A customer information matching method, system, medium and device

Abstract

The application provides a customer information matching method, system, medium and device, relates to the technical field of data processing, and comprises the following steps: obtaining original information text and customer data and a holding position detail of a target customer; performing structured processing on the information to obtain information labels; generating a feature portrait according to the customer data and calculating a holding position exposure portrait according to the holding position detail; calculating customer information matching parameters based on three types of structured objects; and only when the parameters exceed a threshold value, calling a generated constraint template to generate a personalized briefing candidate text containing information matched with the customer. The application integrates customer holding position exposure, features and information content into a unified calculation link, stably identifies the real correlation between information and customer combination structures, improves matching accuracy, reduces the proportion of irrelevant content generation, and guarantees the directional explanation ability of the generated content to the customer holding position structure by means of the constraint template.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a customer information matching method, system, medium, device and program product. Background Technology

[0002] With the development of digital services in financial institutions, personalized information recommendation and content generation have become common capabilities in wealth management customer service. The closest existing technology discloses a personalized financial information generation scheme: receiving and analyzing news in real time, storing users' demographic information, financial portfolios, interests and preferences, using generative artificial intelligence to process news and user information, generating personalized information content of different lengths and levels of detail, and sending it to users.

[0003] However, this type of solution has the following drawbacks: First, the utilization of customer holdings is limited to directly reading the holdings list or simple keyword matching, and it cannot calculate the holdings exposure profile from holdings such as funds and stocks; Second, it does not incorporate information structured tags, customer feature profiles, and holdings exposure profiles into the same calculation chain, making it difficult to reliably identify the true degree of correlation between information and the customer's current portfolio structure, resulting in low recommendation relevance, high repetition, and a lot of noise; Third, the generated content lacks constraint template control, making it difficult to balance targeted explanations of holdings structure with consistency in expression style.

[0004] Therefore, how to improve the matching accuracy between information and the client's current investment portfolio and ensure the stability of the generated content is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a customer information matching method, system, computer-readable storage medium, and electronic device that can improve the matching accuracy of information with the customer's current investment portfolio and reduce the proportion of irrelevant content generated.

[0006] To address the aforementioned technical problems, this application provides a customer information matching method, the specific technical solution of which is as follows: Obtain raw information text and target customer data and holding details; The original information text is processed into a structured format to obtain structured information tags; Generate customer profiles based on the customer data; Calculate the target customer's holdings exposure profile based on the aforementioned holdings details; Based on the information structure tags, the customer feature profile, and the position exposure profile, calculate the customer information matching parameters between the target customer and the original information text; When the customer information matching parameters exceed the set threshold, a constraint template is invoked to generate personalized briefing candidate text; the personalized briefing candidate text contains information matching the target customer.

[0007] Optionally, generating a customer profile based on the customer data includes: A customer profile is generated based on the target customer's basic attribute data, holding data, historical interaction data, browsing or followed topics data, and risk level data. The customer profile includes at least one of the following: customer risk level, historical followed topics, interaction activity level, historical content types received, acceptable copywriting complexity, and service stage.

[0008] Optionally, the original information text is subjected to structured processing to extract information structure tags, including: The original information text is subjected to deduplication, segmentation, entity recognition, and coarse topic classification to obtain preprocessed text; Extract the information structure tags corresponding to the preprocessed text; the information structure tags include at least one of the following: information topic tags, target tags, industry tags, market area tags, sentiment polarity tags, timeliness tags, risk warning tags, and applicable customer type tags.

[0009] Optionally, calculating the target customer's position exposure profile based on the position details includes: Read the target customer's holding details; the holding details include the holdings and their weights; Each holding is mapped to a preset exposure factor space to obtain the exposure value of each holding on each exposure factor; The exposure values ​​of each holding are weighted and aggregated according to the aforementioned holding weights to obtain the holding exposure profile; the holding exposure profile includes at least one of the following exposure dimensions: equity and fixed income exposure, growth and value exposure, large-cap and small-cap exposure, industry and theme exposure, market and region exposure, volatility tolerance characteristics, and sensitivity to hot themes.

[0010] Optionally, based on the structured information tags, the customer feature profile, and the position exposure profile, the calculation of customer information matching parameters between the target customer and the original information text includes: Calculate the holding correlation parameters based on the degree of correlation between the information structure tags and the holding exposure profile; The topic interest parameters are calculated based on the degree of fit between the information structure tags and the customer feature profile; Calculate the risk adaptation parameters based on the degree of matching between the customer risk level in the customer feature profile and the risk warning tags in the information structure tags; Calculate timing parameters based on the degree of match between the current contact timing and the customer service rhythm; Calculate the duplicate sending penalty parameters based on historical duplicate sending data; The customer information matching parameters are obtained by weighting and summing the position association parameters, the topic interest parameters, the risk adaptation parameters, and the timing parameters, and then subtracting the product of the duplication penalty parameter and the preset weight.

[0011] Optionally, before calling the constraint template to generate personalized briefing candidate text, the following steps are also included: At least one correlation point is determined based on the customer information matching parameters; the correlation point is used to characterize the correspondence between the information structured tags and the holding exposure profile; Determine expression style constraints based on employee expression profiles; Based on the aforementioned points of association and the aforementioned style of expression constraints, the generated constraint template is constructed; the generated constraint template includes at least one of the following constraints: points of association that must be mentioned, prohibited expressions, suggested length range, opening and closing styles, risk warning expression format, and copywriting detail level.

[0012] Optionally, calling the constraint template to generate personalized briefing candidate text includes: Generate a reason for sending based on the opening description in the generated constraint template; Generate related explanatory sentences based on the reference points in the generated constraint template; the related explanatory sentences are used to explain the relationship between the original information text and the position exposure profile; Generate a service closing sentence based on the closing remarks in the generated constraint template; The personalized briefing candidate text is obtained by combining the reason for sending, the related explanation, and the service closing sentence.

[0013] This application also provides a customer information matching system, including: The data acquisition module is used to acquire raw information text and customer data and holding details of target customers; The information structuring module is used to perform structuring processing on the original information text and extract information structuring tags; The customer data integration module is used to generate customer feature profiles based on the customer data. The position exposure calculation module is used to calculate the position exposure profile of the target customer based on the position details; The customer information matching parameter module is used to calculate the customer information matching parameters between the target customer and the original information text based on the information structured tags, the customer feature profile, and the holding exposure profile. The personalized briefing generation module is used to generate personalized briefing candidate text by calling the generation constraint template when the customer information matching parameters exceed a set threshold; the personalized briefing candidate text contains information matching the target customer.

[0014] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0015] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described above when it invokes the computer program in the memory.

[0016] This application provides a customer information matching method, comprising: acquiring original information text and customer data and position details of a target customer; performing structured processing on the original information text to obtain information structured tags; generating a customer feature profile based on the customer data; calculating the position exposure profile of the target customer based on the position details; calculating customer information matching parameters between the target customer and the original information text based on the information structured tags, the customer feature profile, and the position exposure profile; when the customer information matching parameters exceed a set threshold, calling a constraint template to generate personalized briefing candidate text; the personalized briefing candidate text contains information information matching the target customer.

[0017] This application obtains structured information tags by structuring the original information text, generates customer feature profiles based on customer data, calculates position exposure profiles based on position details, and then uniformly calculates customer information matching parameters based on the three types of structured objects: information structured tags, customer feature profiles, and position exposure profiles. Only when the customer information matching parameters exceed a set threshold are generation constraint templates invoked to generate personalized briefing candidate texts. This integrates customer position exposure, customer features, and information content into the same calculation chain for constraint, which can reliably identify the true correlation between information and the customer's current portfolio structure, improve the matching accuracy between information and the customer's current investment portfolio, reduce the proportion of irrelevant content generation, and ensure the ability of the generated content to provide targeted explanations of the customer's position structure through generation constraint templates.

[0018] This application also provides a customer information matching system, a computer-readable storage medium, and an electronic device, which have the aforementioned beneficial effects, and will not be elaborated here. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 A flowchart illustrating a customer information matching method provided in this application embodiment; Figure 2 This is a schematic diagram of a customer information matching system provided in an embodiment of this application; Figure 3 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] Before introducing specific embodiments, the technical terms involved in the embodiments of this application will be uniformly explained so that those skilled in the art can understand them.

[0023] Structured information tags refer to a set of computable tags obtained after structuring raw information text. These tags are used to express the topic, attributes, risk characteristics, and scope of application of the information. Structured information tags are used to convert unstructured information text into structured intermediate objects that can be used for matching calculations and generation control.

[0024] A customer profile is a set of customer characteristics built based on their basic attributes, historical behavior, interaction records, and service information. It is used to express a customer's service preferences, behavioral characteristics, and information suitability. Customer profiles primarily describe what types of information and content presentation methods are most suitable for a customer.

[0025] A portfolio exposure profile is a set of asset structure exposure characteristics calculated based on a client's portfolio details. It is not the original portfolio list itself, but rather a structured result formed by mapping portfolio holdings to multiple computable exposure dimensions. The portfolio exposure profile reflects the correlation between a client's current asset allocation structure and market themes.

[0026] Customer information matching parameters refer to the correlation score calculated based on information structured tags, customer feature profiles, and portfolio exposure profiles, used to measure the degree of fit between a piece of information and the target customer. In one feasible implementation, subsequent generation processing can be performed only on information whose customer information matching parameters exceed a set threshold.

[0027] An employee communication profile is a set of communication style features built based on an employee's historical text messages, communication records, or configured templates. These features are used to constrain the expression style of generated content. Employee communication profiles aim to improve the consistency between generated content and the employee's daily communication style.

[0028] Personalized briefings refer to candidate text content generated for target customers based on structured information tags, customer profiles, portfolio exposure profiles, and employee communication profiles. Personalized briefings are used to provide customers with informational explanations related to their current asset structure and service characteristics.

[0029] Compliance verification refers to performing rule-based, model-based, or combined verification on generated content to identify any content that does not comply with business rules, risk rules, or regulatory requirements. In one feasible implementation, when candidate text fails compliance verification, a constraint adjustment and regeneration process can be automatically triggered.

[0030] Feedback flow refers to the process of recording user behavior, employee behavior, and interaction results after content is sent and using this information for subsequent model or parameter updates. Feedback flow is used to update the calculation weights of customer information matching parameters, employee profile parameters, generation constraint strategies, or related model parameters to form a continuous optimization loop.

[0031] See Figure 1 , Figure 1 A flowchart of a customer information matching method provided in this application embodiment, the method including: Step S101: Obtain the original information text and the target customer's customer data and holding details.

[0032] This step involves accessing raw news text from information sources. These sources can include research morning reports, market news flashes, news articles, industry analysis summaries, company announcement summaries, etc. The raw news text can be full text, summary, news flash, or a collection of multiple news articles. Customer data can include basic attribute data of target customers, holding data, historical interaction data, browsing or followed topics data, and risk level data; holding details can include the holdings and their weights, such as the holding's code, name, quantity, market value, and cost price. After the raw news text enters the information processing chain, it can first undergo deduplication, segmentation, entity recognition, and coarse topic classification for subsequent structured processing. Thus, scattered unstructured information inputs and multi-dimensional customer-side data can be aggregated into the same processing chain, providing a complete data foundation for subsequent unified calculations.

[0033] Step S102: Perform structuring processing on the original information text to obtain information structuring tags.

[0034] This step performs structured extraction on the original news text to obtain structured tags. This step can be implemented using rule engines, classification models, vector retrieval models, or large-scale model extraction; those skilled in the art can choose one or more combinations based on implementation conditions. For example, for a market news article, the following can be extracted: the theme is technology growth, the sub-theme is artificial intelligence industry applications, the related industries are semiconductors, software, and computing power, the sentiment is neutral to positive, the timeliness is valid for half a day before the market opens, the risk warning is high short-term volatility, and the applicable customer type is growth-oriented customers. Thus, unstructured news text can be converted into computable, matchable, and controllable structured intermediate objects, allowing the news content to participate in subsequent matching calculations in the form of tags.

[0035] Step S103: Generate customer feature profiles based on customer data.

[0036] This step allows us to obtain the target customer's basic and behavioral attributes, forming a customer profile. This profile can include the customer's risk level, historical topics of interest, recent interaction activity, types of content received historically, acceptable text complexity, and service stage. The customer profile indicates what types of information the customer is best suited to receive. Therefore, we can characterize the customer's information fit from behavioral and service perspectives, providing a behavioral basis for subsequent calculations of the fit between information and the customer.

[0037] In one feasible implementation, when generating customer profiles based on customer data, customer profiles can be generated based on the target customer's basic attribute data, holding data, historical interaction data, browsing or followed topics data, and risk level data. The customer profile can include at least one of the following: customer risk level, historical followed topics, interaction activity level, historical content types received, acceptable text complexity, and service stage. Basic attribute data can include gender, age, province of residence, asset level, occupation, etc.; historical interaction data can include interaction channels, interaction time, interaction content categories, and interaction action types, etc. Therefore, customer profiles can characterize customers from multiple dimensions such as attributes, behavior, preferences, risk, and service progress, providing fine-grained input for subsequent topic interest matching and risk matching calculations.

[0038] Step S104: Calculate the target customer's holdings exposure profile based on the holdings details.

[0039] This step retrieves the client's holdings details and maps the holdings to several exposure factors. For example, Fund A can be mapped to a technology growth exposure of 0.7, a ChiNext exposure of 0.6, and a high volatility level; ETF B can be mapped to a large-cap growth exposure of 0.5 and a Shenzhen core asset exposure of 0.8; and Hong Kong stock fund C can be mapped to a Hong Kong stock growth exposure of 0.9. Then, all of the client's holdings can be aggregated according to their weights to obtain a holding exposure profile. The holding exposure profile differs from the client characteristic profile: the client characteristic profile focuses more on behavioral and service characteristics, while the holding exposure profile focuses more on asset structure and market risk exposure. Therefore, the original holdings list can be upgraded to calculable asset structure exposure characteristics, making the correlation between information and the client's current portfolio structure quantifiable.

[0040] In one feasible implementation, calculating a target customer's holdings exposure profile based on holdings details may include the following steps: Step 1: Read the target client's holdings details, including the holdings and their weightings; Step 2: Map each holding to a preset exposure factor space to obtain the exposure value of each holding on each exposure factor; Step 3: Weight and aggregate the exposure values ​​of each holding according to the holding weight to obtain the holding exposure profile.

[0041] A portfolio exposure profile can include at least one of the following exposure dimensions: equity and fixed income exposure, growth and value exposure, large-cap and small-cap exposure, industry and theme exposure, market and regional exposure, volatility tolerance characteristics, and sensitivity to hot topics. Market and regional exposure can include A-share exposure, Hong Kong stock exposure, and overseas exposure. The preset exposure factor space can be set by those skilled in the art based on implementation experience; for example, the types and number of exposure factors can be set by referring to industry-standard factor classification systems. The exposure values ​​of each holding on each exposure factor can be obtained based on manually maintained fund or stock factor mapping tables, third-party labeled data, aggregated after portfolio penetration, or deduced from historical return and volatility characteristics. Therefore, a portfolio exposure profile can quantitatively characterize a client's current asset allocation structure from multiple dimensions such as asset class, style, size, industry, region, volatility, and sensitivity to hot topics, making the correlation calculation between information and holdings more precise.

[0042] Step S105: Calculate the customer information matching parameters between the target customer and the original information text based on the information structured tags, customer feature profiles, and holdings exposure profiles.

[0043] This step calculates customer information matching parameters by integrating the following information: the correlation between structured information tags and position exposure profiles; the fit between structured information tags and customer characteristic profiles; the current outreach timing and customer service rhythm; historical duplicate sending status; and the timeliness and freshness of the information. In the calculation, the position association parameter, topic interest parameter, risk suitability parameter, and timing parameter are multiplied by their respective weights and summed. Then, the product of the duplicate penalty parameter and its weight is subtracted to obtain the customer information matching parameters. Each weight can be a preset weight or a dynamic weight learned dynamically from historical feedback data. Therefore, a single, comparable parameter value can uniformly express the comprehensive fit between information and the customer across multiple dimensions, including position association, topic interest, risk suitability, outreach timing, and duplicate suppression, providing a quantitative criterion for whether to proceed to the generation stage.

[0044] Step S106: When the customer information matching parameters exceed the set threshold, the generation constraint template is invoked to generate personalized briefing candidate text; the personalized briefing candidate text contains information matching the target customer.

[0045] The generation stage only begins when the customer information matching parameters exceed a set threshold. This set threshold can be set by those skilled in the art based on implementation experience, for example, it can be set to 0.6; it can also be set based on historical feedback data, for example, the lower quartile of the customer information matching parameters corresponding to historically adopted candidate texts can be used as the set threshold; it can also be set separately according to customer groups, which is not limited in this application.

[0046] When generating personalized briefing candidate texts, one or more personalized briefing candidate texts can be generated by calling a generative model or a summarizing model based on the generation constraint template. Alternatively, template filling plus local generation, search-enhanced generation, or generating explanatory points first and then generating the text can be performed. Each candidate text can include at least three parts: a reason for sending, explaining why the information is being sent to this customer; a related explanatory sentence, explaining the relationship between the information and the customer's current portfolio exposure profile; and a service closing sentence, providing a light service-oriented conclusion. In one feasible implementation, short, medium, and long versions of candidate texts can be output simultaneously, as well as text with a title, text with follow-up suggestions, or text for different channels such as WeChat, SMS, and in-app messages. This allows generation resources to be focused on information truly relevant to the customer, reducing the generation of irrelevant content, and controlling the structure and expression of the generated content through constraint templates, ensuring that the generated personalized briefing candidate texts have both targeted explanatory capabilities for portfolio structure and stable sendability.

[0047] In one feasible implementation, this step can generate a reason for sending based on the opening description in the constraint template; generate a related explanation sentence based on the related points mentioned in the constraint template, which is used to explain the relationship between the information and the position exposure profile; generate a service closing sentence based on the closing remarks in the constraint template; and finally combine the reason for sending, the related explanation sentence, and the service closing sentence to obtain a personalized briefing candidate text.

[0048] For example, the reason for sending the message could be, "Based on your current holdings, I've extracted three key points"; the related explanation could focus on "why it's relevant to the client's technology growth positions"; and the closing statement could be, "I'm sending this for your reference first," or "I can review it more thoroughly if needed." This generates a three-part structure: explanation, related explanation, and closing statement. This ensures the briefing explains both why it's being sent and its relationship to the client's holdings, enhancing the explainability and service feel of the personalized briefing.

[0049] The customer information matching method provided in this embodiment incorporates information structured tags, customer feature profiles, and portfolio exposure profiles into the same computational chain to uniformly calculate customer information matching parameters. It uses a set threshold as a generation trigger condition and a generation constraint template to control the generation process, thereby stably identifying the true correlation between information and the customer's current portfolio structure, improving the matching accuracy between information and the customer's current investment portfolio, reducing the proportion of irrelevant content generation, and ensuring the targeted explanatory ability and expression stability of the generated content for the customer's portfolio structure.

[0050] Based on the above embodiments, in one feasible implementation, structuring the original information text and extracting information structure tags may include the following steps: Step 1: Perform deduplication, segmentation, entity recognition, and coarse topic classification on the original information text to obtain preprocessed text; Step 2: Extract the information structure tags corresponding to the preprocessed text; information structure tags may include at least one of the following: information topic tags, target tags, industry tags, market area tags, sentiment polarity tags, timeliness tags, risk warning tags, and applicable customer type tags.

[0051] Deduplication removes duplicate and homogeneous information, segmentation divides long articles into semantically complete paragraphs, entity recognition identifies targets, organizations, and individuals within the information, and coarse topic classification labels the information with primary topics. In practical implementation, information structuring can employ rule extraction combined with a classification model, vector retrieval combined with label mapping, direct extraction using a large model, or a multi-model voting approach to output structured information labels. This embodiment improves the accuracy and stability of structured information labels and reduces the interference of duplicate and noisy information on subsequent matching calculations.

[0052] Based on the above embodiments, in one feasible implementation, step S105 may specifically include the following steps: Step 1: Calculate the holding correlation parameters based on the degree of correlation between information structure tags and holding exposure profiles; Step 2: Calculate topic interest parameters based on the degree of fit between information structure tags and customer feature profiles; Step 3: Calculate the risk matching parameters based on the degree of matching between the customer risk level in the customer feature profile and the risk warning tags in the information structure tags; Step 4: Calculate timing parameters based on the match between the current outreach timing and the customer service rhythm; calculate duplicate penalty parameters based on historical duplicate sending data; Step 5: Calculate the weighted sum of the position association parameters, topic interest parameters, risk adaptation parameters, and timing parameters, and then subtract the product of the duplicate penalty parameters and the preset weights to obtain the client information matching parameters.

[0053] Each preset weight can be set by those skilled in the art based on implementation experience. For example, the weight of the position association parameter can be set to 0.4, the weight of the topic interest parameter can be set to 0.2, the weight of the risk adaptation parameter can be set to 0.2, the weight of the timing parameter can be set to 0.1, and the preset weight of the repeat penalty parameter can be set to 0.3. Alternatively, it can be obtained through machine learning training based on historical feedback data. This application does not limit the specific values ​​of each weight. When calculating the position association parameter, the degree of overlap or similarity between the target tags and industry tags involved in the information structured tags and the industry theme exposure and hot topic sensitivity in the customer's position exposure profile can be calculated. When calculating the timing parameter, it can be determined by combining the timeliness tag of the information with the customer's service stage. When calculating the repeat penalty parameter, the number of times the same topic information is sent to the customer within a set time window can be counted. The more times, the larger the repeat penalty parameter. In addition to the above weighting methods, the calculation of customer information matching parameters can also be performed using machine learning ranking models, vector similarity plus rule reordering, graph association path scoring, etc. As can be seen, the customer information matching parameters can explicitly integrate five types of signals: holding correlation, topic interest, risk suitability, timing of reach, and repetition suppression. This allows the scoring results to reflect the true correlation between the information and the customer's holdings, while also suppressing repetitive interference, thereby improving the accuracy of matching and the customer experience.

[0054] Based on the above embodiments, as a preferred embodiment, before calling the constraint template to generate personalized briefing candidate text, the constraint template can be constructed first, which may specifically include the following steps: First, at least one correlation point can be determined based on the customer information matching parameters. The correlation point is used to characterize the correspondence between the structured tags of the information and the holding exposure profile. For example, when the information topic is technology growth and the exposure value of the technology growth direction in the customer's holding exposure profile is high, "this information is related to the customer's technology growth position" can be determined as a correlation point. The number of correlation points can be one or more. The top few with the greatest matching contribution can be selected as correlation points. The specific number can be set by those skilled in the art based on implementation experience, for example, it can be set to 1 to 3.

[0055] Then, expression style constraints can be determined based on employee expression profiles. Employee expression profiles can be constructed based on the financial manager's historical text messages, frequently used phrases, sentence structure characteristics, length preferences, and professional level preferences. In terms of construction methods, it can be based on statistical analysis of historical text messages, manually configured by employees using commonly used style templates, automatically extracted from WeChat chat records, or a combination of common and personal templates. For example, employee expression profiles can include characteristics such as style tendencies, tone preferences, professional fields, word preferences, and compliance levels; tone can be one of the following: steady and professional, approachable and natural, or concise and direct.

[0056] Next, a constraint template can be constructed based on the relevant points and expression style constraints. The constraint template can include at least one constraint from the following: required relevant points, prohibited expressions, suggested length range, opening and closing styles, risk warning expression format, and text detail levels. For example, the opening statement could be, "Based on your current holdings, I have extracted three key points"; the required relevant points could be, "Why is this relevant to the client's technology growth position?"; prohibited expressions could be, "Immediate deployment," "Guaranteed rise," etc.; the suggested length range could be 120 to 180 characters, which can be set by those skilled in the art according to the reading habits of the publishing channel, for example, 120 to 180 characters for WeChat and less than 70 characters for SMS; the closing could be, "Sent for your reference first," or "I will review it further if needed."

[0057] This embodiment can separate and structure "what to generate" and "how to say it": the correlation points ensure that the generated content revolves around the real relationship between information and customer holdings, the employee expression profile ensures that the generated content is consistent with the daily expression style of the financial manager, and the constraint template solidifies the two into structured instructions that the generation model can directly follow, thereby maintaining the consistency of the employee's tone while personalizing the customer, and reducing situations like mass-sent templates or machine generation.

[0058] Based on the above embodiments, as a preferred execution method, this embodiment provides an implementation method for performing compliance verification after calling the generation constraint template to generate personalized briefing candidate text.

[0059] Specifically, compliance checks can be performed on personalized briefing candidate texts. When a personalized briefing candidate text fails the compliance check, a constraint template is reconstructed and a new personalized briefing candidate text is generated. In exemplary application scenarios, compliance checks can employ keyword blacklist / whitelist rules, classification models to determine risk levels, large-scale model review, or a combination of rule and model verification. Upon verification failure, the system can automatically revert to the constraint template construction step or the briefing generation step, rebuild the constraint template, and regenerate it. For example, it can tighten prohibited output statements or lower the level of detail in the text before regenerating. This allows for the interception of non-compliant content before sending and the continuous correction of constraints through an automatic rollback and regeneration mechanism, improving the sendability of the generated content.

[0060] In one feasible implementation, performing compliance verification on personalized briefing candidate texts may include: verifying whether the personalized briefing candidate texts contain statements that induce transactions, whether they do not match the customer's risk level, whether they deviate from the employee's expression profile by more than a set value, whether they contain prohibited sensitive terms, and whether they fail to cover at least one of the relevant points that must be mentioned; when the personalized briefing candidate texts pass the compliance verification, the personalized briefing candidate texts are output to the financial manager's workbench.

[0061] It is important to note that the deviation threshold between the candidate text and the employee's descriptive profile can be set by those skilled in the art based on implementation experience. For example, the style similarity threshold between the candidate text and the employee's descriptive profile can be set to 0.8; anything below 0.8 is considered a deviation exceeding the set value. Alternatively, it can be set based on the style deviation distribution of historically adopted texts. Therefore, candidate texts can be comprehensively validated from five dimensions: inducing transactions, risk matching, style consistency, sensitive language, and coverage of relevant points. This ensures that the candidate texts output to the financial manager's workbench meet the sending requirements in terms of compliance, suitability, and tone consistency.

[0062] After personalized briefings are sent, feedback data can be recorded. This data can include at least one of the following: whether the personalized briefing candidate text was adopted, whether it was sent, whether the customer clicked on it, whether the customer replied, the customer's response sentiment and subsequent actions, and whether a follow-up service event was formed. Based on the feedback data, at least one of the following can be updated: the calculation weight of customer information matching parameters, employee expression profile, and matching model parameters. In terms of recording granularity, binary feedback of sending or not sending can be recorded, as well as fine-grained feedback such as clicks, replies, and dwell time. In terms of update scope, only the calculation weight of customer information matching parameters can be updated, or the employee expression profile and generation constraint strategy can be updated simultaneously. Thus, the sending results and customer feedback can be fed back to the matching calculation and generation constraint stages, allowing various weights and profile parameters to be continuously adjusted based on actual service effects. This forms a sustainable optimization closed loop that connects matching, generation, compliance, and feedback, continuously improving the relevance of subsequent information matching and the stability of generated content.

[0063] The following section provides a further description of the relevant object data and its function in the customer information matching method provided in this application: The original information text is the input object, and its fields can include information identifier, source, publication time, title, and body content.

[0064] Information structured tags are structured intermediate objects extracted from raw information text. Their fields can include information identifier, industry, topic, event type, sentiment polarity, impact level, and tag list. Information structured tags are generated from the raw information text and flow downstream.

[0065] Customer profile and portfolio exposure profile are two distinct customer-side intermediate objects: Customer profile fields may include customer identifier, risk level, investment preference, product preference, lifecycle stage, interaction summary, and recent active time. It relies on the input of customer base data and interaction data. Customer base data fields may include customer identifier, gender, age, province, asset level, and occupation. Interaction data fields may include interaction identifier, customer identifier, channel, interaction time, content category, and action type. Portfolio exposure profile fields may include customer identifier, industry exposure, target exposure, portfolio weight, exposure level, and concentration. It relies on the input of customer portfolio details. Customer portfolio details fields may include portfolio identifier, customer identifier, target code, target name, portfolio quantity, market value, and cost price.

[0066] The customer information matching parameter is an intermediate score object calculated jointly by the structured results from the information side and the customer side. Its fields may include customer identifier, information identifier, matching score, score level, matching reason, and calculation time.

[0067] Employee expression profiles are not directly involved in scoring, but are entered into the constraint template along with customer information matching parameters; the fields of employee expression profiles may include employee identifier, style tendency, tone preference, professional field, word preference, and compliance level; The fields that can be used to generate constraint templates may include customer identifier, information identifier, employee identifier, tone style, content focus, prohibition rules, required mentions, and template version.

[0068] Personalized briefing candidate texts are the final sendable content candidates, and their fields may include candidate identifier, customer identifier, information identifier, employee identifier, body content, generation time, and version. Compliance verification results may include candidate identifier, compliance status, risk level, risk reason, and verification time.

[0069] Fields for sending results or customer feedback can include record identifier, candidate identifier, sending status, sending time, open status, feedback type, feedback content, and feedback time. Sending results or customer feedback can be used to update ratings and profiles, and interactive data can be imported in a correlated reference manner.

[0070] As can be seen, the data objects in this application can form a complete flow link from raw data to intermediate features, to calculation results, to generated content, and then to verification and feedback, providing structured data support for closed-loop optimization.

[0071] See Figure 2 , Figure 2 This application provides a schematic diagram of a customer information matching system structure, which includes: The data acquisition module is used to acquire raw information text and customer data and holding details of target customers; The information structuring module is used to perform structuring processing on the original information text and extract information structuring tags; The customer data integration module is used to generate customer feature profiles based on the customer data. The position exposure calculation module is used to calculate the position exposure profile of the target customer based on the position details; The customer information matching parameter module is used to calculate the customer information matching parameters between the target customer and the original information text based on the information structured tags, the customer feature profile, and the holding exposure profile. The personalized briefing generation module is used to generate personalized briefing candidate text by calling the generation constraint template when the customer information matching parameters exceed a set threshold; the personalized briefing candidate text contains information matching the target customer.

[0072] This application also provides an embodiment of a computer-readable storage medium and a computer program product. Both the computer-readable storage medium and the computer program product may store a computer program that, when executed by a processor, implements the steps of the method described in the above method embodiments.

[0073] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] The computer-readable storage medium provided in this embodiment includes computer programs that implement the corresponding embodiments of the above methods, and the effects are the same as above.

[0075] This application also provides an electronic device, see [link to document]. Figure 3 The present application provides a structural diagram of an electronic device, such as... Figure 3 As shown, it may include a processor 1410 and a memory 1420.

[0076] The processor 1410 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 1410 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1410 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 1410 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 1410 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0077] The memory 1420 may include one or more computer-readable storage media, which may be non-transitory. The memory 1420 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 1420 is used to store at least the following computer program 1421, which, after being loaded and executed by the processor 1410, is capable of implementing the relevant steps in the methods executed by the electronic device side as disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 1420 may also include an operating system 1422 and data 1423, etc., and the storage method may be temporary storage or permanent storage. The operating system 1422 may include Windows, Linux, Android, etc.

[0078] In some embodiments, the electronic device may further include a display screen 1430, an input / output interface 1440, a communication interface 1450, a sensor 1460, a power supply 1470, and a communication bus 1480.

[0079] certainly, Figure 3 The structure of the electronic device shown does not constitute a limitation on the electronic device in the embodiments of this application. In practical applications, the electronic device may include more than [other components]. Figure 3 More or fewer components as shown, or combinations of certain components.

[0080] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. As the system provided in the embodiments corresponds to the method provided in the embodiments, the description is relatively simple; relevant parts can be found in the method section.

[0081] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

[0082] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A customer information matching method, characterized in that, include: Obtain raw information text and target customer data and holding details; The original information text is processed into a structured format to obtain structured information tags; Generate customer profiles based on the customer data; Calculate the target customer's holdings exposure profile based on the aforementioned holdings details; Based on the information structure tags, the customer feature profile, and the position exposure profile, calculate the customer information matching parameters between the target customer and the original information text; When the customer information matching parameters exceed the set threshold, a constraint template is invoked to generate personalized briefing candidate text; the personalized briefing candidate text contains information matching the target customer.

2. The method according to claim 1, characterized in that, Generating a customer profile based on the customer data includes: A customer profile is generated based on the target customer's basic attribute data, holding data, historical interaction data, browsing or followed topics data, and risk level data. The customer profile includes at least one of the following: customer risk level, historical followed topics, interaction activity level, historical content types received, acceptable copywriting complexity, and service stage.

3. The method according to claim 1, characterized in that, The original information text is subjected to structuring processing to extract information structure tags, including: The original information text is subjected to deduplication, segmentation, entity recognition, and coarse topic classification to obtain preprocessed text; Extract the information structure tags corresponding to the preprocessed text; the information structure tags include at least one of the following: information topic tags, target tags, industry tags, market area tags, sentiment polarity tags, timeliness tags, risk warning tags, and applicable customer type tags.

4. The method according to claim 1, characterized in that, The target customer's portfolio exposure profile, calculated based on the aforementioned portfolio details, includes: Read the target customer's holding details; the holding details include the holdings and their weights; Each holding is mapped to a preset exposure factor space to obtain the exposure value of each holding on each exposure factor; The exposure values ​​of each holding are weighted and aggregated according to the aforementioned holding weights to obtain the holding exposure profile; the holding exposure profile includes at least one of the following exposure dimensions: equity and fixed income exposure, growth and value exposure, large-cap and small-cap exposure, industry and theme exposure, market and region exposure, volatility tolerance characteristics, and sensitivity to hot themes.

5. The method according to claim 1, characterized in that, Based on the structured information tags, the customer feature profile, and the position exposure profile, the customer information matching parameters between the target customer and the original information text are calculated as follows: Calculate the holding correlation parameters based on the degree of correlation between the information structure tags and the holding exposure profile; The topic interest parameters are calculated based on the degree of fit between the information structure tags and the customer feature profile; Calculate the risk adaptation parameters based on the degree of matching between the customer risk level in the customer feature profile and the risk warning tags in the information structure tags; Calculate timing parameters based on the degree of match between the current contact timing and the customer service rhythm; Calculate the duplicate sending penalty parameters based on historical duplicate sending data; The customer information matching parameters are obtained by weighting and summing the position association parameters, the topic interest parameters, the risk adaptation parameters, and the timing parameters, and then subtracting the product of the duplication penalty parameter and the preset weight.

6. The method according to claim 1, characterized in that, Before calling the constraint template to generate personalized briefing candidate text, the following steps are also included: At least one correlation point is determined based on the customer information matching parameters; the correlation point is used to characterize the correspondence between the information structured tags and the holding exposure profile; Determine expression style constraints based on employee expression profiles; Based on the aforementioned points of association and the aforementioned style of expression constraints, the generated constraint template is constructed; the generated constraint template includes at least one of the following constraints: points of association that must be mentioned, prohibited expressions, suggested length range, opening and closing styles, risk warning expression format, and copywriting detail level.

7. The method according to claim 1, characterized in that, The generated personalized briefing candidate texts include those generated by calling the constraint template: Generate a reason for sending based on the opening description in the generated constraint template; Generate related explanatory sentences based on the reference points in the generated constraint template; the related explanatory sentences are used to explain the relationship between the original information text and the position exposure profile; Generate a service closing sentence based on the closing remarks in the generated constraint template; The personalized briefing candidate text is obtained by combining the reason for sending, the related explanation, and the service closing sentence.

8. A customer information matching system, characterized in that, include: The data acquisition module is used to acquire raw information text and customer data and holding details of target customers; The information structuring module is used to perform structuring processing on the original information text and extract information structuring tags; The customer data integration module is used to generate customer feature profiles based on the customer data. The position exposure calculation module is used to calculate the position exposure profile of the target customer based on the position details; The customer information matching parameter module is used to calculate the customer information matching parameters between the target customer and the original information text based on the information structured tags, the customer feature profile, and the holding exposure profile. The personalized briefing generation module is used to generate personalized briefing candidate text by calling the generation constraint template when the customer information matching parameters exceed a set threshold; the personalized briefing candidate text contains information matching the target customer.

9. An electronic device, characterized in that, include: processor; Memory, used to store computer programs; When the computer program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images