Target recommendation method, apparatus, device, and medium
By using large language models and virtual resource recommendation methods, the problem of insufficient understanding of implicit or complex needs in intelligent recommendation systems is solved, achieving high-precision target object recommendation and improving the reliability and user experience of the recommendation system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUTU NETWORK TECH (SHENZHEN) CO LTD
- Filing Date
- 2026-03-23
- Publication Date
- 2026-07-24
AI Technical Summary
In existing technologies, intelligent recommendation systems struggle to effectively understand and reasonably recommend implicit or complex user needs, resulting in insufficient accuracy and reliability of target recommendations.
A virtual resource recommendation method based on a large language model is adopted. Through semantic analysis and deep reasoning, the virtual resource demand information input by the user is obtained. The target recommendation factors are obtained from the knowledge base using a pre-set virtual resource recommendation model, and the target virtual resource objects are selected based on the factor parameters.
It enables high-precision processing of complex and implicit requirements, improves the reliability and accuracy of target object recommendations, and enhances the user experience.
Smart Images
Figure CN122451206A_ABST
Abstract
Description
Technical Field
[0001] This disclosure generally relates to the field of data processing technology, and specifically to a method, apparatus, device, and medium for target recommendation. Background Technology
[0002] With the widespread application of artificial intelligence, intelligent recommendation has also seen significant development in the financial sector. Related technologies typically rely on keyword matching and rule engines for targeted recommendations, such as recommending stocks based on recommendation factors directly associated with keywords like "undervalued" and "high-growth." However, these technologies struggle to effectively understand and appropriately recommend stocks based on implicit or complex needs. Summary of the Invention
[0003] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a target recommendation method, apparatus, device and medium that realizes high-precision recommendation of target objects based on user needs.
[0004] In a first aspect, embodiments of this application provide a target recommendation method, including: Obtain the user's input information regarding virtual resource requirements; Using a pre-defined virtual resource recommendation model, a target recommendation factor is obtained from a virtual resource demand knowledge base based on the virtual resource demand information; the virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information. Obtain the factor parameters corresponding to the target recommendation factor based on virtual resource demand information; Based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor, target virtual resource objects are selected.
[0005] Secondly, embodiments of this application provide a target recommendation device, including: The acquisition module is used to acquire the virtual resource requirement information input by the user; The analysis module is used to obtain target recommendation factors from the virtual resource demand knowledge base based on the virtual resource demand information using a preset virtual resource recommendation model; the virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information. The determination module is used to obtain the factor parameters corresponding to the target recommendation factor based on the virtual resource demand information; The filtering module is used to filter target virtual resource objects based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor.
[0006] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in embodiments of this application.
[0007] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in embodiments of this application.
[0008] Fifthly, embodiments of this application provide a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the method described in embodiments of this application.
[0009] The target recommendation method, apparatus, device, and medium provided in this application acquire virtual resource demand information input by the user, utilize a preset virtual resource recommendation model, and obtain target recommendation factors from a virtual resource demand knowledge base based on the virtual resource demand information; obtain factor parameters corresponding to the target recommendation factors based on the virtual resource demand information; and filter target virtual resource objects based on the target recommendation factors and their corresponding factor parameters, thereby achieving high-precision recommendation of target objects based on user needs. Simultaneously, through the preset virtual resource recommendation model, complex and implicit demand processing is achieved, effectively improving the reliability and accuracy of target object recommendations and enhancing user experience.
[0010] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0011] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This diagram illustrates the implementation environment architecture of the target recommendation method provided in an embodiment of this application. Figure 2 A flowchart illustrating a target recommendation method provided in an embodiment of this application is shown; Figure 3 A flowchart illustrating a target recommendation method provided in another embodiment of this application is shown; Figure 4 A schematic diagram of the target recommendation device provided in one embodiment of this application is shown; Figure 5 A schematic diagram of the structure of a computer system suitable for implementing an electronic device or server according to embodiments of this application is shown. Detailed Implementation
[0012] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.
[0013] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0014] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.
[0015] A knowledge base is a collection of information stored in various computers, organized and structured using specific methods to form a comprehensive and organized knowledge cluster that is easy to operate, use, and retrieve and utilize knowledge. In other words, a knowledge base is a collection of information, including various forms of data and information, such as text, images, audio, and video.
[0016] Large Language Models (LLMs) are deep learning models trained on massive amounts of text data, enabling them to generate natural language text or understand the meaning of language text. These models can provide in-depth knowledge and language production on a wide range of topics through training on large datasets. Their core idea is to learn patterns and structures of natural language through large-scale unsupervised training, mimicking human language cognition and generation processes to some extent.
[0017] For the specific implementation environment of the target recommendation method proposed in this application, please refer to [link / reference]. Figure 1 . Figure 1 The implementation environment architecture diagram of the target recommendation method provided in the embodiments of this application is shown.
[0018] like Figure 1 As shown, the implementation environment architecture includes: terminal device 101 and server 102.
[0019] Terminal device 101 is used to run an application client and provide an interactive interface to the user. This interface receives user-inputted virtual resource requests, displays recommendation factors and target virtual resource objects, etc. Terminal device 101 can be a desktop computer, laptop computer, smartphone, tablet computer, e-book reader, smart glasses, smartwatch, in-vehicle device, ultra-mobile personal computer (UMPC), netbook, cellular phone, personal digital assistant (PDA), augmented reality (AR), virtual reality (VR) device, etc., but is not limited to these.
[0020] Server 102 receives virtual resource demand information sent by terminal device 101 and executes the target recommendation method proposed in this application embodiment to obtain target virtual resource objects recommended for the virtual resource demand information.
[0021] Server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminal device 101 and server 102 are connected directly or indirectly via wired or wireless communication. Optionally, the aforementioned wireless or wired network uses standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to any combination of Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), mobile, wired or wireless network, private network, or virtual private network.
[0022] The target recommendation method proposed in this application can be implemented by a target recommendation device, which can be installed on a terminal device or a server.
[0023] To further illustrate the technical solutions provided in the embodiments of this application, a detailed description is provided below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiments of this application provide method operation instruction steps as shown in the following embodiments or drawings, the method may include more or fewer operation instruction steps based on conventional or non-creative effort. In steps where there is no logically necessary causal relationship, the execution order of these steps is not limited to the execution order provided in the embodiments of this application. In actual processing or when the device executes the method, it may be executed sequentially or in parallel according to the method shown in the embodiments or drawings.
[0024] It should be noted that the acquisition or use of data in the embodiments of this application requires the user's consent. The relevant data can only be obtained after the user's authorization and permission, and the acquisition or use of the data complies with the laws and regulations of the relevant regions.
[0025] Please refer to Figure 2 , Figure 2 A flowchart illustrating a target recommendation method provided in an embodiment of this application is shown. Figure 2 As shown, the method includes: Step 201: Obtain the virtual resource requirement information input by the user.
[0026] It should be noted that virtual resource demand information describes the user's requirements for the recommendation results (i.e., the virtual resource objects to be recommended to the user). This virtual resource demand information can be text or voice; no specific limitation is made here.
[0027] Furthermore, a virtual resource object can be an object that can be deemed valuable in a specific usage scenario. For example, it could be a transaction object in a buying and selling scenario; an object deemed valuable by both parties in a barter scenario; stocks, funds, or other objects used as transaction objects in a financial scenario; or game resource objects (such as game equipment or game skins) set up in a game scenario that can be used as transaction objects. No specific limitations are imposed. This application embodiment primarily uses stocks as an example of a virtual resource object for illustration; other cases are not listed.
[0028] Taking stocks as an example of virtual resource objects, the virtual resource demand information can specifically be stock selection demand information. Specifically, the financial investment application client can provide an intelligent question-and-answer system. Users can input their stock selection demand information in the display interface of the intelligent question-and-answer system on the financial investment application client. This stock selection demand information represents the user's proposed stock screening needs. The intelligent question-and-answer system processes the stock selection demand information through a virtual resource recommendation model and outputs target virtual resource objects to recommend to the user.
[0029] Step 202: Using a pre-defined virtual resource recommendation model, based on virtual resource demand information, obtain target recommendation factors from the virtual resource demand knowledge base; the virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information.
[0030] It should be noted that the preset virtual resource recommendation model is a trained Large Language Model (LLM) for virtual resource recommendation. This preset virtual resource recommendation model, after training, can perform various analytical operations on virtual resource demand information, such as semantic analysis, matching analysis, and parameter analysis; however, this application does not impose specific limitations on these operations.
[0031] Among them, virtual resource recommendation factors refer to factors that can describe virtual resource objects or the characteristics of virtual objects in different dimensions. It can be understood that virtual resource recommendation factors can serve as screening conditions (or conditional factors) in a target virtual resource object screening system to filter virtual resource objects. By limiting certain screening conditions through virtual resource recommendation factors, specific recommended objects can be selected from all candidate virtual resource objects. Furthermore, target recommendation factors refer to virtual resource recommendation factors that can be extracted from virtual resource demand information.
[0032] Taking virtual resource objects as stock objects as an example, virtual resource recommendation factors refer to factors that can describe the characteristics of stock objects or stock objects in different dimensions. For example, virtual resource recommendation factors can be stock selection factors, including indicators of stock objects in different dimensions, such as stock price, trading volume, turnover rate, MA indicator, MACD indicator, price-earnings ratio, net profit, growth rate, etc.
[0033] The virtual resource demand knowledge base refers to the knowledge base built for virtual resource recommendation factors. The virtual resource demand knowledge base can be built in advance and also supports operations such as adding, deleting, modifying and querying it during use.
[0034] Specifically, the virtual resource demand knowledge base may include at least one knowledge entry. Each knowledge entry may include: a baseline demand information and at least one virtual resource recommendation factor associated with that baseline demand information. The baseline demand information refers to virtual resource demand information pre-built or accumulated through historical knowledge question-and-answer sessions, and the virtual resource recommendation factor associated with the baseline demand information refers to a virtual resource recommendation factor identified from the baseline demand information.
[0035] Furthermore, a knowledge entry may also include a recommendation intent corresponding to the baseline demand information. This recommendation intent is used to describe the demand direction represented by the baseline demand information, that is, to describe the recommendation direction of what kind of virtual resources should be recommended to the user based on the baseline demand information.
[0036] Taking virtual resource objects as stock objects as an example, a knowledge entry in the virtual resource demand knowledge base can include a benchmark demand information, at least one virtual resource recommendation factor (such as a stock selection factor) associated with the benchmark demand information, and the recommendation intent (i.e., recommendation direction, such as stock selection direction type) corresponding to the benchmark demand information.
[0037] For example, a knowledge entry in the virtual resource requirements knowledge base specifically includes: Benchmark demand information: Helps to screen stocks with good growth potential; Virtual resource recommendation factors (stock selection factors): Price-to-Earnings Ratio, Net Profit; Recommendation intention (stock selection direction type): growth stocks.
[0038] After obtaining virtual resource demand information, a preset virtual resource recommendation model can be used to retrieve benchmark demand information associated with the virtual resource demand information from the virtual resource demand knowledge base. Then, the virtual resource recommendation factors associated with the retrieved benchmark demand information are used as target recommendation factors.
[0039] In one feasible embodiment, such as Figure 3 As shown, step 202 involves using a pre-defined virtual resource recommendation model to obtain target recommendation factors from the virtual resource demand knowledge base based on virtual resource demand information, including: Step 2021: Based on the virtual resource demand knowledge base, perform semantic retrieval on the virtual resource demand information to obtain multiple sets of initial recommendation information. Each set of initial recommendation information includes an initial recommendation benchmark demand information and its corresponding initial recommendation factor and initial recommendation intent.
[0040] In this process, virtual resource demand information is input into a preset virtual resource recommendation model. The preset virtual resource recommendation model retrieves knowledge from the virtual resource demand knowledge base based on the virtual resource demand to obtain multiple sets of initial recommendation information.
[0041] The initial recommendation information is grouped into <initial recommendation intent - initial recommendation baseline demand information - initial recommendation factor>, where the initial recommendation intent is the recommendation intent corresponding to the initial recommendation baseline demand information.
[0042] Specifically, the virtual resource recommendation model may include a pre-trained language model. This pre-trained language model can calculate the semantic similarity between virtual resource demand information and various baseline demand information in the virtual resource demand knowledge base, thereby obtaining initial recommendation baseline demand information based on semantic similarity. It is understood that this pre-trained language model can be a BERT (Bidirectional Encoder Representations from Transformers) model.
[0043] The recommendation intent is used to describe the demand direction represented by the benchmark demand information, that is, to describe the recommendation direction of what kind of virtual resources to recommend to users based on the benchmark demand information. It may include, but is not limited to, valuation-related and growth-related resources. The specific content can be determined according to the actual content of the benchmark demand information. This application does not make any specific limitations on this.
[0044] In some feasible embodiments, the process of constructing a virtual resource demand knowledge base may specifically include: acquiring massive amounts of benchmark demand information and its corresponding recommendation factors; preprocessing the benchmark demand information and its corresponding recommendation factors, such as word segmentation, text cleaning, and standardization, to obtain the preprocessed benchmark demand information and its corresponding recommendation factors; then, performing cluster analysis on the benchmark demand information to obtain a class of benchmark demand information with high similarity; and determining the recommendation intent corresponding to this class of benchmark demand information by the recommendation intent corresponding to each benchmark demand information in this class, for example, taking the same type of recommendation intent that each benchmark demand information has as the recommendation intent corresponding to this class of benchmark demand information; and standardizing the obtained recommendation intent, etc., which are not specifically limited in this application. K-means clustering algorithm can be used to cluster the benchmark demand information. The clustering distance of the K-means clustering algorithm is used to calculate the similarity between any two benchmark demand information pieces. If the clustering distance is greater than a preset clustering distance, it indicates that the two benchmark demand information pieces are not similar.
[0045] It should be understood that when acquiring initial recommended virtual information, the virtual resource demand knowledge base can calculate the semantic similarity between the virtual resource demand information and the existing benchmark demand information in the virtual resource demand knowledge base through a pre-trained language model, so as to achieve semantic retrieval of virtual resource demand. From the massive benchmark demand information in the virtual resource demand knowledge base, benchmark demand information with semantic similarity greater than or equal to a preset similarity threshold is selected as the basis for further semantic analysis by the subsequent preset virtual resource recommendation model.
[0046] Furthermore, after obtaining at least one benchmark demand information whose semantic similarity to the virtual resource demand information is greater than or equal to a preset similarity threshold, the recommendation intent corresponding to these benchmark demand information and its corresponding recommendation factor are obtained to generate a set of initial recommendation information.
[0047] In one specific embodiment, the initial recommendation information obtained by the preset virtual resource recommendation model from the virtual resource demand knowledge base can be as shown in Table 1 (taking historical virtual resource demand information as an example).
[0048] Table 1
[0049] In a preferred embodiment, historical virtual resource demand information and its corresponding recommendation intent tags and recommendation factors can be used as a training set. The recommendation intent tags corresponding to the historical virtual resource demand information can be determined by the user based on candidate recommendation intent tags; that is, feedback information corresponding to the virtual resource demand information is tracked and obtained. This feedback information is used to show the mapping relationship between the virtual resource demand information and the user's actual demand intent.
[0050] In other words, after the preset virtual resource recommendation model obtains at least one type of candidate recommendation intent corresponding to the virtual resource demand information, it can send at least one type of candidate recommendation intent as a feedback label to the client. The user selects at least one feedback label representing the candidate recommendation intent according to their actual demand intent. In response to the user's selection operation, the selected feedback label can be used as feedback information, that is, feedback information is obtained to reflect the mapping relationship between the virtual resource demand information and the user's actual stock selection intent. It can be understood that the candidate recommendation intent corresponding to the selected feedback label can represent the user's actual stock selection intent. Finally, the recommendation intent label corresponding to the virtual resource demand information is generated according to the feedback information (i.e. the selected feedback label) and used as historical data to train the virtual resource demand knowledge base.
[0051] Step 2022: Using a preset virtual resource recommendation model, evaluate the correlation between the virtual resource demand information and each set of initial recommendation information to obtain the correlation evaluation value corresponding to each set of initial recommendation information.
[0052] In other words, after retrieving multiple sets of initial recommendation information corresponding to virtual resource demand information from the virtual resource demand knowledge base, at least one set of candidate recommendation information is further filtered from the initial recommendation information using a pre-defined virtual resource recommendation model. Specifically, the virtual resource recommendation model may also include a Large Language Model (LLM), which evaluates the correlation between the initial recommendation demand information and the virtual resource demand information in the initial recommendation information to perform a secondary filtering of the initial recommendation information.
[0053] In other words, in order to further improve the accuracy of the target recommendation factors, the pre-set virtual resource recommendation model proposed in this application further quantifies and scores the initial recommendation information obtained from the initial screening, so as to automatically filter out the initial recommendation information with low relevance and retain the recommendation benchmark demand information with high relevance and its corresponding recommendation factors, thereby significantly improving the matching accuracy of the target recommendation factors.
[0054] It should be understood that the correlation evaluation value between the initial recommendation baseline demand information and the virtual resource demand information can be a normalized score, such as a score between 0 and 1. The higher the score, the higher the semantic correlation between the initial recommendation baseline demand information and the virtual resource demand information, that is, the more likely the recommendation factor corresponding to the initial recommendation baseline demand information can be used as the target recommendation factor of the virtual resource demand information.
[0055] For example, given the multiple baseline requirements in Table 1, the relevance assessment values are shown in Table 2: Table 2
[0056] Step 2023: Filter the initial recommendation information based on the relevance evaluation value to obtain multiple sets of candidate recommendation information; each set of candidate recommendation information includes a candidate recommendation baseline demand information and its corresponding candidate recommendation factor and candidate recommendation intent.
[0057] Specifically, by using a correlation evaluation threshold, initial recommendation information with a correlation evaluation value less than the correlation evaluation threshold is filtered out, that is, initial recommendation information with a correlation evaluation value greater than or equal to the correlation evaluation threshold is selected and retained.
[0058] For example, if the relevance assessment threshold is 0.7, then Table 2 will be filtered to obtain Table 3 (excluding baseline requirement information entries with relevance assessment values less than 0.7): Table 3
[0059] Step 2024: For each type of candidate recommendation intent, deduplication is performed on multiple sets of candidate recommendation information to obtain the target recommendation benchmark demand information corresponding to each candidate recommendation intent, and the target recommendation factor corresponding to the target recommendation benchmark demand information.
[0060] It should be noted that, as shown in Tables 1-3 above, the content of the benchmark demand information (or candidate recommended benchmark demand information) "help screen stocks with good growth potential" and "help screen stocks with good growth potential" are basically the same, with only slight differences in expression. Therefore, in order to reduce redundant factor calls and invalid calculations in the later stages, this application further deduplicates the highly repetitive benchmark demand information (or candidate recommended benchmark demand information).
[0061] Specifically, for all candidate recommendation benchmark requirement information under each candidate recommendation intent, the similarity between each candidate recommendation benchmark requirement information is calculated; the similarity includes string repetition rate and / or cosine similarity; candidate recommendation benchmark requirement information with string repetition rate greater than or equal to the first similarity threshold and / or cosine similarity greater than or equal to the second similarity threshold is filtered to obtain target recommendation benchmark requirement information; target recommendation factor is determined based on target recommendation benchmark requirement information.
[0062] In other words, for multiple candidate recommendation information under each candidate recommendation intent, the string repetition rate, or cosine similarity, or string repetition rate and cosine similarity, are calculated for any two candidate recommendation baseline requirement information. Specifically, a text hash algorithm can be used to calculate the string repetition rate between any two candidate recommendation baseline requirement information, and a cosine similarity algorithm can be used to calculate the cosine similarity between any two candidate recommendation baseline requirement information.
[0063] If the string repetition rate between any two candidate recommendation benchmark demand information is greater than or equal to the first similarity threshold, it indicates that the text recorded in the two candidate recommendation benchmark demand information is highly similar, such as "help to screen stocks with good growth potential" and "help to screen stocks with good growth potential". If the cosine similarity between any two candidate recommendation benchmark demand information is greater than or equal to the first similarity threshold, it indicates that the content recorded in the two candidate recommendation benchmark demand information is substantially highly similar, such as "high ROE stocks" and "high return on equity stocks".
[0064] Based on this, after using the correlation evaluation value to screen the initial recommendation benchmark requirement information, the similarity is further used to deduplicate the candidate recommendation benchmark requirement information, which can effectively reduce the repeated calculation of the target recommendation factor, reduce the calculation duplication and multiple calls caused by the duplication of candidate recommendation benchmark requirement information, and reduce the system load.
[0065] In a preferred embodiment, to further ensure the completeness and comprehensive coverage of the target recommendation factors, after the above-mentioned screening based on relevance evaluation value and similarity, the number of target recommendation factors corresponding to each type of candidate recommendation intent is further determined. If the number of target recommendation factors corresponding to each type of candidate recommendation intent is less than a preset number, it indicates that the current candidate recommendation intent has not been sufficiently recommended. In this case, the step of determining at least one set of initial recommendation information corresponding to the virtual resource demand information from the virtual resource demand knowledge base for each type of candidate recommendation intent is returned, and the relevance evaluation threshold, the first similarity threshold, and / or the second similarity threshold are redefined. The redefined relevance evaluation threshold, the first similarity threshold, and / or the second similarity threshold are all less than the previously applied relevance evaluation threshold, the first similarity threshold, and / or the second similarity threshold. That is, by lowering the threshold conditions, the range of target recommendation factors that are screened and retained is expanded, thereby ensuring that the final number of target recommendation factors output for each type of candidate recommendation intent is greater than or equal to the preset number, thus ensuring that each type of candidate recommendation intent has a sufficient number of target recommendation factors, ensuring that each type of candidate recommendation intent can be fully described by a sufficient number of target recommendation factors, and thus selecting accurate target objects.
[0066] It should be understood that, in the embodiments of this application, by introducing a large language model to perform semantic analysis on virtual resource demand information, multi-level and deep semantic parsing of the natural language input by the user is possible. This enables intelligent stock selection recommendations to go beyond keyword or shallow similarity analysis, and to effectively understand the user's implicit and complex demand information. This effectively ensures that the recommendation results match the user's actual recommendation intent, improves the accuracy and reliability of the recommendation results, and enhances the user experience.
[0067] Moreover, in terms of semantic understanding, the embodiments of this application upgrade the analysis based on keywords and shallow similarity in the prior art to a combination of implicit intent recognition using knowledge base and deep reasoning using LLM, thereby enabling the handling of complex composite needs, such as virtual resource demand information that is "anti-inflationary and has the potential for technological breakthroughs".
[0068] It should also be understood that, in order to ensure the comprehensiveness of the target recommendation factors, that is, to ensure that no recommendation intent contained in the virtual resource demand information is omitted, this application performs a separate screening of the benchmark demand information for each type of recommendation intent. For example, it screens separately for "valuation", "growth" and "risk", thereby effectively ensuring that each type of recommendation intent contained in the virtual resource demand information can obtain the target recommendation factor.
[0069] In a preferred embodiment, feedback information corresponding to virtual resource demand information is tracked and obtained. This feedback information is used to indicate the mapping relationship between the virtual resource demand information and the actual recommendation factors. Specifically, after the preset virtual resource recommendation model selects target recommendation factors, these factors are sent to the client. Users can select tags for at least one target recommendation factor according to their actual recommendation intentions. In response to the user's selection, feedback information is obtained to indicate the mapping relationship between the virtual resource demand information and the actual recommendation factors. Recommendation factors corresponding to the virtual resource demand information are generated based on the feedback information and stored as historical data in the virtual resource demand knowledge base.
[0070] In other words, the recommendation factors stored in the virtual resource demand knowledge base are those selected and confirmed by user feedback, rather than the target recommendation factors originally recommended by the preset virtual resource recommendation model. This effectively ensures that the relationship between virtual resource demand information and actual recommendation factors in the virtual resource demand knowledge base is updated with user feedback, effectively solving the problem of lag in traditional static knowledge bases.
[0071] In one feasible embodiment, the feedback information is also used to provide feedback on the mapping relationship between the user's actual recommendation intent and the actual recommendation factors, which will not be elaborated further in this application.
[0072] In a preferred embodiment, the latest policy information is obtained, benchmark demand-related information and its corresponding recommendation factors are generated based on the latest policy information, and the virtual resource demand knowledge base is updated based on the benchmark demand-related information and its corresponding recommendation factors.
[0073] In other words, in this embodiment of the application, to further improve the ability of the preset virtual resource recommendation model to understand complex semantics and increase its accuracy, this application further proposes a scheme to incorporate relevant policy information into the virtual resource demand knowledge base. For example, after the "registration system reform" policy is introduced, the benchmark demand information "registration system reform" and the recommendation factor "relaxed listing standards" are correlated to generate a set of target recommendation information, which is then stored in the virtual resource demand knowledge base.
[0074] Therefore, this application can realize the dynamic iteration of the virtual resource demand knowledge base content through the feedback mechanism, so that the target recommendation logic can continuously adapt and optimize with market and policy changes. That is, under the same virtual resource demand information, different target recommendation factors can be obtained based on different market environments and policy changes to meet the needs under different market environments and policy changes.
[0075] Step 203: Obtain the factor parameters corresponding to the target recommendation factor based on the virtual resource demand information.
[0076] In other words, after obtaining the target recommendation factor, the corresponding factor parameters can be further obtained based on the virtual resource demand information. Optionally, a preset virtual resource recommendation model can be used to further analyze the virtual resource demand information to obtain the initial parameter information corresponding to the target recommendation factor. Then, based on the initial parameter information and the parameter range corresponding to the target recommendation factor, the factor parameters corresponding to the target recommendation factor can be determined.
[0077] Step 204: Filter target virtual resource objects based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor.
[0078] It should be noted that the preset virtual resource recommendation model filters target virtual resource objects by calling the factor function corresponding to the target recommendation factor. The factor function corresponding to the target recommendation factor includes structured tool description information of the target recommendation factor. This structured tool description information includes one or more of the following: factor identifier, factor description, factor summary, and usage examples. Specifically, the factor identifier represents the identification information of the recommendation factor; the factor description describes the screening index corresponding to the recommendation factor; the factor summary describes the object information that the recommendation factor focuses on or reflects, as well as parameter configuration characteristics; and the usage examples describe the functional representation of the recommendation factor.
[0079] In some embodiments, to further dynamically adjust the factor function corresponding to the target recommendation factor to match the current market environment and regulatory policies, for example, changing the description of dividend payout ratio in "high dividend yield" from "dividend payout ratio > 50% over the past 3 years" to "dividend payout ratio > 50% over the past 5 years," this application further proposes updating the structured tool description information of the target recommendation factor based on the associated information in the virtual resource demand knowledge base, in order to enhance the interpretability and robustness of the target recommendation factor.
[0080] Specifically, based on the target recommendation factors and the factor parameters corresponding to the target recommendation factors, target virtual resource objects are screened, including: for each target recommendation factor, determining the associated benchmark demand information from the virtual resource demand knowledge base; generating structured tool description information corresponding to the target recommendation factor based on the associated benchmark demand information; and screening target virtual resource objects based on the structured tool description information corresponding to the target recommendation factors.
[0081] It should be noted that the associated benchmark demand information is the virtual resource demand information corresponding to the same recommendation factor. In other words, the associated benchmark demand information and the virtual resource demand information are related through the recommendation factor, and the recommendation intentions may be the same or different.
[0082] In other words, after obtaining the target recommendation factor, multiple related benchmark demand information corresponding to the target recommendation factor is retrieved from the virtual resource demand knowledge base. Then, by analyzing the related benchmark demand information, multiple descriptive information of the target recommendation factor is determined, including but not limited to one or more of factor description, factor introduction, and usage examples. Then, structured tool description information corresponding to the target recommendation factor is generated using this multiple descriptive information. Finally, a target factor function for the target recommendation factor is generated based on the structured tool description information. By calling the target factor function, object filtering based on the target recommendation factor is implemented, thereby obtaining the target virtual resource object.
[0083] In summary, the target recommendation method provided in this application obtains virtual resource demand information input by the user, utilizes a preset virtual resource recommendation model, and obtains target recommendation factors from a virtual resource demand knowledge base based on the virtual resource demand information; obtains factor parameters corresponding to the target recommendation factors based on the virtual resource demand information; and filters target virtual resource objects based on the target recommendation factors and their corresponding factor parameters, thus achieving high-precision recommendation of target objects based on user needs. Simultaneously, the preset virtual resource recommendation model enables the processing of complex and implicit demands, effectively improving the reliability and accuracy of target object recommendations and enhancing user experience.
[0084] It should be noted that although the operation of the method of the present invention is described in a specific order in the accompanying drawings, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed in order to achieve the desired result.
[0085] Figure 4 A schematic diagram of the target recommendation device provided in an embodiment of this application is shown.
[0086] like Figure 4 As shown, the target recommendation device 10 includes: Module 11 is used to obtain the virtual resource requirement information input by the user; Analysis module 12 is used to obtain target recommendation factors from a virtual resource demand knowledge base based on the virtual resource demand information using a preset virtual resource recommendation model; the virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information; The determination module 13 is used to obtain the factor parameters corresponding to the target recommendation factor based on the virtual resource demand information; The filtering module 14 is used to filter target virtual resource objects based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor.
[0087] In some embodiments, the analysis module 12 is specifically used for: Based on the virtual resource demand knowledge base, semantic retrieval is performed on the virtual resource demand information to obtain multiple sets of initial recommendation information. Each set of initial recommendation information includes an initial recommendation benchmark demand information and its corresponding initial recommendation factor and initial recommendation intent. Using the preset virtual resource recommendation model, the correlation between the virtual resource demand information and each set of initial recommendation information is evaluated to obtain the correlation evaluation value corresponding to each set of initial recommendation information; The initial recommendation information is filtered based on the correlation evaluation value to obtain multiple sets of candidate recommendation information; each set of candidate recommendation information includes a candidate recommendation baseline requirement information and its corresponding candidate recommendation factor and candidate recommendation intent; For each type of candidate recommendation intent, multiple sets of candidate recommendation information are deduplicated to obtain target recommendation benchmark demand information corresponding to each candidate recommendation intent, and target recommendation factor corresponding to the target recommendation benchmark demand information.
[0088] In some embodiments, the analysis module 12 is specifically used for: For each type of candidate recommendation intent, the similarity between the baseline requirements of each candidate recommendation is calculated; the similarity includes string repetition rate and / or cosine similarity. The candidate recommendation benchmark requirement information is filtered out if the string repetition rate is greater than or equal to the first similarity threshold and / or the cosine similarity is greater than or equal to the second similarity threshold, to obtain the target recommendation benchmark requirement information; The target recommendation factor is determined based on the target recommendation benchmark requirement information.
[0089] In some embodiments, the filtering module 14 is further configured to: For each of the target recommendation factors, relevant benchmark demand information is determined from the virtual resource demand knowledge base; Based on the associated benchmark demand information, generate structured tool description information corresponding to the target recommendation factor; Based on the structured tool description information and factor parameters corresponding to the target recommendation factor, target virtual resource objects are filtered.
[0090] In some embodiments, the analysis module 12 is further configured to: Track and obtain feedback information corresponding to the virtual resource demand information; the feedback information is used to provide feedback on one or more of the following: the mapping relationship between the virtual resource demand information and the user's actual recommendation intention, the mapping relationship between the virtual resource demand information and the actual recommendation factor, and the mapping relationship between the user's actual recommendation intention and the actual recommendation factor; The virtual resource requirements knowledge base is updated based on the feedback information.
[0091] In some embodiments, the analysis module 12 is further configured to: Obtain the latest policy information, and generate the benchmark demand-related information and its corresponding recommendation factors based on the latest policy information; The virtual resource requirement knowledge base is updated based on the baseline requirement information and its corresponding recommendation factors.
[0092] It should be understood that the modules or modules described in the target recommendation device 10 are related to the reference. Figure 2 The steps in the described method correspond to each other. Therefore, the operations and features described above for the method also apply to the target recommendation device 10 and the modules contained therein, and will not be repeated here. The target recommendation device 10 can be pre-implemented in the browser or other security applications of an electronic device, or it can be loaded into the browser or other security applications of an electronic device by means of downloading. The corresponding modules in the target recommendation device 10 can cooperate with the modules in the electronic device to implement the solution of the embodiments of this application.
[0093] The division of modules or units mentioned in the detailed description above is not mandatory. In fact, according to the embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.
[0094] The following is for reference. Figure 5 , Figure 5 A schematic diagram of the structure of a computer system suitable for implementing the embodiments of this application is shown. like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 502 or programs loaded from storage section 508 into random access memory (RAM) 503. RAM 503 also stores various programs and data required for the system's operating instructions. CPU 501, ROM 502, and RAM 503 are interconnected via bus 504. Input / output (I / O) interface 505 is also connected to bus 504.
[0095] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.
[0096] Specifically, according to embodiments of this application, the flowchart above refers to... Figure 2 The described process can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program contains program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs the functions defined in the system of this application.
[0097] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0098] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operational instructions of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two connected blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified functions or operational instructions, or using a combination of dedicated hardware and computer instructions.
[0099] The units or modules described in the embodiments of this application can be implemented in software or hardware. The described units or modules can also be housed in a processor; for example, a processor can be described as including an acquisition module, an analysis module, a determination module, and a filtering module. The names of these units or modules do not necessarily limit the specific unit or module itself; for example, an acquisition module can also be described as "acquiring virtual resource requirement information input by the user."
[0100] In another aspect, this application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments, or may exist independently and not assembled into the electronic device. The aforementioned computer-readable storage medium stores one or more programs that, when used by one or more processors, execute the target recommendation method described in this application.
[0101] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.
Claims
1. A target recommendation method, characterized in that, include: Obtain the user's input information regarding virtual resource requirements; Using a pre-defined virtual resource recommendation model, target recommendation factors are obtained from the virtual resource demand knowledge base based on the virtual resource demand information. The virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information; Obtain the factor parameters corresponding to the target recommendation factor based on virtual resource demand information; Based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor, target virtual resource objects are selected.
2. The target recommendation method according to claim 1, characterized in that, The step of using a preset virtual resource recommendation model to obtain target recommendation factors from a virtual resource demand knowledge base based on the virtual resource demand information includes: Based on the virtual resource demand knowledge base, semantic retrieval is performed on the virtual resource demand information to obtain multiple sets of initial recommendation information. Each set of initial recommendation information includes an initial recommendation benchmark demand information and its corresponding initial recommendation factor and initial recommendation intent. Using the preset virtual resource recommendation model, the correlation between the virtual resource demand information and each set of initial recommendation information is evaluated to obtain the correlation evaluation value corresponding to each set of initial recommendation information; The initial recommendation information is filtered based on the correlation evaluation value to obtain multiple sets of candidate recommendation information; each set of candidate recommendation information includes a candidate recommendation baseline requirement information and its corresponding candidate recommendation factor and candidate recommendation intent; For each type of candidate recommendation intent, multiple sets of candidate recommendation information are deduplicated to obtain target recommendation benchmark demand information corresponding to each candidate recommendation intent, and target recommendation factor corresponding to the target recommendation benchmark demand information.
3. The target recommendation method according to claim 2, characterized in that, For each type of candidate recommendation intent, the process of deduplicating multiple sets of candidate recommendation information to obtain target recommendation benchmark demand information corresponding to each candidate recommendation intent, and target recommendation factors corresponding to the target recommendation benchmark demand information, includes: For each type of candidate recommendation intent, the similarity between the baseline requirements of each candidate recommendation is calculated; the similarity includes string repetition rate and / or cosine similarity. The candidate recommendation benchmark requirement information is filtered out if the string repetition rate is greater than or equal to the first similarity threshold and / or the cosine similarity is greater than or equal to the second similarity threshold, to obtain the target recommendation benchmark requirement information; The target recommendation factor is determined based on the target recommendation benchmark requirement information.
4. The target recommendation method according to claim 1, characterized in that, The step of filtering target virtual resource objects based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor includes: For each of the target recommendation factors, relevant benchmark demand information is determined from the virtual resource demand knowledge base; Based on the associated benchmark demand information, generate structured tool description information corresponding to the target recommendation factor; Based on the structured tool description information and factor parameters corresponding to the target recommendation factor, target virtual resource objects are filtered.
5. The target recommendation method according to claim 1, characterized in that, Also includes: Track and obtain feedback information corresponding to the virtual resource demand information; The feedback information is used to provide feedback on one or more of the following: the mapping relationship between the virtual resource demand information and the user's actual recommendation intention; the mapping relationship between the virtual resource demand information and the actual recommendation factor; and the mapping relationship between the user's actual recommendation intention and the actual recommendation factor. The virtual resource requirements knowledge base is updated based on the feedback information.
6. The target recommendation method according to claim 2, characterized in that, Also includes: Obtain the latest policy information, and generate the benchmark demand-related information and its corresponding recommendation factors based on the latest policy information; The virtual resource requirement knowledge base is updated based on the baseline requirement information and its corresponding recommendation factors.
7. A target recommendation device, characterized in that, include: The acquisition module is used to acquire the virtual resource requirement information input by the user; The analysis module is used to obtain target recommendation factors from the virtual resource demand knowledge base based on the virtual resource demand information using a preset virtual resource recommendation model; the virtual resource demand knowledge base includes multiple benchmark demand information and virtual resource recommendation factors corresponding to the benchmark demand information. The determination module is used to obtain the factor parameters corresponding to the target recommendation factor based on the virtual resource demand information; The filtering module is used to filter target virtual resource objects based on the target recommendation factor and the factor parameters corresponding to the target recommendation factor.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the target recommendation method as described in any one of claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the target recommendation method as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the target recommendation method as described in any one of claims 1-6.