An insurance information recommendation method, device, equipment and storage medium

Abstract

The application discloses an insurance information recommendation method and device, equipment and a storage medium. The method comprises the following steps: obtaining the behavior data of a user, and determining the knowledge understanding degree and the knowledge coverage range of the user in the insurance field according to the behavior data. A plurality of candidate insurance information matched with the knowledge understanding degree and the first score of each candidate insurance information are determined. The second score of each candidate insurance information is determined based on the knowledge coverage range. The recommendation score of each candidate insurance information is determined by combining the first score and the second score, and the candidate insurance information with a recommendation score greater than or equal to a preset score threshold is determined as target insurance information and output. It can be seen that the present scheme can help users improve their cognition of insurance knowledge and help users understand new insurance knowledge, thereby increasing the probability that users are satisfied with the recommended insurance information.

Description

Technical Field

[0001] This application relates to the field of computers, and more particularly to a method, apparatus, device, and storage medium for recommending insurance information. Background Technology

[0002] Insurance information recommendation refers to the process of filtering suitable insurance information and recommending it to users based on their interests. Existing insurance information recommendation methods are usually based on tag similarity matching mechanisms. That is, first, user interest tags for insurance information are generated based on user browsing and clicking behavior, and information tags are set for each piece of insurance information. Then, suitable insurance information is filtered and recommended to users based on the similarity between the two types of tags.

[0003] However, existing insurance information recommendation methods are relatively rigid, only considering the relevance of recommended insurance information to user interests. This may lead to user dissatisfaction with the recommended insurance information, thereby affecting the recommendation effect. Summary of the Invention

[0004] This application provides an insurance information recommendation method, apparatus, device, and storage medium, which can recommend insurance information that helps users deepen their understanding of the insurance field and learn new insurance knowledge based on the user's level of knowledge and the scope of their knowledge in the insurance field, thereby improving user satisfaction with the recommendation results and enhancing the recommendation effect of insurance information.

[0005] In a first aspect, embodiments of this application provide an insurance information recommendation method, including: Acquire user behavior data, and determine the user's level of understanding and knowledge coverage in the insurance field based on the behavior data; Based on the level of knowledge understanding, multiple candidate insurance information matching the level of knowledge understanding are determined, and a first score is given for each candidate insurance information. The first score of each candidate insurance information is used to represent the degree to which each candidate insurance information improves the level of knowledge understanding. A second score is determined for each candidate insurance information based on the knowledge coverage, and the second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage. Based on the first score and the second score, a recommended score is determined for each candidate insurance information. The target insurance information is determined from the multiple candidate insurance information and the target insurance information is output. The recommendation score of the target insurance information is greater than or equal to a preset score threshold.

[0006] Optionally, before acquiring the user's behavioral data, the method further includes: Obtain information on insurance types, insurance products, and insurance terms, and determine the relationships between these information. An insurance knowledge graph is constructed using the insurance type information, the insurance product information, and the insurance terms information as entities, and the related relationships as relations.

[0007] Optionally, determining the user's level of knowledge and scope of knowledge regarding the insurance field based on the behavioral data includes: The analysis model is invoked to analyze the behavioral data and determine the user's level of knowledge and understanding regarding the insurance field. Based on the behavioral data and the insurance knowledge graph, a user knowledge graph for the insurance field is determined, and the knowledge coverage is determined based on the user knowledge graph.

[0008] Optionally, before determining multiple candidate insurance information items matching the level of knowledge understanding based on the level of knowledge understanding, the method further includes: Obtain insurance information to be recommended, and call a large language model to decompose the insurance information to be recommended, resulting in multiple insurance information to be screened; The label determination model is invoked to determine the stage label of each piece of insurance information to be screened among the multiple pieces of insurance information to be screened. The stage label of each piece of insurance information to be screened is used to characterize the knowledge understanding stage to which each piece of insurance information to be screened is applicable and the degree of improvement of each piece of insurance information to the applicable knowledge understanding stage.

[0009] Optionally, the step of determining multiple candidate insurance information entries matching the knowledge understanding level, and the first score of each candidate insurance information entry, includes: Based on the knowledge understanding level corresponding to the knowledge understanding stage and the knowledge understanding stage applicable to each piece of insurance information to be screened, multiple candidate insurance information is determined from the multiple pieces of insurance information to be screened; Based on the level of knowledge understanding and the degree to which each candidate insurance information improves the applicable knowledge understanding stage, a first score is determined for each candidate insurance information.

[0010] Optionally, determining the second score for each candidate insurance information based on the knowledge coverage includes: Determine the coordinate information of each candidate insurance information in the insurance knowledge graph; Based on the knowledge coverage and the coordinate information of each candidate insurance information in the insurance knowledge graph, a second score is determined for each candidate insurance information.

[0011] Optionally, after determining the target insurance information from the plurality of candidate insurance information and outputting the target insurance information, the method further includes: In response to a browsing operation targeting the insurance information, the user's knowledge coverage is redefined based on the coordinate information of the target insurance information in the insurance knowledge graph, and the tag determination model is updated based on the target insurance information.

[0012] Secondly, embodiments of this application provide an insurance information recommendation device, comprising: The behavior data acquisition module is used to acquire user behavior data and determine the user's level of knowledge and knowledge coverage in the insurance field based on the behavior data. The first score determination module is used to determine multiple candidate insurance information that match the knowledge understanding level based on the knowledge understanding level, and a first score for each candidate insurance information among the multiple candidate insurance information, wherein the first score of each candidate insurance information is used to represent the degree to which each candidate insurance information improves the knowledge understanding level; The second score determination module is used to determine a second score for each candidate insurance information based on the knowledge coverage scope. The second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage scope. The recommendation score determination module is used to determine the recommendation score for each candidate insurance information based on the first score and the second score; The target insurance information determination module is used to determine the target insurance information from the multiple candidate insurance information and output the target insurance information, wherein the recommended score of the target insurance information is greater than or equal to a preset score threshold.

[0013] Thirdly, embodiments of this application provide an electronic device, the device including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps of the insurance information recommendation method described above.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform any of the implementation steps of the insurance information recommendation method described above.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: In this embodiment, user behavior data is first acquired, and the user's level of knowledge understanding and knowledge coverage in the insurance field are determined based on this data. Then, multiple candidate insurance information items matching the user's knowledge understanding level are identified, and a first score is determined for each candidate insurance information item to evaluate its improvement on the user's knowledge understanding. Further, a second score is determined for each candidate insurance information item based on its knowledge coverage to evaluate the novelty of each item compared to the knowledge within its coverage. Based on this, a recommended score for each candidate insurance information item is determined by combining the first and second scores. Finally, target insurance information with a recommended score greater than or equal to a preset score threshold is identified from the multiple candidate insurance information items, and this target insurance information is output.

[0016] Therefore, this embodiment of the application can determine candidate insurance information that matches the user's level of knowledge in the insurance field, ensuring that the recommended insurance information is appropriate to the user's level of knowledge and helps to further improve the user's understanding of insurance. Simultaneously, it can evaluate candidate insurance information based on the breadth of the user's knowledge in the insurance field, thereby reducing the duplication of recommended content with the user's existing insurance knowledge. Finally, by comprehensively considering the expected improvement of each candidate insurance information in the user's insurance understanding and its novelty compared to the user's existing knowledge, the final insurance information recommended to the user is determined. This method ensures that the recommended insurance information not only helps the user improve their understanding of insurance knowledge but also helps them learn new insurance knowledge, increasing the probability of user satisfaction with the recommended insurance information. Attached Figure Description

[0017] Figure 1 A flowchart illustrating an insurance information recommendation method provided in this application embodiment; Figure 2 A flowchart illustrating an application scenario of an insurance information recommendation method provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an insurance information recommendation device provided in an embodiment of this application. Detailed Implementation

[0018] As mentioned earlier, in practical applications, existing insurance information recommendation methods typically generate user interest tags based on browsing and clicking behaviors, assign corresponding information tags to each piece of insurance information, and then filter and recommend relevant insurance information to users based on the similarity between interest tags and information tags. However, this insurance information recommendation method is relatively rigid, only considering the relevance of recommended insurance information to user interests, which may lead to user dissatisfaction with the recommended insurance information, thus affecting the recommendation effect.

[0019] For example, after a user reads several articles introducing the basic concepts of critical illness insurance, the system can determine that the user is interested in critical illness insurance based on their reading behavior and continuously recommend product information, claims cases, and other content to that user. However, the user may not yet understand basic concepts such as the sum insured and payment period in insurance products, making it difficult to comprehend the recommended product information, and therefore, they may not be satisfied with the recommended insurance information. Furthermore, existing insurance information recommendation methods, after determining that a user is interested in critical illness insurance, typically continuously recommend related insurance information, resulting in monotonous recommendations and potentially reducing user satisfaction. Therefore, recommending insurance information solely based on the relevance between insurance information and user interests may decrease user satisfaction with the recommended insurance information and affect the effectiveness of insurance information recommendations.

[0020] Based on this, embodiments of this application provide an insurance information recommendation method, including: acquiring user behavior data, and determining the user's level of knowledge understanding and knowledge coverage in the insurance field based on the behavior data; determining multiple candidate insurance information items matching the user's level of knowledge understanding, and a first score for each candidate insurance information item, the first score representing the degree to which the corresponding candidate insurance information item improves the level of knowledge understanding; determining a second score for each candidate insurance information item based on the knowledge coverage, the second score representing the novelty of the corresponding candidate insurance information item compared to the knowledge within the knowledge coverage; determining a recommendation score for each candidate insurance information item based on the first score and the second score; determining target insurance information from the multiple candidate insurance information items and outputting the target insurance information item, wherein the recommendation score of the target insurance information item is greater than or equal to a preset score threshold.

[0021] As can be seen, this insurance information recommendation method can determine candidate insurance information that matches the user's level of knowledge, ensuring that the candidate insurance information is appropriate for the user's understanding of insurance knowledge. Based on this, the method evaluates the degree to which each candidate insurance information enhances the user's understanding and its novelty compared to the user's existing knowledge, and comprehensively determines the recommendation score. Finally, it outputs the target insurance information whose recommendation score reaches a preset threshold. This helps ensure that the insurance information recommended to the user not only deepens their understanding of insurance knowledge but also helps them learn new insurance information, reducing the duplication of recommended content with the user's existing knowledge, thereby increasing user satisfaction with the recommendation results and improving the effectiveness of insurance information recommendation.

[0022] It should be noted that the embodiments of this application do not limit the executing entity of the insurance information recommendation method. For example, the insurance information recommendation method of this application embodiment can be applied to data processing devices such as terminal devices or servers. The terminal device can be an electronic device such as a smartphone, computer, personal digital assistant (PDA), or tablet computer. The server can be a standalone server, a cluster server, or a cloud server.

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0024] Figure 1 This is a flowchart illustrating an insurance information recommendation method provided in an embodiment of this application. (In conjunction with...) Figure 1 As shown, the insurance information recommendation method provided in this application embodiment may specifically include the following steps S101-S105.

[0025] S101: Obtain user behavior data and determine the user's level of understanding and knowledge coverage in the insurance field based on the behavior data.

[0026] In this embodiment, before acquiring user behavior data, reference information in the insurance field, such as insurance type information, insurance product information, and insurance clause information, as well as the relationships between these reference information, can be pre-acquired. Subsequently, the reference information, including insurance type information, insurance product information, and insurance clause information, can be treated as entities, and the relationships between these entities as relations, to construct an insurance knowledge graph. For example, insurance type information may include information on critical illness insurance, medical insurance, and life insurance; insurance product information may include relevant information on specific insurance products under each insurance type.

[0027] After constructing the insurance knowledge graph, user behavioral data can be obtained. For example, behavioral data can include insurance questions searched by the user, the time spent on insurance information pages, and insurance information repeatedly viewed by the user. Subsequently, an analytical model can be invoked to analyze the user's behavioral data, obtaining a cognitive grading vector representing the user's level of understanding of insurance knowledge. The dimensions of this cognitive grading vector can include the user's understanding of insurance terminology, the current stage of knowledge understanding, and the range of insurance types already explored. For example, the current stage of knowledge understanding can include four stages: S1, S2, S3, and S4. If a user's cognitive grading vector has values ​​corresponding to the current stage of knowledge understanding of [S1: 0.7, S2: 0.1, S3: 0.1, S4: 0.1], then based on the maximum value of 0.7, the user's current stage of knowledge understanding can be determined to be stage S1. The analytical model can be a recurrent neural network model or a Transformer model.

[0028] Furthermore, based on user behavioral data, entities corresponding to the insurance knowledge acquired by users through reading can be marked in the insurance knowledge graph. A user knowledge graph is then formed based on these marked entities and the relationships between them. Subsequently, the portion of the user knowledge graph covered in the insurance knowledge graph can be defined as the user's knowledge coverage in the insurance field, representing the distribution of the user's acquired knowledge within the entire insurance knowledge system.

[0029] In this way, by analyzing user behavior data, we can understand the user's level of understanding of insurance knowledge and the scope of their knowledge in the insurance field. This provides a basis for subsequently identifying candidate insurance information that matches the user's level of knowledge understanding, as well as evaluating the extent to which the candidate insurance information improves the user's level of knowledge understanding and its novelty compared to existing knowledge.

[0030] S102: Based on the level of knowledge understanding, determine multiple candidate insurance information that match the level of knowledge understanding, and the first score of each candidate insurance information. The first score of each candidate insurance information is used to represent the degree to which each candidate insurance information improves the level of knowledge understanding.

[0031] In this embodiment, before determining multiple candidate insurance information items, the insurance information to be recommended can be obtained in advance, and a large language model can be used to decompose the insurance information to be recommended, resulting in multiple insurance information items to be screened. For example, the insurance information to be recommended can be articles such as insurance knowledge introductions, insurance product comparisons, or insurance guides; the large language model can be a large language model adapted from the insurance and medical fields, used to decompose the article according to the knowledge content contained in it, so that each insurance information item to be screened can independently express one insurance information item.

[0032] Taking an insurance recommendation article titled "How to Buy Insurance for Patients with Thyroid Nodules" as an example, the large language model can break down the content into multiple pieces of insurance information to be screened, such as "What is the TI-RADS classification of thyroid nodules", "The general underwriting conclusions corresponding to different classifications", and "How to combine different types of insurance when there are thyroid nodules".

[0033] After obtaining multiple insurance information entries to be screened, a label determination model can be invoked to determine the stage label for each entry. The stage label characterizes which knowledge understanding stage each entry applies to, and the degree to which each entry improves the knowledge understanding stage it applies to. For example, if the stage label for an entry is [S1→S2: 0.6, S2→S3: 0.3, S3→S4: 0.1], then based on the maximum value of 0.6 in the stage label, the knowledge understanding stage applicable to this entry is determined to be S1. The probability that this entry can improve the knowledge understanding stage of a user with a knowledge understanding stage of S1 to S2 is estimated to be 0.6, i.e., the degree of improvement is 0.6.

[0034] After completing the above preparations, based on the user's current knowledge understanding stage determined in step S101, multiple pieces of insurance information that match the user's current knowledge understanding stage can be identified as multiple candidate insurance information. For example, when the user's knowledge understanding level corresponds to stage S1, the insurance information applicable to stage S1 can be identified as candidate insurance information.

[0035] Subsequently, a first score can be determined for each candidate insurance information based on the user's level of knowledge understanding and the degree to which each candidate insurance information improves the applicable stage of knowledge understanding. Specifically, the expected target state vector corresponding to the stage label of the candidate insurance information can be determined, and the weighted distance between the cognitive grading vector corresponding to the user's level of knowledge understanding and the expected target state vector can be calculated. Based on this, the first score is determined to evaluate how much the candidate insurance information is expected to improve the user's level of knowledge understanding.

[0036] It is worth noting that, in this embodiment, the first score can be defined as the promotion utility score. .

[0037] In this way, candidate insurance information can be matched to the user's level of knowledge and understanding first, and then the first score can be used to evaluate the degree to which each candidate insurance information improves the level of knowledge and understanding, providing a basis for determining the recommendation score in combination with the novelty of the candidate insurance information.

[0038] S103: Determine a second score for each candidate insurance information based on the knowledge coverage scope. The second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage scope.

[0039] In this embodiment, after obtaining multiple candidate insurance information entries, the coordinate information of each candidate insurance information entry in the insurance knowledge graph can be determined. Specifically, each candidate insurance information entry can be associated with a corresponding entity in the insurance knowledge graph through semantic matching, thereby determining the coordinate information of each candidate insurance information entry. This coordinate information is used to represent the position of the knowledge involved in the candidate insurance information entry in the insurance knowledge graph.

[0040] Subsequently, the semantic distance between the entities in the user knowledge graph determined in step S101 and the coordinate information of each candidate insurance information can be used to determine the second score of each candidate insurance information. The semantic distance reflects the difference between the knowledge involved in the candidate insurance information and the user's existing knowledge. A larger semantic distance indicates that the candidate insurance information is more novel than the knowledge within its coverage, resulting in a higher second score; a smaller semantic distance indicates that the candidate insurance information is closer to the user's existing knowledge, resulting in a lower second score.

[0041] For example, if a user's knowledge mainly covers critical illness insurance but not life insurance, the second score for candidate insurance information related to critical illness insurance will be lower; for candidate insurance information related to life insurance, the second score will be higher.

[0042] It is worth noting that, in this embodiment, the second score can be defined as the information gain score. .

[0043] In this way, based on the difference between the coordinate information of the candidate insurance information and the user's knowledge coverage, a second score can be used to evaluate the novelty of the knowledge that each candidate insurance information can bring to the user. This helps to reduce the duplication of recommended content with the user's existing knowledge and expand the user's insurance knowledge coverage.

[0044] S104: Based on the first score and the second score, determine the recommended score for each candidate insurance information.

[0045] In this embodiment, after obtaining the first score and the second score for each candidate insurance information, the first score and the second score can be weighted and summed to obtain the recommended score for each candidate insurance information. The calculation process for the recommended score can be shown in the following formula (1): (1) in, For first place, For second place, and This represents a dynamic coefficient, the value of which depends on user behavior data and has no fixed value. For example, when user behavior data shows a clear interest in a particular product, the coefficient can be increased. The value can be increased if user behavior data shows a significant lack of understanding of a certain type of insurance knowledge. The value of .

[0046] In this way, by dynamically adjusting the weights corresponding to the first and second scores, and combining the values ​​of the first and second scores to calculate the recommendation score for each candidate insurance information, the recommendation score can comprehensively reflect the role of each candidate insurance information in improving the user's knowledge understanding and expanding the user's knowledge coverage, thereby providing a basis for determining the insurance information worth recommending to the user.

[0047] S105: Determine the target insurance information from multiple candidate insurance information and output the target insurance information. The recommended score of the target insurance information is greater than or equal to the preset score threshold.

[0048] In this embodiment, after obtaining the recommendation score for each candidate insurance information, the recommendation score of each candidate insurance information can be compared with a preset score threshold. Candidate insurance information with a recommendation score greater than or equal to the preset score threshold is determined as the target insurance information to be recommended to the user. The preset score threshold can be adjusted upwards or downwards according to actual recommendation needs. After determining the target insurance information, it can be arranged in descending order of recommendation score, generating a recommendation list and outputting it to the client to display the corresponding target insurance information to the user.

[0049] Furthermore, in this embodiment, in response to a browsing operation on the target insurance information, the user's knowledge related to the target insurance information can be supplemented into the user's knowledge graph based on the coordinate information of the target insurance information in the insurance knowledge graph. Based on the supplemented user knowledge graph, the user's knowledge coverage of insurance knowledge is then redefined. In other words, after detecting that a user has browsed the target insurance information, the user's knowledge coverage can be redefined, allowing the evaluation of the novelty of insurance information when recommending it to the user later to take into account the browsing activity.

[0050] Furthermore, in response to browsing actions targeting insurance information, the label determination model can be updated based on that information. Specifically, data on changes in the knowledge understanding stage in response to browsing actions targeting insurance information can be collected, and this data can be used to fine-tune the label determination model.

[0051] In this way, the target insurance information can be determined and output based on the recommendation score, and in response to browsing operations related to the target insurance information, the user's knowledge coverage and the updated tag determination model can be updated, thereby improving the matching degree between the subsequent recommended insurance information and the user.

[0052] Based on the relevant content of steps S101 to S105 above, it is known that in this embodiment of the application, reference information in the insurance field, such as insurance type information, insurance product information, and insurance clause information, as well as the relationships between the reference information, can be obtained in advance to construct an insurance knowledge graph. Subsequently, user behavior data can be obtained, and an analysis model can be called to determine the user's level of knowledge understanding in the insurance field. Combining the behavior data and the insurance knowledge graph, a user knowledge graph recording the knowledge already mastered by the user can be determined, thereby determining the user's knowledge coverage. On this basis, a large language model can be used to decompose the insurance information to be recommended, obtaining multiple pieces of insurance information to be screened. A label determination model can be used to determine the stage label of each piece of insurance information to be screened. Then, based on the user's level of knowledge understanding and the stage labels of each piece of insurance information to be screened, multiple candidate insurance information matching the user can be determined, and a first score can be determined for each candidate insurance information to evaluate its improvement on the user's level of knowledge understanding. At the same time, the coordinate information of each candidate insurance information in the insurance knowledge graph can be determined, and combined with the knowledge coverage represented by the user knowledge graph, a corresponding second score can be determined to evaluate its novelty relative to the user's already mastered knowledge. Furthermore, the weights corresponding to the first and second scores can be dynamically adjusted. A weighted summation is used to determine the recommendation score for each candidate insurance information. Candidate insurance information with a recommendation score greater than or equal to a preset score threshold is identified as target insurance information and output. In response to browsing operations targeting target insurance information, the user knowledge graph can be updated based on the coordinate information of the target insurance information, redefining its knowledge coverage. The label determination model is also updated based on data showing changes in the user's knowledge comprehension stage after browsing. In this way, by leveraging the insurance knowledge graph and the user knowledge graph, the knowledge involved in candidate insurance information can be compared and correlated with the user's existing knowledge. The role of candidate insurance information in deepening user understanding and providing new knowledge is comprehensively considered to determine the target insurance information. Simultaneously, subsequent recommended information is adjusted based on user feedback after browsing, improving user satisfaction with the recommended insurance information and enhancing the recommendation effect.

[0053] Based on the insurance information recommendation method provided in the above embodiments, this application embodiment can also provide an application scenario for the insurance information recommendation method. The application scenario of the insurance information recommendation method will be described below with reference to the embodiments and accompanying drawings.

[0054] Figure 2 A flowchart illustrating an application scenario of an insurance information recommendation method provided in this application embodiment. For example... Figure 2 As shown, process ① represents the construction of the insurance knowledge graph. In this process, insurance reference information such as insurance type information, insurance product information, and insurance clause information, as well as the relationships between these insurance reference information, can be obtained. Using insurance reference information as entities and relationships as relations, an insurance knowledge graph is constructed.

[0055] Process ② represents the breakdown of insurance information and the determination of stage labels. In this process, insurance information to be recommended can be obtained, and a large language model can be used to break it down into multiple pieces of insurance information that can independently express knowledge points. Subsequently, a label determination model can be used to determine the stage label of each piece of insurance information to be selected, so as to characterize the applicable knowledge understanding stage of the insurance information and the degree of improvement it provides to the applicable knowledge understanding stage.

[0056] Process ③ represents the process of determining the user's level of knowledge understanding and the scope of knowledge coverage. In this process, behavioral data such as the user's search history, dwell time, and review history related to insurance information can be acquired. An analytical model is then used to analyze this behavioral data, resulting in a cognitive grading vector that characterizes the user's level of knowledge understanding. This cognitive grading vector reflects the user's understanding of insurance terminology, current stage of knowledge comprehension, and the range of insurance types already explored. Simultaneously, by combining behavioral data and an insurance knowledge graph, entities corresponding to the knowledge the user has confirmed through reading behavior can be marked in the insurance knowledge graph. These entities are then combined with relevant relationships to form a user knowledge graph, which is used to determine the user's knowledge coverage in the insurance field.

[0057] Process ④ represents the matching, scoring, and recommendation output of candidate insurance information. In this process, multiple candidate insurance information items are determined based on the user's level of knowledge understanding and the applicable knowledge understanding stage for each item of insurance information to be screened. Subsequently, a first score is determined for each candidate insurance information item by combining the user's level of knowledge understanding and the stage label of each item, to evaluate the degree to which the candidate insurance information item improves the user's knowledge understanding. Simultaneously, the coordinate information of each candidate insurance information item in the insurance knowledge graph is determined, and the semantic distance between this coordinate information and the already covered entity set in the user's knowledge graph is calculated. Based on this, a second score is determined for each candidate insurance information item, to evaluate the novelty of each candidate insurance information item relative to the knowledge covered. On this basis, the weights corresponding to the first and second scores can be dynamically adjusted, and a recommendation score is determined by weighted summation. Candidate insurance information items with recommendation scores greater than or equal to a preset score threshold are identified as target insurance information, arranged and output in descending order of recommendation scores.

[0058] Step 5 represents the knowledge coverage and label determination model update process. In this process, after a user browses target insurance information, the relevant content related to this browsing is added to the user's knowledge graph based on the coordinate information of the target insurance information in the insurance knowledge graph. The knowledge coverage is then redefined based on the updated user knowledge graph. Simultaneously, data on changes in the user's knowledge understanding stage after browsing the target insurance information can be collected, and the label determination model can be updated accordingly.

[0059] Furthermore, based on the insurance information recommendation method provided in the above embodiments, this application embodiment can also provide an insurance information recommendation device. The insurance information recommendation device will now be described in conjunction with the embodiments and accompanying drawings.

[0060] Figure 3 This is a schematic diagram of the structure of an insurance information recommendation device provided in an embodiment of this application, combined with... Figure 3 As shown, the insurance information recommendation device 300 may include: The behavior data acquisition module 301 is used to acquire user behavior data and determine the user's level of knowledge understanding and knowledge coverage in the insurance field based on the behavior data. The first score determination module 302 is used to determine multiple candidate insurance information that match the knowledge understanding level based on the knowledge understanding level, and a first score for each candidate insurance information among the multiple candidate insurance information, wherein the first score of each candidate insurance information is used to represent the degree of improvement of the knowledge understanding level by each candidate insurance information; The second score determination module 303 is used to determine a second score for each candidate insurance information based on the knowledge coverage scope. The second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage scope. The recommendation score determination module 304 is used to determine the recommendation score for each candidate insurance information based on the first score and the second score; The target insurance information determination module 305 is used to determine target insurance information from the multiple candidate insurance information and output the target insurance information, wherein the recommended score of the target insurance information is greater than or equal to a preset score threshold.

[0061] Optionally, the insurance information recommendation device 300 further includes: The information acquisition module is used to acquire insurance type information, insurance product information, and insurance clause information, and to acquire the relationship between the insurance type information, the insurance product information, and the insurance clause information; The knowledge graph construction module is used to construct an insurance knowledge graph using the insurance type information, the insurance product information, and the insurance terms information as entities, and the relationship as a relation.

[0062] Optionally, the behavior data acquisition module 301 is specifically used for: The analysis model is invoked to analyze the behavioral data and determine the user's level of knowledge and understanding regarding the insurance field. Based on the behavioral data and the insurance knowledge graph, a user knowledge graph for the insurance field is determined, and the knowledge coverage is determined based on the user knowledge graph.

[0063] Optionally, the insurance information recommendation device 300 further includes: The filtering module is used to obtain insurance information to be recommended and to call a large language model to decompose the insurance information to be recommended, so as to obtain multiple insurance information to be filtered. The label determination module is used to call the label determination model to determine the stage label of each piece of insurance information to be screened among the multiple pieces of insurance information to be screened. The stage label of each piece of insurance information to be screened is used to characterize the knowledge understanding stage applicable to each piece of insurance information to be screened and the degree of improvement of each piece of insurance information to the applicable knowledge understanding stage.

[0064] Optionally, the first score determination module 302 is specifically used for: Based on the knowledge understanding level corresponding to the knowledge understanding stage and the knowledge understanding stage applicable to each piece of insurance information to be screened, multiple candidate insurance information is determined from the multiple pieces of insurance information to be screened; Based on the level of knowledge understanding and the degree to which each candidate insurance information improves the applicable knowledge understanding stage, a first score is determined for each candidate insurance information.

[0065] Optionally, the second score determination module 303 is specifically used for: Determine the coordinate information of each candidate insurance information in the insurance knowledge graph; Based on the knowledge coverage and the coordinate information of each candidate insurance information in the insurance knowledge graph, a second score is determined for each candidate insurance information.

[0066] Optionally, the insurance information recommendation device 300 further includes: An update module is used to respond to a browsing operation for the target insurance information, redetermine the user's knowledge coverage based on the coordinate information of the target insurance information in the insurance knowledge graph, and update the tag determination model based on the target insurance information.

[0067] Furthermore, embodiments of this application also provide an electronic device, including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, the one or more programs including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps of the insurance information recommendation method described above.

[0068] Furthermore, embodiments of this application also provide a computer-readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform any of the implementation steps of the insurance information recommendation method described above.

[0069] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on describing the differences from other embodiments. The same or similar parts between the various embodiments can be referred to mutually.

[0070] The system disclosed in the embodiments is described in a relatively simple manner because it corresponds to the method disclosed in the embodiments. For relevant details, please refer to the method section.

[0071] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 limitations, 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.

[0072] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for recommending insurance information, characterized in that, include: Acquire user behavior data, and determine the user's level of understanding and knowledge coverage in the insurance field based on the behavior data; Based on the level of knowledge understanding, multiple candidate insurance information matching the level of knowledge understanding are determined, and a first score is given for each candidate insurance information. The first score of each candidate insurance information is used to represent the degree to which each candidate insurance information improves the level of knowledge understanding. A second score is determined for each candidate insurance information based on the knowledge coverage, and the second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage. Based on the first score and the second score, a recommended score is determined for each candidate insurance information. The target insurance information is determined from the multiple candidate insurance information and the target insurance information is output. The recommendation score of the target insurance information is greater than or equal to a preset score threshold.

2. The insurance information recommendation method according to claim 1, characterized in that, Before acquiring user behavior data, the method further includes: Obtain information on insurance types, insurance products, and insurance terms, and determine the relationships between these information. An insurance knowledge graph is constructed using the insurance type information, the insurance product information, and the insurance terms information as entities, and the related relationships as relations.

3. The insurance information recommendation method according to claim 2, characterized in that, The process of determining the user's level of knowledge and scope of knowledge regarding the insurance field based on the behavioral data includes: The analysis model is invoked to analyze the behavioral data and determine the user's level of knowledge and understanding regarding the insurance field. Based on the behavioral data and the insurance knowledge graph, a user knowledge graph for the insurance field is determined, and the knowledge coverage is determined based on the user knowledge graph.

4. The insurance information recommendation method according to claim 2, characterized in that, Before determining multiple candidate insurance information entries matching the level of knowledge understanding based on the level of knowledge understanding, the method further includes: Obtain insurance information to be recommended, and call a large language model to decompose the insurance information to be recommended, resulting in multiple insurance information to be screened; The label determination model is invoked to determine the stage label of each piece of insurance information to be screened among the multiple pieces of insurance information to be screened. The stage label of each piece of insurance information to be screened is used to characterize the knowledge understanding stage to which each piece of insurance information to be screened is applicable and the degree of improvement of each piece of insurance information to the applicable knowledge understanding stage.

5. The insurance information recommendation method according to claim 4, characterized in that, The determination of multiple candidate insurance information entries matching the knowledge understanding level based on the knowledge understanding level, and the first score of each candidate insurance information entry, includes: Based on the knowledge understanding level corresponding to the knowledge understanding stage and the knowledge understanding stage applicable to each piece of insurance information to be screened, multiple candidate insurance information is determined from the multiple pieces of insurance information to be screened; Based on the level of knowledge understanding and the degree to which each candidate insurance information improves the applicable knowledge understanding stage, a first score is determined for each candidate insurance information.

6. The insurance information recommendation method according to claim 3, characterized in that, The determination of the second score for each candidate insurance information based on the knowledge coverage includes: Determine the coordinate information of each candidate insurance information in the insurance knowledge graph; Based on the knowledge coverage and the coordinate information of each candidate insurance information in the insurance knowledge graph, a second score is determined for each candidate insurance information.

7. The insurance information recommendation method according to claim 2, characterized in that, After determining the target insurance information from the multiple candidate insurance information and outputting the target insurance information, the method further includes: In response to a browsing operation targeting the insurance information, the user's knowledge coverage is redefined based on the coordinate information of the target insurance information in the insurance knowledge graph, and the tag determination model is updated based on the target insurance information.

8. An insurance information recommendation device, characterized in that, include: The behavior data acquisition module is used to acquire user behavior data and determine the user's level of knowledge and knowledge coverage in the insurance field based on the behavior data. The first score determination module is used to determine multiple candidate insurance information that match the knowledge understanding level based on the knowledge understanding level, and a first score for each candidate insurance information among the multiple candidate insurance information, wherein the first score of each candidate insurance information is used to represent the degree to which each candidate insurance information improves the knowledge understanding level; The second score determination module is used to determine a second score for each candidate insurance information based on the knowledge coverage scope. The second score is used to represent the novelty of each candidate insurance information relative to the knowledge within the knowledge coverage scope. The recommendation score determination module is used to determine the recommendation score for each candidate insurance information based on the first score and the second score; The target insurance information determination module is used to determine the target insurance information from the multiple candidate insurance information and output the target insurance information, wherein the recommended score of the target insurance information is greater than or equal to a preset score threshold.

9. An electronic device, characterized in that, The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of the insurance information recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when executed by a terminal device, implements the steps of the insurance information recommendation method according to any one of claims 1 to 7.

Images