retrieval results corresponding to the retrieval request information based on the target user's feedback on the second resource, a retrieval plan can be generated based on the second resource and its display order. The retrieval plan may include a list composed of resource description information of the second resource, the order of which corresponds to the display order of the second resource. When the resource description information of the second resource is triggered, it will redirect to the corresponding second resource. The generated retrieval plan can be displayed to the target user, who can then interact with it, such as clicking to select a retrieval plan that meets their needs. Based on the target user's feedback on the retrieval plan, the resource retrieval results corresponding to the retrieval request information are determined.
[0039] In one embodiment of the present invention, such as Figure 2As shown, this can be accomplished through a system comprising multiple modules, including a retrieval and interaction GAI (Generative Artificial Intelligence) module, a knowledge graph management GAI module, and an information resource content description database. The system's composition, connection structure, functional modules, and operating modes are as follows: 1. System Structure: The system consists of a human-computer interaction interface, a multimodal large model, three GAI modules, one information resource access control module, and a database and knowledge base system component. The three GAI modules are: a retrieval interaction GAI module, an information resource management GAI module, and a knowledge graph management GAI module. The database and knowledge base system component includes an information resource database, an information resource content description database, an information resource retrieval record database, and a retrieval behavior knowledge graph database. A dedicated resource catalog database is included in the information resource database. 2. Connection Method and Operation Mode: The human-computer interaction interface connects to the retrieval interaction GAI module, enabling bidirectional data transmission; the retrieval interaction GAI module connects to the multimodal large model and the retrieval behavior knowledge graph library, allowing it to read data from the retrieval behavior knowledge graph library and transmit data bidirectionally with the multimodal large model; the multimodal large model connects to the retrieval interaction GAI module, the information resource management GAI module, and the knowledge graph management GAI module, enabling bidirectional data transmission; the information resource access control module connects to the retrieval interaction GAI module and the information resource management GAI module, enabling bidirectional data transmission and also connecting with the information resource retrieval... The system connects to the record library, allowing data to be written to it. The Information Resource Management GAI module connects to the Multimodal Large Model, the Information Resource Access Control module, and the Information Resource Content Description Library, enabling bidirectional data transmission. It also connects to the Knowledge Graph Management GAI module, allowing data to be read from these modules. The Knowledge Graph Management GAI module connects to the Multimodal Large Model, enabling bidirectional data transmission. Furthermore, it connects to the Retrieval Interaction GAI module, the Information Resource Retrieval Record Library, and the Retrieval Behavior Knowledge Graph Library, allowing data to be read from the Information Resource Retrieval Record Library, written to the Retrieval Behavior Knowledge Graph Library, and transmitted to the Retrieval Interaction GAI module.
[0040] The functions of each module are as follows: (1) Human-computer interaction interface: With the support of the search interaction GAI module, a unified access interaction interface is provided for all users, and information resource search responses can be generated in text, graph and audio formats according to user needs; (2) Multimodal large model: Provides multimodal information training and basic information generation for the retrieval interaction GAI module, information resource management GAI module, and knowledge graph management GAI module; (3) Search and Interaction GAI Module: Based on the user search information sent by the human-computer interaction interface, extract the knowledge graph from the search behavior knowledge graph library, analyze the user's search needs and behavioral characteristics, and extract the description information of relevant resources from the information resource content description library through the information resource access control module and the information resource management GAI module, so that the user can better understand the resource catalog and content description information that match their needs, and provide the user with information resources that meet security requirements, and generate multimodal information resource search responses for the human-computer interaction interface; (4) Resource access control module: Based on the analysis of user identity, retrieval needs and behavioral characteristics, and in accordance with the regulations for the use of information resources, determine the accessible resources and access control requirements, transmit the user's needs and the above requirements to the information resource management GAI module for execution, and store the user's access and retrieval records in the information resource retrieval record database; (5) Information Resource Management GAI Module: Extract the retrieval behavior knowledge graph library through the knowledge graph management GAI module, analyze its retrieval needs and behavioral characteristics, and extract each information resource from the information resource library according to the above retrieval needs and behavioral characteristics. Generate the information resource description that best matches the user's retrieval needs and behavioral characteristics in advance, store it in the information resource content description library, and retrieve the corresponding information resources according to the user needs and security control requirements sent by the resource access control module and transmit them to the resource access control module. (6) Knowledge Graph Management GAI Module: Based on the record data in the information resource retrieval record library, and on the basis of multimodal large model training, generate a retrieval behavior knowledge graph and store it in the retrieval behavior knowledge graph library, and provide relevant knowledge graphs to the retrieval interaction GAI module and the information resource management GAI module; (7) Information Resource Repository: Stores various information resources and provides secure access; (8) Information resource content description library: Stores the description information of various information resource contents pre-generated by the information resource management GAI module according to the knowledge graph of user search behavior, and provides secure access; (9) Information resource retrieval record database: Stores user access and retrieval record information sent by the information resource access control module and provides secure access; (10) Search behavior knowledge graph library: Stores the user search behavior knowledge graph generated by the knowledge graph management GAI module and provides secure access.
[0041] In one example, these modules will be used to describe the embodiments of the present invention in detail: First, the target user can input their search query information through a human-computer interaction interface. This query information can be text, voice, or graphics. The retrieval interaction GAI module sends the target user's user identifier to the resource access control module to authenticate the target user's identity and access permissions. After successful authentication, the resource access control module sends the authentication result back to the retrieval interaction GAI module. The retrieval interaction GAI module can then call a pre-trained multimodal large model to perform semantic parsing on the search query information, obtaining the semantic parsing result. This result, along with the target user's user identifier, is then sent to the knowledge graph GAI module to obtain the resource identifier of the first resource the target user is likely to access. The knowledge graph GAI module pre-constructs the target user's knowledge graph based on their historical resource access records. The target user's knowledge graph can be G={id, r, f, s}, where: id is the user's registration code; r is the resource directory the user has searched; f is the search frequency for each resource per month; and s is the search order for different resources. After obtaining the target user's user identifier, the Knowledge Graph (GAI) module can obtain the target user's knowledge graph and, based on the target user's knowledge graph and the semantic parsing results of the retrieval request information, determine the resource identifier of the first resource that the target user may want to retrieve.
[0042] After determining the first resource identifier, the Knowledge Graph GAI module can send the first resource identifier to the Retrieval Interaction GAI module. Upon receiving the first resource identifier, the Retrieval Interaction GAI module can send the first resource identifier and the target user's user identifier to the Information Resource Management GAI module. The Information Resource Management GAI module stores resource description information pre-described based on the user's knowledge graph of each resource in the resource library. The Information Resource Management GAI module can determine the resource description information of the target user for the first resource corresponding to the first resource identifier based on the target user's user identifier. Then, it can call the multimodal big model, inputting the resource description information of the first resource corresponding to the target user and the retrieval request information into the multimodal big model, and obtaining the semantic matching result output by the multimodal big model. Then, based on the semantic matching result, the first resource corresponding to a semantic matching degree higher than a preset matching degree threshold can be determined as the second resource, and the determined second resource can be fed back to the Retrieval Interaction GAI module.
[0043] After obtaining the second resource from the Information Resource Management GAI module, the retrieval interaction GAI module, in order to achieve accurate matching even with limited historical user data, can further determine the user cluster corresponding to the target user, the preferred resources of that cluster, and the display order of those preferred resources. Specifically, it can identify users in the same cluster as the target user, determine their preferred resources and their display order, and then, using a collaborative filtering algorithm, add the preferred resources to the second resource according to their display order, resulting in the added second resource.
[0044] Then, the retrieval interaction GAI module can generate a retrieval plan based on the second resource and its display order. The retrieval plan can include a list of resource descriptions of the second resources, arranged in the same order as the display order of the second resources. Once the resource descriptions are triggered, the user will be redirected to the corresponding second resource. After generating the retrieval plan, the retrieval interaction GAI module can display it to the target user through a human-computer interaction interface using text, graphics, audio, etc. The target user can interact with the retrieval plan through the interface, such as selecting a plan that matches their needs. Based on the target user's feedback on the retrieval plan, the resource retrieval results corresponding to the retrieval requirements are determined.
[0045] In this embodiment of the invention, after the target user inputs search query information, a pre-trained large model is used to perform semantic parsing on the search query information to obtain semantic parsing results. Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier is determined. Since the first resource identifier can be determined by combining the knowledge graph constructed from the user's historical resource search records, the first resource corresponding to the determined first resource identifier can be more in line with the target user's search habits. Then, the first resource corresponding to the first resource identifier can be determined from the resource library. Using the pre-trained large model, a second resource is determined from the first resource based on the resource description information of the first resource and the search query information. Since the target user's knowledge graph can reflect the target user's description preferences for resource retrieval to a certain extent, when describing resources in conjunction with the target user's knowledge graph, resource description information that matches the target user's resource retrieval description preferences can be obtained. When matching this resource description information with the search query information input by the target user, the second resource that the target user wants to retrieve can be determined more accurately from the resource library based on the matching results. Finally, the resource retrieval results corresponding to the search query information can be finally determined based on the target user's feedback on the second resource. Because it can combine the search habits and preferences of target users to accurately understand their needs and provide them with accurate search results, it can effectively solve the technical problem in related technologies that rely on literal matching and manual classification rules to retrieve government information resources, lack semantic understanding of user intent, and make it difficult to accurately understand user needs, resulting in search results that deviate from the user's actual goals.
[0046] Figure 3 The resource retrieval device 300 shown can achieve Figure 1 The method described in the embodiment achieves the same technical effect, and can be specifically referred to in the above description. Figure 1 The resource retrieval method of the illustrated embodiment will not be described in detail here.
[0047] Figure 4 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Please refer to it. Figure 4 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.
[0048] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.
[0049] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.
[0050] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a resource retrieval device at the logical level. The processor executes the program stored in memory and specifically performs the following operations: Obtain the search request information input by the target user; The retrieval request information is semantically parsed using a pre-trained large model to obtain the semantic parsing results; Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier is determined; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records. The first resource corresponding to the first resource identifier is determined from the resource library, and a second resource is determined from the first resource based on the resource description information of the first resource and the retrieval requirement information using the pre-trained large model; wherein the matching degree between the resource description information of the second resource and the retrieval requirement information is higher than a preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the knowledge graph of the target user; Based on the target user's feedback on the second resource, the resource retrieval results corresponding to the retrieval request information are determined.
[0051] The above is as stated in this application. Figure 1The resource retrieval method disclosed in the embodiments described above can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in one or more embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in one or more embodiments of this application can be directly implemented by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.
[0052] The electronic device can also perform Figure 1 The resource retrieval method described herein will not be elaborated further in this application.
[0053] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by a portable electronic device including multiple applications, enable the portable electronic device to perform... Figure 1 The methods of the embodiments shown are not described in detail here.
[0054] This application also proposes a computer program product, which is stored in a storage medium and executed by at least one processor to implement... Figure 1 The methods of the embodiments shown are not described in detail here.
[0055] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.
[0056] In summary, the above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this application should be included within the scope of protection of one or more embodiments of this application.
[0057] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.
[0058] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined in the embodiments of this application, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0059] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0060] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
Claims
1. A resource retrieval method, characterized in that, The method includes: Obtain the search request information input by the target user; The retrieval request information is semantically parsed using a pre-trained large model to obtain the semantic parsing results; Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier is determined; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records. The first resource corresponding to the first resource identifier is determined from the resource library, and a second resource is determined from the first resource based on the resource description information of the first resource and the retrieval requirement information using the pre-trained large model; wherein the matching degree between the resource description information of the second resource and the retrieval requirement information is higher than a preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the knowledge graph of the target user; Based on the target user's feedback on the second resource, the resource retrieval results corresponding to the retrieval request information are determined.
2. The method according to claim 1, characterized in that, The target user's knowledge graph includes the target user's resource browsing order for historical resources; Determining the first resource identifier based on the semantic parsing result and the target user's knowledge graph includes: Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier and the display order corresponding to the first resource identifier are determined. After determining the second resource from the first resource, the method further includes: The display order of the second resource is determined based on the display order corresponding to the first resource identifier.
3. The method according to claim 2, characterized in that, Before determining the resource retrieval results corresponding to the retrieval request information based on the target user's feedback on the second resource, the method further includes: Determine the user cluster corresponding to the target user; the user cluster is obtained by clustering users based on the user's knowledge graph; Based on the knowledge graph of the target user's co-cluster users, determine the preferred resources of the target user's co-cluster users and the resource display order of the preferred resources; wherein, the target user's co-cluster users are users in the user cluster corresponding to the target user; Using a collaborative filtering algorithm, the preferred resource is added to the second resource according to the resource display order of the preferred resource, thus obtaining the second resource after addition.
4. The method according to claim 3, characterized in that, The step of determining the resource retrieval results corresponding to the retrieval request information based on the target user's feedback on the second resource includes: A search plan is generated based on the second resource and its display order; wherein the search plan includes a list of resource description information of the second resource; the arrangement order of the resource description information of the second resource in the list is consistent with the display order of the second resource; the resource description information of the second resource will jump to the corresponding second resource after being triggered; Based on the target user's feedback on the search plan, the resource search results corresponding to the search request information are determined.
5. The method according to claim 1, characterized in that, Before performing semantic parsing on the retrieval request information using a pre-trained large model to obtain the semantic parsing result, the method further includes: The identity of the target user and the access rights of the target user to the resource library to be searched are authenticated based on the user identifier of the target user.
6. The method according to any one of claims 1 to 5, characterized in that, The target user's knowledge graph includes at least one of a resource retrieval catalog and a resource retrieval frequency.
7. A resource retrieval device, characterized in that, The device includes: The acquisition module is used to acquire the search request information input by the target user; The parsing module is used to perform semantic parsing on the retrieval request information using a pre-trained large model to obtain semantic parsing results; The first determining module is used to determine a first resource identifier based on the semantic parsing result and the target user's knowledge graph; the target user's knowledge graph is constructed based on the target user's historical resource retrieval records. The matching module is used to determine the first resource corresponding to the first resource identifier from the resource library, and to determine the second resource from the first resource based on the resource description information of the first resource and the retrieval requirement information using the pre-trained large model; wherein the matching degree between the resource description information of the second resource and the retrieval requirement information is higher than a preset matching degree threshold; the resource description information is obtained by describing the resources in the resource library based on the knowledge graph of the target user; The second determining module is used to determine the resource retrieval results corresponding to the retrieval request information based on the feedback from the target user to the second resource.
8. The apparatus according to claim 7, characterized in that, The target user's knowledge graph includes the target user's browsing order of historical resources; the first determining module is used for: Based on the semantic parsing results and the target user's knowledge graph, a first resource identifier and the display order corresponding to the first resource identifier are determined. The device further includes: The third determining module is used to determine the display order of the second resource according to the display order corresponding to the first resource identifier after determining the second resource from the first resource.
9. The apparatus according to claim 8, characterized in that, The device further includes: The fourth determining module is used to determine the user cluster corresponding to the target user before determining the resource retrieval result corresponding to the retrieval request information based on the target user's feedback on the second resource; the user cluster is obtained by clustering users based on the user's knowledge graph; The fifth determining module is used to determine the preferred resources of the target user's peers and the resource display order of the preferred resources based on the knowledge graph of the target user's peers; wherein, the peers of the target user are users in the user cluster corresponding to the target user; An addition module is used to add the preferred resource to the second resource according to the resource display order of the preferred resource using a collaborative filtering algorithm, so as to obtain the second resource after addition.
10. The apparatus according to claim 9, characterized in that, The second determining module is used for: A search plan is generated based on the second resource and its display order; wherein the search plan includes a list of resource description information of the second resource; the arrangement order of the resource description information of the second resource in the list is consistent with the display order of the second resource; the resource description information of the second resource will jump to the corresponding second resource after being triggered; Based on the target user's feedback on the search plan, the resource search results corresponding to the search request information are determined.
11. The apparatus according to claim 7, characterized in that, The device further includes: The authentication module is used to authenticate at least one of the following, based on the target user's user identifier and the target user's access permissions to the resource library to be retrieved, before the semantic parsing of the retrieval request information is performed by the pre-trained large model to obtain the semantic parsing result.
12. The apparatus according to any one of claims 7 to 11, characterized in that, The target user's knowledge graph includes at least one of a resource retrieval catalog and a resource retrieval frequency.
13. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 6.
14. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store computer-executable instructions that, when executed by a processor, implement the steps of the method described in any one of claims 1 to 6.
15. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1 to 6.