A data processing method, apparatus and equipment
By identifying user intent and benefit usage scenarios, and using a large language model to filter and combine benefits, the problem of inaccurate benefit recommendations in existing technologies has been solved, thereby improving the success rate of redemption and user experience.
Patent Information
- Application Number
- CN202610804688.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-04
- Publication Date
- 2026-11-10
AI Technical Summary
The existing benefits recommendation scheme cannot accurately match user needs, resulting in a low success rate of benefits redemption and a lack of benefits combination capabilities, leading to a fragmented user interaction experience.
By identifying user intent and benefit usage scenarios, a pre-set large language model is used to filter candidate benefits, and a benefit combination strategy is constructed based on logical dependencies to generate structured interaction data.
It improved the success rate of rights redemption, enhanced user understanding and interactive experience, overcame the illusion of large models, and achieved accurate combination recommendations.
Smart Images

Figure CN123071018A_ABST
Abstract
Description
Technical Field
[0001] This document relates to the field of computer technology, and in particular to a data processing method, apparatus, and device. Background Technology
[0002] With the booming development of the "membership economy," major internet service platforms are offering users an increasing variety and number of redeemable benefits. How to accurately recommend benefits to users has become a key focus for these platforms.
[0003] For example, relevance matching can be performed based on users' search keywords to recommend benefits to users based on keyword relevance. However, as the types and number of benefits become increasingly large, the compatibility between benefits selected solely through keyword relevance matching and users is low, affecting the success rate of benefit redemption. Therefore, a more reliable benefit recommendation scheme is needed to improve the success rate of benefit redemption. Summary of the Invention
[0004] The purpose of the embodiments in this specification is to provide a more reliable rights recommendation scheme to improve the success rate of rights redemption.
[0005] To achieve the above technical solution, the embodiments in this specification are implemented as follows: This specification provides a data processing method, comprising: responding to a user's rights query request, determining the user's intent and rights usage scenario corresponding to the rights query request; determining candidate rights matching the user based on the user's intent and the user's rights redemption eligibility, and determining multiple candidate rights types corresponding to the rights usage scenario; using a preset large language model, filtering the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types to obtain multiple target rights, and constructing a rights combination strategy based on the logical dependencies between the multiple target rights; and performing rights redemption interaction processing with the user based on the rights combination strategy.
[0006] This specification provides a data processing apparatus comprising: an intent recognition module, which, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the rights query request; a rights determination module, which, based on the user's intent and the user's rights redemption eligibility, determines candidate rights matching the user and determines multiple candidate rights types corresponding to the rights usage scenario; a strategy determination module, which, using a preset large language model, filters the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types to obtain multiple target rights, and constructs a rights combination strategy based on the logical dependencies between the multiple target rights; and a redemption interaction module, which performs rights redemption interaction processing with the user based on the rights combination strategy.
[0007] This specification provides an embodiment of a data processing device, comprising: a processor; and a memory arranged to store computer-executable instructions, wherein, when executed, the executable instructions cause the processor to: respond to a user's rights query request, determine the user's intent and rights usage scenario corresponding to the rights query request; determine candidate rights matching the user based on the user's intent and the user's rights redemption eligibility, and determine multiple candidate rights types corresponding to the rights usage scenario; use a preset large language model to filter the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types to obtain multiple target rights, and construct a rights combination strategy based on the logical dependencies between the multiple target rights; and perform rights redemption interaction processing with the user based on the rights combination strategy.
[0008] This specification also provides a storage medium for storing computer-executable instructions. When executed by a processor, these instructions implement the following process: in response to a user's rights query request, determining the user's intent and rights usage scenario corresponding to the rights query request; based on the user's intent and the user's rights redemption eligibility, determining candidate rights matching the user and determining multiple candidate rights types corresponding to the rights usage scenario; using a preset large language model, filtering the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types to obtain multiple target rights, and constructing a rights combination strategy based on the logical dependencies between the multiple target rights; and performing rights redemption interaction processing with the user based on the rights combination strategy.
[0009] This specification also provides a computer program product, including a computer program that, when executed by a processor, implements the following process: in response to a user's rights query request, determining the user's intent and rights usage scenario corresponding to the rights query request; based on the user's intent and the user's rights redemption eligibility, determining candidate rights matching the user, and determining multiple candidate rights types corresponding to the rights usage scenario; using a preset large language model, filtering the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types to obtain multiple target rights, and constructing a rights combination strategy based on the logical dependencies between the multiple target rights; and performing rights redemption interaction processing with the user based on the rights combination strategy. Attached Figure Description
[0010] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a schematic diagram illustrating the implementation environment of one data processing method described in this specification. Figure 2 This is a flowchart illustrating the processing procedure of one data processing method described in this specification. Figure 3 This is a flowchart illustrating the process of determining a candidate right in this specification. Figure 4 This is a schematic diagram of a rights strategy generation system described in this specification; Figure 5 This is a schematic diagram of a hybrid retrieval combined with constraint filtering process described in this specification; Figure 6 This is a flowchart illustrating a rights exchange interaction process described in this specification; Figure 7 This is a schematic diagram of the system architecture of a rights strategy generation system described in this specification; Figure 8 This is a schematic diagram of a rights exchange interaction process described in this specification; Figure 9 This is a flowchart illustrating the screening process for one of the target rights in this specification; Figure 10 This is a flowchart illustrating the process of determining a user's intent as described in this specification. Figure 11 This is a schematic diagram illustrating the data processing of a rights strategy generation system as described in this specification; Figure 12 This is a schematic diagram of a data processing device described in this specification; Figure 13 This is a schematic diagram of a data processing device described in this specification. Detailed Implementation
[0011] This specification provides a data processing method, apparatus, and device through its embodiments.
[0012] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.
[0013] This specification provides a more reliable benefit recommendation scheme to improve the success rate of benefit redemption. In practical applications, benefit recommendation can be processed through keyword search. For example, when a user searches for benefits, keyword search can be used to match the user's input keywords with information such as benefit titles and descriptions, and an inverted index can be built based on the relevance. Finally, recommended benefits can be filtered by relevance and displayed in a sorted list. However, this scheme lacks semantic understanding, so the retrieved benefits may not match the user's needs. Furthermore, because it can only recommend one benefit at a time, it cannot link multiple different types of benefits to form specific benefit combinations, lacking benefit combination capabilities. In addition, since users need to click on the details page of each benefit to view the redemption rules, it is prone to user churn, and the information density is low. Alternatively, benefit recommendation can be processed through a collaborative filtering recommendation system. For example, benefit recommendations can be based on a user's historical behavior matrix. For instance, a recommendation rule can be constructed based on the user's historical behavior matrix that "a user who redeemed benefit A also redeemed benefit B." Based on this recommendation rule, if a user redeems benefit A, benefit B can also be recommended to that user. However, the above solutions recommend benefits based on historical behavior matrices. Therefore, they cannot effectively recommend newly launched benefits or those used infrequently, leading to a cold start problem. Furthermore, since the collaborative recommendation system directly provides the recommendation results without explaining why the benefit is recommended or how to use it most cost-effectively (i.e., lacking a "guide" attribute), it lacks explanatory power. Alternatively, an interactive system built on a large model could be used for benefit recommendations, such as calling the large model's API to recommend benefits to users. However, due to the illusion problem, the large model might fabricate non-existent benefits or incorrect benefit redemption rules. Moreover, the benefit recommendations generated by the large model cannot be directly redeemed, nor can the user's current eligibility for benefit redemption be verified in real time, failing to achieve closed-loop benefit management.
[0014] In summary, existing benefit recommendation schemes, when faced with vague user input (such as "where to go on the weekend"), can only return a list of benefits containing keywords, failing to combine different types of benefits, resulting in a low success rate for decision-making. Furthermore, users often abandon benefits due to a lack of understanding of complex redemption rules (such as usage thresholds and redemption ratios), leading to high-value benefits remaining dormant and low utilization rates. Additionally, the disconnect between recommended benefits and their intended use cases forces users to navigate between multiple benefit redemption pages to confirm their applicability, resulting in a fragmented user experience. Therefore, this specification provides a more reliable benefit recommendation scheme. This scheme, on the one hand, can quickly and accurately filter out redeemable candidate benefits matching the user's intent based on the user's intent and redemption eligibility, improving the efficiency and accuracy of subsequent target benefit selection. On the other hand, it can leverage the deep semantic understanding capabilities of a pre-set large language model to filter multiple candidate benefit types corresponding to the benefit's usage scenario, and construct benefit combination strategies based on the logical dependencies between the selected target benefits. This overcomes the illusion of a large model while providing more accurate combined benefit recommendations to users, increasing the success rate of benefit redemption. For specific processing details, please refer to the following embodiments.
[0015] The data processing methods described in one or more embodiments of this specification are applicable to the data acquisition implementation environment, such as... Figure 1 As shown, the implementation environment includes at least: Client 100 and server 200. Furthermore, server 200 can be configured with various network models and algorithms, among which: Client 100 can run on terminal devices, which can be mobile phones, personal computers, tablets, e-book readers, wearable devices, devices that interact with information based on AR (Augmented Reality) and VR (Virtual Reality), and laptop computers, etc. Client 100 can be installed on terminal devices, and Client 100 can be an application, a browser, or a subroutine embedded in an application, etc.
[0016] Server 200 can run on a server, which can be one or more servers, a server cluster consisting of several servers, or a cloud server on a cloud computing platform. Server 200 can be installed on the server. Server 200 can be an application or a subroutine embedded in an application. Various network models and algorithms can be integrated into server 200, or server 200 can call one or more of various network models and algorithms to perform corresponding operations.
[0017] In addition, it may include a database 300, which may be set in the server on which the server 200 runs, or outside the server on which the server 200 runs. The database 300 may store various types of redeemable rights and their descriptions, users' rights redemption qualifications, and model parameters of a preset large language model, etc.
[0018] In this implementation environment, users can trigger a rights query request through client 100. Server 200 responds to the user's rights query request, determines the user's intent and rights usage scenario corresponding to the rights query request, and determines candidate rights matching the user based on the user's intent and rights redemption qualifications. It also determines multiple candidate rights types corresponding to the rights usage scenario. Using a preset large language model, based on the user's rights redemption qualifications and multiple candidate rights types, it filters the candidate rights to obtain multiple target rights. Based on the logical dependencies between multiple target rights, it constructs a rights combination strategy. Based on the rights combination strategy, server 200 can interact with the user through client 100 to process rights redemption.
[0019] like Figure 2 As shown in the embodiments of this specification, a data processing method is provided. The execution subject of this method can be a server, which can be a single independent server or a server cluster composed of multiple servers. The server can be a backend server for services such as instant messaging, commodity trading, and financial management, or a backend server for an internet service application. Specifically, the method may include the following steps: In step S202, in response to the user's rights query request, the user's intent and rights usage scenario corresponding to the rights query request are determined.
[0020] In this context, "rights" can refer to a set of rights with completely different attributes, lifecycles, and redemption methods (such as rights redeemed for physical goods requiring logistics transmission, rights redeemed for virtual coupons requiring activation and use, and rights used in preset scenarios requiring identity verification).
[0021] In practice, users can trigger a rights query request through the client. The client can send the rights query request to the server. The server responds to the user's rights query request and can perform intent recognition processing based on the rights query data corresponding to the rights query request to determine the user's intent and rights usage scenario corresponding to the rights query request.
[0022] For example, a user can enter "What are the redeemable benefits at the amusement park?" in the client. The server can identify the user's intent by performing intent recognition on the query data. The user's intent can be "redeem amusement park-related benefits", and the scenario for using the benefits can be an amusement park scenario.
[0023] When performing intent recognition, the server can extract keywords from the user's input rights query data and determine the user intent and rights usage scenario corresponding to the extracted keywords based on the preset correspondence between the extracted keywords and intents, as well as the preset correspondence between the keywords and scenarios. Based on the determined user intent and rights usage scenario corresponding to the keywords, the server can determine the user intent and rights usage scenario corresponding to the rights query request.
[0024] Furthermore, the method for determining user intent and rights usage scenarios described above is an optional and feasible method. In actual application scenarios, there can be a variety of different methods. Different methods can be selected according to different actual application scenarios. This specification does not specifically limit this method in the embodiments.
[0025] In step S204, based on the user's intent and the user's eligibility for redemption, candidate rights matching the user are determined, and multiple candidate right types corresponding to the right usage scenarios are determined.
[0026] Among them, the redemption qualification can be used to characterize the eligibility conditions for users to redeem rights. For example, if the redemption condition is a points condition, then the user's redemption qualification can include user points, which can be service rewards obtained by users in the process of using Internet services. Candidate rights can include redemption rights for physical goods, redemption rights for virtual coupons, and usage rights for preset scenarios. Redemption rights for physical goods can be used to redeem physical goods such as electrical appliances and daily necessities. Redemption rights for virtual coupons can be used to redeem virtual coupons such as phone bills, video application membership cards, and utility payment vouchers. Usage rights for preset scenarios can be used to redeem the right to use scenarios such as preset waiting areas and fast lanes. The types of candidate rights can include various types such as travel, accommodation, and catering. The types of candidate rights can vary according to the different usage needs of the actual application scenario. This specification does not specifically limit this in the embodiments.
[0027] In practice, the server can use a pre-trained screening model to select candidate benefits that match the user from the current redeemable benefits based on the benefit description information, user intent, and the user's benefit redemption qualifications.
[0028] Alternatively, the server can obtain the user's historical rights redemption records and determine the user's rights redemption preferences based on these records. This allows the server to filter out redeemable rights that match the user's redemption qualifications and preferences from the currently available redeemable rights. Based on the relevance between the selected redeemable rights and the user's intent, the server can then perform further filtering to obtain candidate rights that match the user.
[0029] Furthermore, the above-mentioned method for determining candidate rights that match the user is an optional and feasible method. In actual application scenarios, there can be a variety of different methods. Different methods can be selected according to different actual application scenarios. This specification does not specifically limit this method in the embodiments.
[0030] The server can determine multiple candidate benefit types corresponding to the benefit usage scenario based on the preset correspondence between scenario and benefit type. For example, the preset correspondence between scenario and benefit type can be shown in Table 1 below.
[0031] Table 1
[0032] Assuming the current user's benefit usage scenario is an amusement park scenario, then, based on the preset correspondence in Table 1 above, it can be determined that the multiple candidate benefit types corresponding to the benefit usage scenario may include travel, accommodation, and dining.
[0033] Alternatively, the server can determine multiple candidate right types corresponding to the right usage scenario based on the correlation between the right usage scenario and each right type. In addition, there are many other methods for determining multiple candidate right types, which may vary depending on the actual application scenario. This specification does not limit these methods in detail.
[0034] In step S206, a preset large language model is used to filter candidate rights based on the user's rights redemption qualifications and multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between multiple target rights, a rights combination strategy is constructed.
[0035] In practice, the server can use a pre-set large language model to select target rights that match each candidate right type and are redeemable by the user based on the user's right redemption qualifications.
[0036] After obtaining multiple target benefits, the server can construct a benefit combination strategy based on the logical dependencies between the multiple target benefits. For example, assuming that the candidate benefit types include travel and accommodation, the usage time of target benefit 1 corresponding to travel can be earlier than that of target benefit 2 corresponding to accommodation. That is, the benefit combination strategy can be target benefit 1 - target benefit 2.
[0037] Alternatively, assuming that target benefit 3 and target benefit 4 are prerequisites for redeeming target benefit 5, that is, users can only redeem target benefit 5 after redeeming target benefit 3 and target benefit 4, then the benefit combination strategy can be target benefit 3 + target benefit 4 - target benefit 5.
[0038] Furthermore, when there are a large number of target interests and the logical dependencies between them are complex, the server can use a pre-built strategy determination model to construct an interest combination strategy based on multiple target interests and the logical dependencies between them. The strategy determination model can be a model built based on a pre-set deep learning algorithm.
[0039] In step S208, based on the rights and benefits combination strategy, the rights and benefits exchange interaction is processed with the user.
[0040] In implementation, the server can use a pre-set large language model to generate interactive data based on the benefit combination strategy, and feed the interactive data back to the user to conduct benefit redemption interaction with the user. The interactive data can be structured text and operation guidance generated based on the benefit combination strategy and targeting the user's intent, including step descriptions, reasons for recommendation, and specific benefit redemption instructions for each target benefit.
[0041] In this way, the logical connection capability of natural language in large language models, namely the Chain of Thought (CoT), can be used to generate recommendation reasons for multiple target benefits (e.g., "Because you are a member of level x, we recommend that you claim coupon a first, and then place an order, which is more favorable than paying directly"). This avoids the black box problem in benefit recommendation systems that use collaborative filtering, overcomes the lack of explainability in benefit recommendations, and improves the success rate of benefit redemption.
[0042] This specification provides a data processing method that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate rights types corresponding to the rights usage scenario. Using a pre-defined large language model, it filters the candidate rights based on the user's rights redemption eligibility and multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to redeem rights. This approach, on the one hand, allows for quick and accurate filtering of redeemable candidate rights matching the user's intent based on their intent and redemption eligibility, improving the efficiency and accuracy of subsequent target rights filtering. On the other hand, it leverages the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate rights types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model and provides more accurate recommendations for combined rights, increasing the success rate of rights redemption.
[0043] In practical applications, the specific processing method for determining the candidate rights matching the user based on the user's intent and the user's rights redemption qualifications in step S204 above can vary. The following provides one optional processing method, such as... Figure 3 As shown, the specific process may include the following steps S2042 to S2044.
[0044] In step S2042, a second candidate right that matches the user's intent is selected from the first candidate right.
[0045] In practice, the server can filter the currently redeemable rights based on their launch time, historical redemption count, etc., to obtain the first candidate rights. For example, the server can identify redeemable rights with a redemption count greater than a preset threshold, as well as redeemable rights with a redemption count less than the preset threshold and a launch time within a preset time period (such as the last 3 days, the last week, etc.) as the first candidate rights.
[0046] Then, the server can filter the first candidate rights based on the matching degree between the right description information of the first candidate rights and the user's intent, and obtain the second candidate rights.
[0047] In practical applications, there are various ways to process the second candidate right that matches the user's intent from the first candidate right in step S2042 above. The following is an optional processing method, which may include the following steps A1 to A3.
[0048] In step A1, the first candidate right corresponding to the right usage scenario and the right description information corresponding to the first candidate right are obtained.
[0049] The description of the rights and benefits may include the launch time, the number of historical redemptions, the conditions for redeeming the rights and benefits, the scope of application of the rights and benefits, and the validity period of the redemption.
[0050] In implementation, the server can pre-build a preset correspondence between scenarios and benefit types, and obtain multiple candidate benefit types corresponding to the benefit usage scenarios. Then, the server can filter the redeemable benefits corresponding to each candidate benefit type based on the launch time, redemption conditions, and historical redemption frequency of the redeemable benefits corresponding to each candidate benefit type, and obtain the first candidate benefit.
[0051] In step A2, feature extraction processing is performed on the right description information corresponding to the first candidate right to obtain a first feature vector, and feature extraction processing is performed on the user intent to obtain a second feature vector.
[0052] In implementation, the server can build a rights strategy generation system based on data processing methods, such as... Figure 4 As shown, users can input a rights query request through the user interaction interface on the user terminal. The rights strategy generation system can interact with the user on rights redemption based on the data processing of the intent recognition module, the hybrid retrieval combined with constraint filtering module, and the strategy generation module.
[0053] The server can first construct and vectorize heterogeneous rights knowledge, storing the vectors corresponding to different redeemable rights in the rights database in a format such as... Figure 4 In the knowledge base shown, this allows for the breaking down of technical barriers between different interest types through heterogeneous fusion capabilities (i.e., through a unified vector space and metadata table structure), enabling cross-business line combined recommendations.
[0054] For example, the server can perform feature extraction processing based on the feature description information for redeemable rights of various different types. For instance, the server can extract features from... Figure 4 The metadata of each redeemable right is obtained from the rights library shown, and this metadata is used as the rights description information. Then, based on the metadata, features such as the name, redemption conditions, scope of application, and redemption validity period of each redeemable right are extracted. Then, using a vector transformation model (such as the Embedding model), the extracted features are transformed into high-dimensional vectors and stored in the knowledge base.
[0055] The metadata of redeemable rights is shown in Table 2 below. Table 2 can be used to store the basic attributes of rights. The server can use pre-designed general fields to be compatible with redeemable rights of different rights types.
[0056] Table 2
[0057] In this way, the server can directly obtain the vector corresponding to the first candidate right from the knowledge base and determine the vector as the first feature vector. At the same time, the server can also use vector transformation models (such as the Embedding model) to perform feature extraction processing on the user's intent to obtain the second feature vector.
[0058] In step A3, the first candidate rights are screened based on the semantic similarity between the first feature vector and the second feature vector, and the second candidate rights are determined based on the screened first candidate feature vector.
[0059] In implementation, the server can determine the cosine similarity between the first feature vector and the second feature vector as the semantic similarity between the first feature vector and the second feature vector, so as to filter the first candidate rights based on the semantic similarity, and can determine the filtered first candidate feature vector as the second candidate rights.
[0060] In step S2044, the second candidate rights are screened based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, and the candidate rights that match the user are determined based on the screened second candidate rights.
[0061] In practice, the server can use a hybrid retrieval combined with a constraint filtering module to select candidate benefits that match the user from the benefits pool using a "funnel-shaped" filtering mechanism.
[0062] The hybrid retrieval combined with the constraint filtering module can include a semantic recall layer and a hard constraint filtering layer. The semantic recall layer can be used to recall second candidate benefits based on semantic relevance. For example, the server can filter out the first candidate benefit from the full benefit pool, and then retrieve the second candidate benefit by using the semantic similarity between the first candidate benefit and the user's intent. That is, the server can construct a set of semantically related benefits consisting of the second candidate benefit through vector similarity retrieval.
[0063] Specifically, the server can combine the core intent (Implicit_Needs) and entity keywords from the user's intent. The Entity_Keywords are concatenated, and then the concatenated data is transformed into a query vector V_q, i.e., the second feature vector, through a vector transformation model. Simultaneously, the server can calculate the cosine similarity between V_q and the first feature vector V_b corresponding to the first candidate benefit in the vector database. Based on the calculated cosine similarity, a predetermined number (e.g., top 100, TOP-K, K=100) of the first candidate benefits are selected as second candidate benefits. A semantically relevant benefit set S_{raw} is then constructed based on these second candidate benefits. In this way, the semantic recall layer can solve the problem of inaccurate keyword matching in benefit recommendation. For example, if the user inputs the benefit query data "raising children," benefits that do not contain the keyword "children" but contain the attribute "parent-child" can be recalled.
[0064] Then, the server can use a hard constraint filtering layer to eliminate second candidate rights that do not meet the redemption qualifications or are not within the redemption validity period, based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, and obtain candidate rights that match the user.
[0065] In practical applications, the specific processing method for screening the second candidate right based on the user's corresponding right redemption qualification and the right redemption conditions corresponding to the second candidate right in step S2044 above can be varied. The following provides an optional processing method, which may include the following steps B1 to B2.
[0066] In step B1, based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, the user level verification and user points verification are performed on the second candidate rights respectively to obtain the level verification result and the points verification result.
[0067] In implementation, the server can use "hard constraint filtering" to ensure that the target equity in the equity portfolio strategy generated by the large language model is executable in the physical world, thus overcoming the illusion of "non-exchangeable equity" that generative AI is prone to produce.
[0068] Specifically, such as Figure 5 As shown, the server can traverse each second candidate interest B_i in the semantically relevant interest set S_{raw}, and obtain the level verification result and the point verification result through the following integral verification and level verification: 1. Points Verification: Determine whether the user's points in the corresponding rights redemption qualification are not less than the rights points in the rights redemption conditions corresponding to the second candidate rights. If the user's points are not less than the rights points, that is, user points User_Assets.points>= rights points Bi_i.cost_points, then the points verification result is passed; otherwise, the points verification result can be failed.
[0069] 2. Level Verification: Determine whether the user's level in the corresponding rights redemption qualification is not lower than the rights level in the rights redemption conditions corresponding to the second candidate rights. If the user's level is not lower than the rights level, that is, user level User_Assets.level>= rights level B_i.min_level, then the level verification result is passed; otherwise, the level verification result can be failed.
[0070] In step B2, the second candidate rights are screened based on the grade verification results and the points verification results.
[0071] In implementation, such as Figure 5 As shown, the server can identify the second candidate interest whose level verification result and points verification result both pass as the second candidate interest after screening.
[0072] In addition, the server can also perform inventory and expiration date verification, audience rule verification, and redemption location verification for the second candidate benefits. Among these, inventory and expiration date verification can be performed by querying the real-time inventory system to ensure that the benefits are currently redeemable.
[0073] Furthermore, when the number of available rights in the set S_{valid} constructed based on the second candidate rights after screening is large, to improve the efficiency of determining the candidate rights combination strategy, the server can use algorithms such as Wide&Deep, DIN, and MaskNet to further sort and filter the available rights set S_{valid} to obtain the final candidate set (i.e., the set of available rights candidates). The server can then determine the rights in the final candidate set as candidate rights that match the user. In this way, by combining hybrid retrieval with a constraint filtering module, candidate rights that are "available and desired by the user" can be filtered from a large number of heterogeneous rights.
[0074] In practical applications, the specific processing methods for interacting with users on the basis of the rights and interests combination strategy in step S208 above can vary. The following provides one optional processing method, such as... Figure 6 As shown, the specific process may include the following steps S2082 to S2084.
[0075] In step S2082, target interaction data is generated using a preset large language model based on user intent and the logical dependencies between multiple target benefits in the benefit combination strategy.
[0076] In implementation, the server can utilize the generation capabilities of a pre-set large language model, combined with multiple target interests in the equity portfolio strategy and the logical dependencies between these target interests, to generate natural language strategies, i.e. target interaction data. The large language model can be trained or prompted to follow the CoT (Concept-Solution-Action Suggestion) thought chain structure.
[0077] For example, the generation logic of a large language model can be: 1. Identify the scenario: such as a family trip.
[0078] 2. Route planning: e.g., travel -> accommodation -> dining.
[0079] 3. Filling rights: (1). Travel: We recommend using travel-related target benefits 1.
[0080] (2). Accommodation: Recommended to redeem accommodation-related target benefit 2 (this target benefit requires 1000 user points).
[0081] (3). Catering: We recommend claiming Catering Target Benefit 3 (this target benefit can be a virtual coupon benefit).
[0082] 4. Result Generation: Generate target interactive data, which may include preset slot markers for subsequent front-end rendering.
[0083] Specifically, the server can input the following into the large language model: 1. User Intent; 2. User's eligibility for redeeming benefits (User_Assets): For example, if a user has 5200 points and is level 3; 3. Equity portfolio strategy Retrieved_Benefits: A JSON data list of equity interests constructed from multiple target equity interests; 4. System Instructions: You are a seasoned member lifestyle planner. Based on user intent and redemption eligibility, and using the provided benefits combination strategy, please write a "Benefits Usage Guide".
[0084] 5. Writing Requirements: (1) Scenario-based narrative: Construct specific usage scenarios for the rights and benefits (e.g., "For your weekend family trip, we suggest this combination...").
[0085] (2) Logical arrangement: rights and interests are arranged in the time sequence of "before use -> during use -> after use".
[0086] (3) Asset verification: The generated interactive text clearly informs the user how many points this scheme consumes in total.
[0087] 6. Example of Generated Text: "For this weekend's family time, I've planned a fantastic travel itinerary for you! First, before you set off, you can use Target Benefit 1 (corresponding to preset slot 1) exclusive to Membership Level 3 to skip the queues with your child. Once you arrive at your destination, I recommend redeeming Target Benefit 2 (corresponding to preset slot 2), which only costs 1000 points and saves you 50 yuan. When you're tired from playing, you'll still have enough points to redeem Target Benefit 3 (corresponding to preset slot 3), which can be redeemed directly at a store near the park. This guide will cost you a total of 3000 points, so you have plenty of points to spare. Have a great trip!" In this way, based on the above input data, the large language model can generate target interactive data for users that can realize the interaction of redeeming rights.
[0088] In practical applications, the specific processing methods for generating target interaction data in step S2082 above, which utilizes a preset large language model, based on user intent and the logical dependencies between multiple target rights in the rights combination strategy, can be varied. The following provides an optional processing method, which may specifically include the following steps C1 to C2.
[0089] In step C1, the interactive components corresponding to each target benefit are obtained.
[0090] Interactive components can be used to trigger the execution of rights and benefits redemption operations.
[0091] In step C2, using a pre-defined large language model, target interaction data is generated based on user intent, interactive components corresponding to target benefits, and logical dependencies between multiple target benefits in the benefit combination strategy.
[0092] In practice, the writing requirements for large language models may also include: (4) Use dynamic slots: When you mention a specific benefit, you need to use the format [CARD:benefit_id] to mark it, and the system will automatically replace it with the click card.
[0093] In this way, the server can fill the preset slots with clickable cards, which can be configured with interactive components. Users can then trigger the redemption of benefits in a timely manner through these interactive components, avoiding the problem of low success rate of benefit redemption caused by jumping back and forth between multiple pages.
[0094] In addition, interactive components can also be used to display benefit descriptions so that users can fully understand the benefit information.
[0095] In step S2084, based on the target interaction data, a rights exchange interaction is performed with the user.
[0096] In implementation, such as Figure 7 As shown, the benefits guide generation system can instantly render the target interactive data generated by the benefits generation module within the generation and rendering module, thus integrating the text guide with clickable benefits cards for display. For example, in response to a user's benefits query request, the system can filter out three target benefits—travel benefits 1, dining benefits 1, and dining benefits 2—based on the user's usage scenario and intent. Furthermore, based on these three target benefits and their corresponding interactive components, the system can generate a guide such as... Figure 8 The target interaction data shown allows users to redeem benefits through these three redemption cards.
[0097] like Figure 7As shown, the user terminal can be used to input benefit query requests, display target interactive data, and perform benefit redemption operations. The benefit strategy generation system can be logically divided into four layers: 1) Intent understanding layer: i.e. Figure 4 The intent recognition module can be used to understand users' explicit needs and implicit scenarios based on a preset large language model.
[0098] 2) Strategy orchestration layer: i.e. Figure 4 The hybrid search engine, combined with a constraint filtering engine, can obtain candidate benefits that match the user through vector search and constraint filtering.
[0099] 3) Strategy Generation Layer: Strategy generation engine Figure 4 The strategy generation model is responsible for generating "strategy" text and arranging benefit combinations. It can generate benefit combination strategies based on a preset large language model and generate target interaction data based on the benefit combination strategies.
[0100] 4) Generate rendering layer: This is the card rendering module, which can be used to merge and display text guides and clickable benefit cards.
[0101] The data storage layer can be used to store various types of data, including user profile data, rights and interests data, vector knowledge data, etc.
[0102] The generation rendering layer can replace the preset slots in the generated target interaction number with visual, interactive benefit cards (Action Cards). The final presentation interface can be a smooth suggestion text, interspersed with specific benefit claim or benefit redemption buttons, which users can click to perform the benefit redemption operation.
[0103] In practical applications, step S206 above utilizes a pre-set large language model to filter candidate rights based on the user's rights redemption qualifications and multiple candidate rights types, obtaining multiple target rights. Based on the logical dependencies between these target rights, a rights combination strategy is constructed. The specific processing methods can vary; the following provides one optional processing method, such as... Figure 9 As shown, the specific process may include the following steps S2062 to S2064.
[0104] In step S2062, a preset enhanced retrieval algorithm is used to obtain supplementary knowledge of rights corresponding to candidate rights based on a preset knowledge base, and preset prompt information is constructed based on the supplementary knowledge of rights.
[0105] In practice, to make the generated guides (i.e. target interaction data) user-friendly and professional, the server can also use the Retrieval-augmented Generation (RAG) algorithm to add the retrieved rights and interests supplementary knowledge to the "guide generation prompt information". In other words, preset prompt information can be built based on the rights and interests supplementary knowledge.
[0106] For example, taking the rights and benefits of "tissue paper" as a physical commodity, using a pre-set enhanced retrieval algorithm based on a pre-set knowledge base, the supplementary knowledge of the rights and benefits corresponding to the commodity can be: A Complete Guide to Buying Facial Tissue: A high-quality facial tissue is crucial to both quality of life and health. Selection requires consideration from multiple perspectives. I. Material: The foundation of quality and feel. 1. Virgin wood pulp: The most common raw material used in high-quality facial tissues, resulting in a delicate, soft, smooth, and resilient paper. 2. Virgin bamboo pulp: Possesses natural antibacterial properties and strong absorbency, with outstanding environmental characteristics. 3. Comparison of pure wood pulp and virgin pulp: "Pure wood pulp" may contain recycled materials; while "virgin pulp" refers to pulp made from natural plant fibers without the use of recycled pulp. Purchase recommendation: Prioritize products clearly labeled "100% virgin wood pulp" or "virgin bamboo pulp." II. Brand: Market reputation and quality assurance. Purchase recommendation: Major brands offer high-quality basic series at competitive prices; choose according to your budget and preferences, and pay attention to promotional activities. III. Technical Specifications and Selection Tips: 1. Standards: Look for the highest standards for facial tissues, as hygiene requirements are more stringent. 2. Number of Ply and Thickness: Generally, more layers mean better absorbency and durability. Three-ply facial tissues offer a good balance in thickness, absorbency, and cost-effectiveness. Embossing can also increase thickness and absorbency. 3. Wet Toughness: This reflects the paper's durability when wet. Facial tissues that don't easily tear when wet are less likely to shed lint. 4. Avoid Additives: Avoid tissues with a pungent fragrance or excessive whiteness. Avoid fluorescent whitening agents and excessive fragrances. 5. Practical Methods to Distinguish Quality: Feel the texture (high-quality facial tissues are soft and delicate, and don't shed lint easily); observe burning (normal tissue ash is grayish-white). IV. Applicable Scenarios: Choose according to your needs. Daily Use / Living Room / Office: Choose medium-thickness, soft, skin-friendly three-ply virgin wood pulp facial tissues. Kitchen Use: It is recommended to choose "wet-strength paper" or four-ply or higher kitchen-specific paper, which is strong and doesn't easily tear when wet. For Travel: Choose small packs of soft facial tissues or handkerchiefs. V. Environmental Considerations: Choosing bamboo pulp paper or products with biodegradable packaging is more environmentally friendly. Summary and recommendations: When purchasing tissues, follow the principles of "safety first, choose according to needs, and pay attention to cost-effectiveness." Ensure the safety of the materials (virgin wood pulp / bamboo pulp) and compliance with standards. Determine the thickness and function based on the usage scenario. Choose from reputable brands and take advantage of promotions.
[0107] Based on the aforementioned supplementary knowledge of benefits, the preset prompt message could be: You are a seasoned member lifestyle planner. Please write a "Benefits Usage Guide" based on the user's intent and redemption eligibility, and the provided benefit combination strategy. When generating the guide, you should refer to the retrieved professional knowledge: {retrieved_knowledge}.
[0108] Here, retrieved_knowledge is a placeholder, which can be replaced with the retrieved supplementary knowledge of rights and interests in practical applications.
[0109] In step S2064, a preset large language model is used to filter candidate rights based on preset prompt information, the user's rights redemption qualifications, and multiple candidate rights types, to obtain multiple target rights.
[0110] In implementation, the use of an enhanced retrieval architecture and the selection of genuinely redeemable candidate benefits enables the large language model to more accurately select genuine benefits that meet user needs, thus avoiding the model illusion problem.
[0111] Thus, as Figure 7 As shown, this rights and benefits strategy generation system can first vectorize heterogeneous rights and benefits to build a rights and benefits library. When a user inputs rights and benefits query data, the system can combine user profiles (membership level, points balance, and other rights and benefits redemption qualifications) with rights and benefits query data, and use a large language model for intent recognition and reasoning. Subsequently, the system can retrieve relevant rights and benefits to supplement knowledge through enhanced retrieval, and generate structured "strategy text" by LLM. Finally, specific rights and benefits cards are dynamically embedded into the strategy to generate target interaction data and display it to the user. Through this method of fusion rendering of structured strategies and dynamic interactive components, key entities in the unstructured text generated by LLM can be mapped in real time to structured objects (UI cards) of the business system, realizing a seamless jump from "reading the strategy" to "one-click execution" and shortening the user's decision-making path.
[0112] In practical applications, the specific processing method for determining the user intent corresponding to the rights query request in step S202 above can vary. The following provides one optional processing method, such as... Figure 10 As shown, the specific process may include the following steps S2022 to S2024.
[0113] In step S2022, the rights query data corresponding to the rights query request is obtained.
[0114] In step S2024, a preset large language model is used to perform intent recognition processing on the user based on the rights query data and the user's rights redemption qualifications, so as to obtain the user intent corresponding to the rights query request.
[0115] In implementation, such as Figure 11 As shown, the server can receive user-triggered benefit query requests and input the benefit query data corresponding to the request (e.g., "I want to take my family to Area 1, what discounts are available?") and the user's current state constraint context: C_{user}=\{Level, Points, Other\} (membership level, points balance, other benefit redemption qualifications) into the intent recognition module. Furthermore, during intent recognition processing, if the user's input benefit query data is detected to be ambiguous, a thought chain can be used for reasoning and completion to obtain the user's intent.
[0116] In this way, the server can utilize a pre-defined large language model, based on the rights query data and the user's rights redemption qualifications, and output a standardized intent object, that is, the user intent corresponding to the rights query request, by prompting the engineering requirements.
[0117] For example, a structured prompt could be: "Analyze user intent to determine if cross-category combinations are involved. The current user has 5000 points and is at level 3. They are traveling from the current city to region 1." As shown in Table 3 below, in addition to user level and user points, the user's redemption qualifications may also include information such as the user's permanent residence area and user preference information.
[0118] Table 3
[0119] In this way, structured prompts can constrain the output of the large language model, prevent illusions, and ensure integration with business systems. In addition, through the deep semantic understanding of the large language model, even if the user inputs ambiguous scenario words (such as the process of "through search-strategy-decision"), it can be accurately mapped to specific multi-rights types, overcoming the problem of "insufficient semantic understanding" in rights recommendation methods such as keyword matching.
[0120] The prompt template used to construct structured prompts can be: 1. Input can include: (1) Rights Inquiry Data: "{Natural language input by the user, such as: I want to take my child to the amusement park this weekend, what can I use?}" (2) Current Time: "{System Current Time}" 2. System instructions may include: You are a professional membership benefits intent identification assistant. Please analyze users' benefits query data and their benefits redemption qualifications, extract the following key information, and output it in JSON format, without including any explanatory text: (1) Core Intent: The user's core goal (e.g., travel, dining, shopping).
[0121] (2) Entity_Keywords: The specific items or locations to be searched (e.g., amusement park, family).
[0122] (3) Implicit Needs: Needs inferred from common sense (e.g., taking care of children -> need “parent-child services” or “multi-person packages”).
[0123] (4) Sort_Preference: Whether the user has implied price sensitivity (e.g., cost-effectiveness, most expensive, free).
[0124] An output example could be: JSON { "Core_Intent":"Entertainment", "Entity_Keywords":["amusement park", "weekend"], "Implicit_Needs": ["Family benefits", "Multi-person tickets", "Parking discounts"], "Sort_Preference":"None" } In this way, the server can perform vector semantic retrieval and hard constraint filtering based on user intent, which includes core intent, entity keywords, implicit needs, and ranking preferences, to obtain candidate benefits that match the user intent.
[0125] In practical applications, a user's eligibility for redeeming benefits meets the redemption conditions corresponding to the benefit combination strategy. The redemption conditions corresponding to the benefit combination strategy are determined based on the redemption conditions corresponding to multiple target benefits.
[0126] In implementation, the server can filter out redeemable benefit combinations based on the user's redemption eligibility during the target benefit selection process. This means the user's redemption eligibility must meet the sum of the redemption conditions for multiple selected target benefits. For example, suppose a benefit combination strategy includes travel-related benefit 1, dining-related benefit 2, and accommodation-related benefit 3, and the redemption condition for benefit 1 is 200 points, benefit 2 is 300 points, and benefit 3 is 600 points. Therefore, the total redemption condition for this benefit combination strategy is 200 + 300 + 600 = 1100 points. The user's redemption eligibility must then be no less than 1100 points. This allows the user to redeem all the target benefits displayed in the target interaction data, ensuring a smooth redemption experience.
[0127] In summary, by using a large language model to perform deep intent analysis on users' natural language query data, it is possible to automatically generate "strategy-type" solutions containing multiple heterogeneous benefit combinations, realizing the transformation from "people looking for benefits" to "strategies looking for people". This helps to improve user experience and is easily identifiable by users, thus having high business value.
[0128] This specification provides a data processing method that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate rights types corresponding to the rights usage scenario. Using a pre-defined large language model, it filters the candidate rights based on the user's rights redemption eligibility and multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to redeem rights. This approach, on the one hand, allows for quick and accurate filtering of redeemable candidate rights matching the user's intent based on their intent and redemption eligibility, improving the efficiency and accuracy of subsequent target rights filtering. On the other hand, it leverages the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate rights types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model and provides more accurate recommendations for combined rights, increasing the success rate of rights redemption.
[0129] The above describes the data processing method provided in the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide a data processing device, such as... Figure 12 As shown.
[0130] The data processing device includes: an intent recognition module 1201, a rights determination module 1202, a strategy determination module 1203, and a redemption interaction module 1204, wherein: The intent recognition module 1201, in response to a user's rights query request, determines the user intent and rights usage scenario corresponding to the rights query request; The rights determination module 1202 determines candidate rights that match the user based on the user's intent and the user's rights redemption qualifications, and determines multiple candidate rights types corresponding to the rights usage scenarios; The strategy determination module 1203 uses a preset large language model to filter the candidate rights based on the user's rights redemption qualifications and the multiple candidate rights types, to obtain multiple target rights, and constructs a rights combination strategy based on the logical dependencies between the multiple target rights. The redemption interaction module 1204 performs redemption interaction processing with the user based on the aforementioned rights combination strategy.
[0131] In this embodiment of the specification, the rights determination module 1202 is used for: From the first candidate benefits, select the second candidate benefits that match the user's intent; Based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, the second candidate rights are screened, and based on the screened second candidate rights, candidate rights that match the user are determined.
[0132] In this embodiment of the specification, the rights determination module 1202 is used for: Obtain the first candidate right corresponding to the right usage scenario, and the right description information corresponding to the first candidate right; The feature extraction process is performed on the right description information corresponding to the first candidate right to obtain a first feature vector, and the feature extraction process is performed on the user intent to obtain a second feature vector; Based on the semantic similarity between the first feature vector and the second feature vector, the first candidate interest is screened, and based on the screened first candidate feature vector, the second candidate interest is determined.
[0133] In this embodiment of the specification, the rights determination module 1202 is used for: Based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, the user level verification and user points verification are performed on the second candidate rights respectively to obtain the level verification result and the points verification result. Based on the grade verification result and the score verification result, the second candidate rights are screened.
[0134] In this embodiment of the specification, the exchange interaction module 1204 is used for: Using the preset large language model, based on the user intent and the logical dependencies between multiple target benefits in the benefit combination strategy, target interaction data is generated; Based on the target interaction data, the user is engaged in a rights exchange interaction process.
[0135] In this embodiment of the specification, the exchange interaction module 1204 is used for: Obtain the interactive component corresponding to each of the target benefits, and the interactive component is used to trigger the execution of the benefit redemption operation; Using the preset large language model, the target interaction data is generated based on the user intent, the interactive components corresponding to the target benefit, and the logical dependencies between multiple target benefits in the benefit combination strategy.
[0136] In the embodiments described in this specification, the strategy determination module 1203 is used for: Using a preset enhanced retrieval algorithm and based on a preset knowledge base, supplementary knowledge of the rights corresponding to the candidate rights is obtained, and preset prompt information is constructed based on the supplementary knowledge of the rights. Using the preset large language model, based on the preset prompt information, the user's rights redemption qualifications, and the multiple candidate rights types, the candidate rights are filtered to obtain the multiple target rights.
[0137] In this embodiment of the specification, the intent recognition module 1201 is used for: Obtain the rights query data corresponding to the rights query request; Using the preset large language model, based on the rights query data and the user's rights redemption qualifications, the user's intent is processed to obtain the user intent corresponding to the rights query request.
[0138] In this embodiment of the specification, the user's eligibility for redeeming rights meets the redemption conditions corresponding to the rights combination strategy, and the redemption conditions corresponding to the rights combination strategy are determined based on the redemption conditions corresponding to the multiple target rights.
[0139] This specification provides a data processing device that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate rights types corresponding to the rights usage scenario. Using a pre-defined large language model, it filters the candidate rights based on the user's rights redemption eligibility and the multiple candidate rights types, obtaining multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to process rights redemption. This approach, on the one hand, allows for quick and accurate filtering of redeemable candidate rights matching the user's intent based on their intent and redemption eligibility, improving the efficiency and accuracy of subsequent target rights filtering. On the other hand, it leverages the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate rights types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model while providing more accurate recommendations for combined rights, increasing the success rate of rights redemption.
[0140] The above are the data processing apparatuses provided in the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide a data processing device, such as... Figure 13 As shown.
[0141] The data processing device can provide terminal equipment or servers, etc., for the above embodiments.
[0142] like Figure 13 As shown, device 1300 mainly consists of a communication interface 1302, a user interface 1304, a processor 1306, and a data storage 1308. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 1310. The communication interface 1302 enables device 1300 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 1302 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 1302 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 1302 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 1302 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.
[0143] User interface 1304 includes receiving user input and providing output to the user. Therefore, user interface 1304 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 1304 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 1304 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 1300 may support remote access from other devices via communication interface 1302 or another physical interface (not shown). User interface 1304 may be configured to receive user input, the position and movement of which may be indicated by indicators or cursors described herein. User interface 1304 can also be configured as a display device for rendering or displaying text fragments.
[0144] Processor 1306 may include one or more general-purpose processors and / or special-purpose processors.
[0145] Data storage 1308 may include one or more volatile storage components and may be integrated wholly or partially with processor 1306. Data storage 1308 may include removable and non-removable components.
[0146] Processor 1306 is capable of executing program instructions 1318 (e.g., compiled or uncompiled program logic and / or machine code) stored in data store 1308 to perform the various functions described herein. Data store 1308 may contain a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 1300, enable device 1300 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Execution of program instructions 1318 by processor 1306 may result in processor 1306 using data 1312.
[0147] For example, program instructions 1318 may include an operating system 1322 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 1300 and one or more applications 1320 (e.g., a browser, social application, or game application). Similarly, data 1312 may include operating system data 1316 and application data 1314. Operating system data 1316 is primarily accessible to the operating system 1322, while application data 1314 is primarily accessible to one or more applications 1320. Application data 1314 may reside in a file system visible or hidden from the user of device 1300.
[0148] Application 1320 can communicate with operating system 1312 through one or more application programming interfaces (APIs). These APIs help application 1320 read and / or write application data 1314, transmit or receive information via communication interface 1302, receive or display information on user interface 1304, etc.
[0149] In some terminology, application 1320 may be simply referred to as "app". Furthermore, application 1320 can be downloaded to device 1300 through one or more online app stores or app markets. However, applications can also be installed on device 1300 in other ways, such as through a web browser or a physical interface on device 1300 (e.g., a USB port).
[0150] Specifically, in this embodiment, the data processing device includes a memory and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the data processing device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following: In response to a user's rights inquiry request, determine the user's intent and rights usage scenario corresponding to the rights inquiry request; Based on the user's intent and the user's eligibility for redemption of rights, determine candidate rights that match the user, and determine multiple candidate right types corresponding to the right usage scenarios; Using a pre-defined large language model, the candidate rights are filtered based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between the multiple target rights, a rights combination strategy is constructed. Based on the aforementioned benefits combination strategy, the user will be interacted with to exchange benefits.
[0151] The various embodiments in this specification 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 data processing device 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.
[0152] This specification provides a data processing device that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate rights types corresponding to the rights usage scenario. Using a pre-defined large language model, it filters the candidate rights based on the user's rights redemption eligibility and multiple candidate rights types, obtaining multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to process rights redemption. This approach, on the one hand, allows for quick and accurate filtering of redeemable candidate rights matching the user's intent based on their intent and redemption eligibility, improving the efficiency and accuracy of subsequent target rights filtering. On the other hand, it leverages the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate rights types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model while providing more accurate recommendations for combined rights, increasing the success rate of rights redemption.
[0153] Furthermore, based on the above Figures 1 to 11 This specification also provides a storage medium for storing computer-executable instruction information in one or more embodiments. In one specific embodiment, the storage medium may be a USB flash drive, optical disc, hard disk, etc. When the computer-executable instruction information stored in the storage medium is executed by a processor, it can realize the following process: In response to a user's rights inquiry request, determine the user's intent and rights usage scenario corresponding to the rights inquiry request; Based on the user's intent and the user's eligibility for redemption of rights, determine candidate rights that match the user, and determine multiple candidate right types corresponding to the right usage scenarios; Using a pre-defined large language model, the candidate rights are filtered based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between the multiple target rights, a rights combination strategy is constructed. Based on the aforementioned benefits combination strategy, the user will be interacted with to exchange benefits.
[0154] The various embodiments in this specification 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 above-described storage medium embodiment is basically similar to the method embodiment, so the description is relatively simple; relevant parts can be referred to the description of the method embodiment.
[0155] This specification provides a storage medium that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate rights types corresponding to the rights usage scenario. Using a pre-defined large language model, based on the user's rights redemption eligibility and multiple candidate rights types, it filters the candidate rights to obtain multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to redeem rights. In this way, on the one hand, it can quickly and accurately filter out redeemable candidate rights matching the user's intent based on the user's intent and rights redemption eligibility, improving the efficiency and accuracy of subsequent target rights filtering. On the other hand, it can utilize the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate rights types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model and can more accurately recommend combined rights to the user, improving the success rate of rights redemption.
[0156] Furthermore, based on the above Figures 1 to 11 This specification also provides one or more embodiments of a computer program product, including a computer program, which, when executed by a processor, can perform the following processes: In response to a user's rights inquiry request, determine the user's intent and rights usage scenario corresponding to the rights inquiry request; Based on the user's intent and the user's eligibility for redemption of rights, determine candidate rights that match the user, and determine multiple candidate right types corresponding to the right usage scenarios; Using a pre-defined large language model, the candidate rights are filtered based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between the multiple target rights, a rights combination strategy is constructed. Based on the aforementioned benefits combination strategy, the user will be interacted with to exchange benefits.
[0157] The various embodiments in this specification 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 above-described embodiment of a computer program product is relatively simple in description because it is fundamentally similar to the method embodiment; relevant parts can be referred to the description of the method embodiment.
[0158] This specification provides a computer program product that, in response to a user's rights query request, determines the user's intent and rights usage scenario corresponding to the request. Based on the user's intent and rights redemption eligibility, it identifies candidate rights matching the user and multiple candidate right types corresponding to the rights usage scenario. Using a pre-defined large language model, it filters the candidate rights based on the user's rights redemption eligibility and multiple candidate right types, obtaining multiple target rights. Based on the logical dependencies between these target rights, it constructs a rights combination strategy and interacts with the user to process rights redemption. This approach, on the one hand, allows for quick and accurate filtering of redeemable candidate rights matching the user's intent based on their intent and redemption eligibility, improving the efficiency and accuracy of subsequent target right filtering. On the other hand, it leverages the deep semantic understanding capabilities of the pre-defined large language model to filter multiple candidate right types corresponding to the rights usage scenario and construct a rights combination strategy based on the logical dependencies between the filtered target rights. This overcomes the illusion of a large model while providing more accurate recommendations for combined rights, increasing the success rate of rights redemption.
[0159] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0160] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.
[0161] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.
[0162] 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.
[0163] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.
[0164] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0165] The embodiments described herein are illustrated with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0166] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0167] These computer program instructions may also be loaded onto a computer or other programmable device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0168] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0169] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0170] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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 herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0171] 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 process, method, article, or apparatus. Without further limitations, the presence of additional identical elements in the process, method, article, or apparatus that includes said elements is not excluded.
[0172] It should also be noted that the terms "one," "an," and "the" do not specifically refer to the singular; they can also include the plural. Ordinal numbers such as "first," "second," etc., do not necessarily indicate order; often they are used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.
[0173] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving data sent by B, or it can be understood as A indirectly receiving data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending data directly to A, or it can be understood as B indirectly sending data to A through other entities such as C. Here, C can be a single subject, or two or more subjects.
[0174] This specification uses specific terms to describe embodiments thereof. Terms such as "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations within this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those different embodiments or examples, without contradiction.
[0175] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.
[0176] It should be noted that the user data obtained in this manual is authorized by the user and does not involve user privacy.
[0177] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0178] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0179] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. 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 in the method embodiments.
[0180] The above description is merely an embodiment of this specification and is not intended to limit this document. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.
Claims
1. A data processing method, comprising: In response to a user's rights inquiry request, determine the user's intent and rights usage scenario corresponding to the rights inquiry request; Based on the user's intent and the user's eligibility for redemption of rights, determine candidate rights that match the user, and determine multiple candidate right types corresponding to the right usage scenarios; Using a pre-defined large language model, the candidate rights are filtered based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between the multiple target rights, a rights combination strategy is constructed. Based on the aforementioned benefits combination strategy, the user will be interacted with to exchange benefits.
2. The method according to claim 1, wherein determining the candidate benefits matching the user based on the user's intent and the user's benefit redemption eligibility includes: From the first candidate benefits, select the second candidate benefits that match the user's intent; Based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, the second candidate rights are screened, and based on the screened second candidate rights, candidate rights that match the user are determined.
3. The method according to claim 2, wherein selecting a second candidate benefit that matches the user's intent from the first candidate benefits comprises: Obtain the first candidate right corresponding to the right usage scenario, and the right description information corresponding to the first candidate right; The feature extraction process is performed on the right description information corresponding to the first candidate right to obtain a first feature vector, and the feature extraction process is performed on the user intent to obtain a second feature vector; Based on the semantic similarity between the first feature vector and the second feature vector, the first candidate interest is screened, and based on the screened first candidate feature vector, the second candidate interest is determined.
4. The method according to claim 2, wherein the step of filtering the second candidate benefit based on the user's corresponding benefit redemption qualification and the benefit redemption conditions corresponding to the second candidate benefit includes: Based on the user's corresponding rights redemption qualifications and the rights redemption conditions corresponding to the second candidate rights, the user level verification and user points verification are performed on the second candidate rights respectively to obtain the level verification result and the points verification result. Based on the grade verification result and the score verification result, the second candidate rights are screened.
5. The method according to claim 1, wherein the step of performing the rights exchange interaction with the user based on the rights combination strategy includes: Using the preset large language model, based on the user intent and the logical dependencies between multiple target benefits in the benefit combination strategy, target interaction data is generated; Based on the target interaction data, the user is engaged in a rights exchange interaction process.
6. The method according to claim 5, wherein generating target interaction data using the preset large language model, based on the user intent and the logical dependencies between multiple target benefits in the benefit combination strategy, includes: Obtain the interactive component corresponding to each of the target benefits, and the interactive component is used to trigger the execution of the benefit redemption operation; Using the preset large language model, the target interaction data is generated based on the user intent, the interactive components corresponding to the target benefit, and the logical dependencies between multiple target benefits in the benefit combination strategy.
7. The method according to claim 1, wherein the step of using a preset large language model to filter the candidate rights based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights includes: Using a preset enhanced retrieval algorithm and based on a preset knowledge base, supplementary knowledge of the rights corresponding to the candidate rights is obtained, and preset prompt information is constructed based on the supplementary knowledge of the rights. Using the preset large language model, based on the preset prompt information, the user's rights redemption qualifications, and the multiple candidate rights types, the candidate rights are filtered to obtain the multiple target rights.
8. The method according to claim 1, wherein determining the user intent corresponding to the rights query request includes: Obtain the rights query data corresponding to the rights query request; Using the preset large language model, based on the rights query data and the user's rights redemption qualifications, the user's intent is processed to obtain the user intent corresponding to the rights query request.
9. The method according to claim 1, wherein the user's eligibility for redeeming rights meets the redemption conditions corresponding to the rights combination strategy, and the redemption conditions corresponding to the rights combination strategy are determined based on the redemption conditions corresponding to the plurality of target rights.
10. A data processing apparatus, comprising: The intent recognition module, in response to a user's rights query request, determines the user's intent and the rights usage scenario corresponding to the rights query request; The rights determination module determines candidate rights that match the user based on the user's intent and the user's rights redemption qualifications, and determines multiple candidate rights types corresponding to the rights usage scenarios; The strategy determination module uses a preset large language model to filter the candidate rights based on the user's rights redemption qualifications and the multiple candidate rights types, thereby obtaining multiple target rights, and constructs a rights combination strategy based on the logical dependencies between the multiple target rights. The redemption interaction module, based on the aforementioned benefits combination strategy, performs benefits redemption interaction processing with the user.
11. A data processing apparatus, the data processing comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to: In response to a user's rights inquiry request, determine the user's intent and rights usage scenario corresponding to the rights inquiry request; Based on the user's intent and the user's eligibility for redemption of rights, determine candidate rights that match the user, and determine multiple candidate right types corresponding to the right usage scenarios; Using a pre-defined large language model, the candidate rights are filtered based on the user's rights redemption qualifications and the multiple candidate rights types to obtain multiple target rights. Based on the logical dependencies between the multiple target rights, a rights combination strategy is constructed. Based on the aforementioned benefits combination strategy, the user will be interacted with to exchange benefits.
