Product output method and device based on historical data, equipment and medium

Abstract

The application relates to the technical field of intelligent decision-making, and discloses a product output method and device based on historical data, equipment and a medium, the method comprising the following steps: constructing a corresponding customer portrait according to pre-acquired customer information of a target customer, matching the customer portrait with a preset customer historical database to obtain a matching customer case; generating a target script according to the matching customer case, and generating customer emotional data according to the customer portrait; constructing a three-dimensional scene image according to the target script and the customer emotional data; performing scene interaction expansion on the three-dimensional scene image according to a key node of the target script to obtain an expanded scene video; performing emotional enhancement and color adjustment on the expanded scene video to obtain a target rendering video; and performing product recommendation by using the target rendering video. The application can be applied to a business system platform such as financial technology, and can improve the accuracy and efficiency of product recommendation.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, and in particular to a product output method, apparatus, device, and medium based on historical data. Background Technology

[0002] With the digital transformation of the banking and insurance industries, the insurance sector is comprehensively promoting refined customer management. Historical data on customers' actual insurance purchases, claims, and coverage has become a core resource for agents' online and offline marketing conversion. Product marketing recommendations based on historical data rely on real customer policy data, accident records, family protection needs, and claims payment information to construct marketing materials. Compared to standardized product scripts, this approach more easily lowers the decision-making threshold for customers. By aggregating customer case data from all channels through a centralized big data platform, marketing models for case matching, case generation, and precise push notifications are trained, enabling intelligent association and distribution of customer leads with insurance cases.

[0003] However, existing product output methods, such as using text examples and graphic posters to introduce products to customers, are mostly generic templates. They cannot dynamically adjust the narrative perspective according to user profiles (such as age and family structure), nor can they capture users' emotional needs (such as fear and sense of responsibility). They are difficult to reproduce the emotional details in real scenarios, and the stories lack appeal. Furthermore, traditional video production requires manual scriptwriting, shooting, and editing, which is time-consuming and cannot be generated in batches quickly, resulting in low efficiency in product recommendations.

[0004] Therefore, improving the accuracy and efficiency of product output has become an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a product output method, apparatus, equipment, and medium based on historical data, with the main purpose of solving the problems of poor accuracy and efficiency in product output.

[0006] Firstly, to achieve the above objectives, the present invention provides a product output method based on historical data, comprising: Based on the customer information of the target customers to be acquired, a corresponding customer profile is constructed, and the customer profile is matched with a preset customer history database to generate a matching customer case. Based on the matched customer cases, a target script for the target customer is generated, and customer sentiment data is generated based on the customer profile. A 3D scene image is constructed based on the target script and the customer emotion data; Based on the key nodes of the target script, the 3D scene image is extended with scene interaction to generate an extended scene video; The extended scene video is enhanced with emotion and its color is adjusted to obtain a target rendered video, which is then used for product recommendations.

[0007] Secondly, the present invention also provides a product output device based on historical data, comprising: The historical data matching module is used to construct a corresponding customer profile based on the customer information of the target customer to be acquired, and to match the customer profile with the preset customer history database to generate a matching customer case. The customer sentiment data generation module is used to generate a target script for the target customer based on the matched customer case, and to generate customer sentiment data based on the customer profile. A 3D scene image construction module is used to construct a 3D scene image based on the target script and the customer emotion data. The scene interaction extension module is used to extend the 3D scene image based on the key nodes of the target script to generate an extended scene video. The product recommendation module is used to enhance the emotion and adjust the color of the extended scene video to obtain a target rendered video, and then use the target rendered video to recommend products.

[0008] Thirdly, the present invention also provides an electronic device, the electronic device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the product output method based on historical data described above.

[0009] Fourthly, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the product output method based on historical data described above.

[0010] This invention can be applied to business scenarios such as fintech. By constructing a customer profile of the target customer, it can be matched with a preset customer history database to obtain more matching customer cases. Based on the matching customer cases, a target script is generated, and customer emotional data is generated based on the customer profile. This enables the subsequent 3D scene images to have emotional narrative capabilities, which is conducive to improving the emotional identification of the target customer and thus realizing personalized product recommendations. Based on the key nodes of the target script, the 3D scene images are extended with scene interaction capabilities. Multiple ending scenes can be automatically redrawn according to customer selections, and interactive content with a thousand faces can be generated in batches for different customers. This saves the manpower cost of manually creating multiple versions of branch materials and significantly improves the efficiency of material production. The extended scene videos are enhanced with emotion and color adjustment to obtain target rendering videos. The target rendering videos are used for product recommendations, which improves the understanding of the case, personalizes the content, and has a high degree of user resonance, effectively improving the accuracy and efficiency of product recommendations. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram illustrating an application environment of a product output method based on historical data according to an embodiment of the present invention. Figure 2 A flowchart illustrating a product output method based on historical data provided in an embodiment of the present invention; Figure 3 for Figure 2 A flowchart illustrating the process of building a customer profile in step S1. Figure 4 for Figure 2 A flowchart illustrating the process of generating target scripts for target customers based on matched customer cases in step S2. Figure 5 for Figure 2 A flowchart illustrating the process of generating customer sentiment data in step S2. Figure 6 for Figure 2 A flowchart illustrating the process of constructing a 3D scene image based on the target script and customer emotional data in step S3. Figure 7 for Figure 2 A flowchart illustrating the process of extending the scene interaction of the 3D scene image in step S4. Figure 8 for Figure 2 A flowchart illustrating the process of enhancing the emotion and adjusting the color of the extended scene video in step S5 to obtain the target rendered video. Figure 9 A functional block diagram of a product output device based on historical data provided in an embodiment of the present invention; Figure 10A schematic diagram of an electronic device for implementing a product output method based on historical data, according to an embodiment of the present invention; Figure 11 Another structural schematic diagram of an electronic device for implementing a product output method based on historical data, provided as an embodiment of the present invention.

[0013] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0014] It should be noted that in the technical solutions disclosed in this invention, the acquisition of user information (personal image data (e.g., facial videos or pictures, facial feature videos or pictures, etc.) and personal privacy information (e.g., name, ID number, occupation, address, etc.)) is all completed with the user's knowledge and consent, and the acquisition of the relevant user information is legal and compliant.

[0015] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0016] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0017] This application provides a product output method based on historical data. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the device provided in this application: a server, a terminal, etc. In other words, the product output method based on historical data can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server 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, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0018] The product output method based on historical data of this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0019] Reference Figure 2 The diagram shown is a flowchart illustrating a product output method based on historical data according to an embodiment of the present invention. In this embodiment, the product output method based on historical data includes: S1. Construct a corresponding customer profile based on the customer information of the target customer to be acquired, and match the customer profile with the preset customer history database to generate a matching customer case.

[0020] In this embodiment of the invention, the target customer can refer to a potential or existing customer group that has insurance needs, purchasing power, and willingness, and meets the underwriting conditions of the insurance company. Customer information includes basic user information (such as the target customer's age, occupation, family structure, and income); behavioral preference information (such as the target customer's focus on insurance types such as "critical illness protection" or "children's education funds"); and may also include the target customer's historical interaction data (such as the types of insurance products they have browsed).

[0021] In detail, customer profiling is a comprehensive and clear description of target customers based on real customer information. It captures information such as user preferences, interests, and behavioral habits, making personalized recommendations more accurate. Customer profiling can include the target customer's attribute characteristics, behavioral characteristics, and preference characteristics.

[0022] Specifically, participate Figure 3 As shown, the step of constructing a corresponding customer profile based on the pre-acquired target customer's customer information includes: S31. Extract the attribute features, behavioral features, and historical interaction features from the customer information; S32. Construct feature triples based on the attribute features, the behavioral features, and the historical interaction features; S33. Generate a customer profile of the target customer based on the feature triplet.

[0023] In this embodiment of the invention, the attribute features are extracted from the customer's basic information, and include the target customer's age, gender, geographical location, family status, and occupation; the behavioral features describe the target customer's behavioral preference information, including browsing history, search records, click behavior, and purchase intention; the historical interaction features describe the target customer's preference data for insurance in historical interaction data, including insurance product type, historical preference data, etc.

[0024] In detail, a feature triple is obtained by unifying the dimensions of the attribute features, the behavioral features, and the historical interaction features to the same dimension, and then constructing a matrix using the unified attribute features, behavioral features, and historical interaction features as row features.

[0025] Furthermore, the attribute features, behavioral features, and historical interaction features in the feature triple can be used as feature tags for target customers. By describing the user profile in detail through feature tags, personalized product recommendations can be made.

[0026] In another embodiment of the present invention, the customer history database can be a standardized collection of claims cases established by insurance companies, insurance intermediaries, or corporate customer service and claims departments. It is a knowledge base that structures and archives historical claims cases according to dimensions such as type, scenario, and result. Insurance agents can recommend corresponding insurance products to target customers based on historical data, thereby increasing the target customers' willingness to purchase.

[0027] For example, each customer's historical data in the customer history database includes, but is not limited to, insurance type, claim type, customer profile of historical customers, time points, claim information, processing procedures, and processing results.

[0028] Specifically, the step of matching the customer profile with a preset customer history database to obtain matched customer cases includes: Convert the customer history data in the customer history database into data features; Calculate the feature similarity between the data features and the feature triples in the customer profile, and select historical data of the target customer based on the feature similarity; Based on the customer profile, the historical data of the target customer is adjusted to obtain a matching customer case.

[0029] In this embodiment of the invention, a preset vector encoder can be used to convert customer history data in the customer history database into feature vectors to obtain data features for each customer history data. The cosine similarity between the data features and each feature label in the feature triplet is calculated to obtain the feature similarity between the data features and the customer profile. The customer history data with the highest similarity is selected as the target customer history data.

[0030] In detail, the cosine similarity between the data features and the attribute features, behavioral features, and historical interaction features in the feature triples is calculated separately. The average value of the cosine similarities is calculated to obtain the feature similarity between the data features and the feature triples in the customer profile.

[0031] Furthermore, information adjustment involves modifying the customer information in the target customer's historical data based on the personalized characteristics of the target customer in the customer profile. For example, "if the target customer is a ride-hailing driver, then focus on accident insurance; if the target customer is a middle-aged family breadwinner, then focus on children's education funds." It can also be that the target customer faces the risk of family income interruption and education fund protection. Based on the specific target customer characteristic tags in the customer profile, the information is adjusted to obtain more targeted matching customer cases.

[0032] S2. Generate the target script for the target customer based on the matched customer case, and generate customer sentiment data based on the customer profile.

[0033] In this embodiment of the invention, the target script is a personalized narrative script with key emotional conflicts, turning points, and endings, and may include characters, scenes, etc. For example, the target script could depict a target customer experiencing a serious car accident while picking up passengers, facing enormous medical expenses during hospitalization, being unable to work leading to an interruption of family income, suffering from illness, and feeling helpless in the face of high medical costs. For details, please refer to Figure 4 As shown, the step of generating the target script for the target customer based on the matched customer case includes: S41. Extract the structured customer profile and unstructured data from the matched customer cases; S42. Generate multi-dimensional semantic information based on the structured customer profile and the unstructured data; S43. Perform multi-head self-attention calculation on the multi-dimensional semantic information to obtain comprehensive semantic features; S44. Map the comprehensive semantic features to text to generate the target script for the target customer.

[0034] In detail, structured customer profiles match customer profiles in customer cases, such as age, occupation, preferences, family structure, etc., while unstructured data matches communication dialogues, claims information, processing procedures, and results between customers and insurance companies and insurance agents in customer cases.

[0035] Furthermore, a pre-defined Large Language Model (LLM) can be used to extract semantic information from the structured customer profile and the unstructured data respectively, to obtain multi-dimensional semantic information corresponding to the matching customer case.

[0036] For example, a large language model may include an embedding layer, which converts structured customer profiles and unstructured data into semantic feature vectors of corresponding dimensions, and uses the semantic feature vectors of structured customer profiles and unstructured data as multi-dimensional semantic information.

[0037] Furthermore, multi-head self-attention computation on multi-dimensional semantic information can integrate the semantic information of structured customer profiles and unstructured data to obtain comprehensive feature information. Then, using a feedforward neural network, the comprehensive semantic features are mapped to the feature space corresponding to the industry knowledge base, such as insurance pain points, claims rules, and customer psychological models, to complete the text mapping of features and obtain the target script.

[0038] For example, a structured customer profile could be a 28-year-old office worker who is married, has a mortgage, and a 2-year-old child. Unstructured data could be someone who has been delaying buying insurance because they feel it's unnecessary at their age, are afraid of high premiums, hidden clauses, or disputes over claims, and have heard that many people who bought insurance didn't receive compensation, showing extreme distrust of insurance. Their core demand is to provide a safety net for their family, but they are afraid of making mistakes and are hesitant.

[0039] In detail, multi-head self-attention computation of multi-dimensional semantic information first involves concatenating the semantic feature vectors of structured customer profiles and unstructured data to obtain concatenated features. These concatenated features are then linearly projected to generate Q (Query), K (Key), and V (Value) vectors, which are then equally divided into multiple heads along the feature dimensions. Each self-attention head independently executes the self-attention mechanism to obtain a single-head output feature. Finally, the single-head output features are concatenated along the feature dimensions, and residual connections and layer normalization are performed on the concatenated features to obtain comprehensive semantic features.

[0040] By utilizing a large language model, structured customer profiles and unstructured data are converted into semantic features, resulting in multi-dimensional semantic information. Based on these multi-dimensional semantic features, multi-head self-attention computation and text mapping are performed to generate a target script including characters and scenarios. For example, the characters in the target script could be: Character 1: A 28-year-old working professional with a monthly mortgage payment of 4,000 yuan, young children, a reliable and family-oriented individual, extremely pragmatic, with a strong stereotype about insurance, believing that "insurance is all a scam, claims are difficult, and it's easy to pay but hard to get paid," and harboring wishful thinking that they are young and won't get hurt. Character 2: A colleague of the same age who has been in a serious car accident and experienced the accident firsthand. An insurance professional advisor is also included, being objective and rational, not pushing sales, only explaining rules and real-life cases, and responsible for eliminating information gaps. The scenario could be that the character suddenly suffers an acute serious illness, suspending work and being hospitalized for treatment.

[0041] Furthermore, customer sentiment data refers to the emotional characteristics of customers corresponding to their profiles. It describes the stable and continuous inner emotional tendencies and psychological feelings of customers. For example, Person 1: A 28-year-old working professional with a monthly mortgage payment of 4,000, an annual income of 100,000, and a 2-year-old child. Their customer sentiment data could include: wishful thinking, fear of making mistakes, lack of security, strong sense of responsibility, aversion to sales pitches, anxiety, cautious decision-making, and fear of troubling their family. Based on this sentiment data, targeted product recommendations can be made.

[0042] For details, please refer to Figure 5 As shown, generating customer sentiment data based on the customer profile includes: S51. Identify the behavioral status of the target customer based on the behavioral characteristics of the customer profile; S52. Calculate the customer sentiment tag corresponding to the customer profile based on the behavioral state; S53. Construct customer sentiment data based on the customer sentiment tags.

[0043] In this embodiment of the invention, the behavioral state is the real-time psychological state of the target customer. For example, when the behavioral characteristics of the customer profile are underestimating risk and not wanting to spend money, the behavioral state may be one of wishful thinking, and the corresponding customer emotional label is "giving up with wishful thinking". When the attribute characteristics are 28 years old, monthly salary of 8k, mortgage of 4000, child of 2 years old, and no commercial insurance, the customer emotional label may be "having needs but worried about the burden".

[0044] In detail, semantic reasoning can be performed using a pre-built large language model to obtain behavioral states and corresponding customer sentiment tags. Alternatively, feature vectors corresponding to feature triples in customer profiles can be matched with preset sentiment tags to aggregate all sentiment tags and obtain corresponding customer sentiment data.

[0045] Specifically, a large language model is used to semantically match behavioral features with preset candidate behavioral state labels. The behavioral state corresponding to the best semantically matched candidate behavioral state label is then selected as the target customer's behavioral state. The large language model includes a rule base between feature combinations and candidate states; the behavioral state of the target customer is identified through the rules in this rule base.

[0046] In this embodiment of the invention, customer emotional data can enable subsequent 3D scene images to have emotional narrative capabilities, which is conducive to enhancing the emotional identification of target customers and thus realizing personalized product recommendations.

[0047] S3. Construct a three-dimensional scene image based on the target script and the customer emotion data.

[0048] In this embodiment of the invention, the 3D scene data is a dynamic 3D scene corresponding to the generated target script and user emotional data. For example, for a target customer whose emotional data indicates a strong sense of family responsibility: a warm living room 3D walkthrough scene is generated as the visual background of the target script; for a target customer whose emotional data indicates anxiety about serious illness: a realistic single-person hospital room 3D scene is generated to visualize the disease risk storyline.

[0049] For details, please refer to Figure 6 As shown, the step of constructing a 3D scene image based on the target script and the customer emotion data includes: S61. Determine the core entities of the scene based on the target script, and generate a scene depth map based on the core entities of the scene; S62. Construct a sparse three-dimensional point cloud skeleton based on the scene depth map, and perform three-dimensional reconstruction of the neural radiation field on the sparse three-dimensional point cloud skeleton to obtain a three-dimensional scene. S63. Based on the customer emotion data, perform emotion modeling on the characters in the three-dimensional scene to generate a three-dimensional scene image.

[0050] In detail, the core entities of the scene include, but are not limited to, sofas, coffee tables, etc., as well as the characters appearing in the scene. For example, in a family living room of an 80㎡ small apartment, there is a light beige fabric three-seater sofa against the wall, a wooden coffee table, a TV background cabinet, floor-to-ceiling windows with white gauze curtains, a warm yellow floor lamp, a rug, cushions, and potted plants. There is natural light during the day. There are two characters in the living room, and their relationship is father and daughter, each located on a sofa in the living room.

[0051] Specifically, spatial object entities such as sofas, coffee tables, beds, counters, vehicles, and human entities such as character names and identities are extracted from the target script, as well as action entities such as sitting, standing, arguing, bowing, waving, etc. Then, the spatial distance, posture, and actions between characters are analyzed through semantic fusion of human entities and position and actions; and single-person behavior and multi-person interactive behavior are distinguished.

[0052] In this embodiment of the invention, a scene depth map is generated using a depth map model, and the spatial scale (such as sofa height, floor height 2.8m) is locked to avoid furniture proportion imbalance. At the same time, 6 to 12 fixed-view reference images (front, left and right sides, distant view, and top view) are generated in batches, each image carrying corresponding depth information. The scene depth maps from multiple perspectives are fused to generate a sparse 3D point cloud skeleton.

[0053] Among these methods, the scene depth map coordinates from multiple perspectives can be unified into a world frame before multi-view point cloud fusion registration, downsampling, and filtering and denoising are performed to construct a sparse 3D point cloud skeleton. This can avoid spatial scale distortion during 3D reconstruction and improve the accuracy of 3D reconstruction.

[0054] Furthermore, the sparse 3D point cloud skeleton is reconstructed using Neural Radiance Field (NeRF) to generate a complete indoor 3D skeleton. At the same time, the lighting can be changed and the wall color can be modified to obtain a high-definition static rendering image from any angle, that is, a 3D scene image.

[0055] Specifically, based on the customer emotion data, emotion modeling can be performed on the characters in the 3D scene to obtain a 3D scene image, including: Extract the point cloud of the character area in the 3D scene; The point cloud expression coefficients corresponding to the point cloud of the person's region are determined based on the customer's emotional data. The point cloud position of the character area is adjusted using the point cloud expression coefficients to obtain a three-dimensional scene image.

[0056] In this embodiment of the invention, point clouds in a 3D scene are classified to obtain a point cloud of the person's region. The point cloud of the person's region is then aligned with a preset standard human skeleton model to obtain a skeleton point cloud model. Based on customer emotional data, preset posture parameters (such as limb, torso, and hand movements) and preset facial expression coefficients (such as eyebrows, orbicularis oculi muscle, corners of the mouth, and eyelids) corresponding to the skeleton point cloud model are determined, which are the point cloud expression coefficients corresponding to the person's region point cloud.

[0057] This can be achieved by using pre-trained point cloud models based on deep learning or geometric rule segmentation methods to extract point clouds representing human figures in a 3D scene. For example, a pre-trained point cloud model aggregates local geometric features from neighboring points in 3D space, calculates multi-layer neighborhood features for each point cloud in the 3D scene, outputs the class probability of each point cloud based on these multi-layer neighborhood features, and determines the human figure region point cloud based on the class probability. Geometric rule segmentation achieves classification through clustering algorithms. For example, Euclidean clustering can be used to segment point cloud instances. A fixed spatial distance threshold is set, and continuous points within a 3D space with a distance less than the threshold are clustered into clusters representing the same object. The candidate clusters for all human figure region point clouds in the scene are then discretized, thus allocating the human figure region point clouds.

[0058] Furthermore, the point cloud positions of the character's limbs, torso, and hand movements are adjusted using posture parameters. The target point cloud positions of the character's eyebrows, orbicularis oculi muscle, corners of the mouth, and eyelids are determined using facial expression coefficients. The limbs, torso, and hand movements of the character in the point cloud are then adjusted to the corresponding target point cloud positions to obtain a 3D scene image with adjusted point cloud positions.

[0059] For example, when a customer's emotional data indicates anxiety about serious illness, the body, torso, and hand movements can be adjusted by pre-setting posture parameters to have the shoulders tucked in and chest hunched, hands crossed over the abdomen, body slightly leaning forward, and arms close to the torso. More detailed adjustments could include repeated rubbing of the fingers; the eyebrows, orbicularis oculi muscle, corners of the mouth, and eyelids could be adjusted to have the eyebrows slightly furrowed, the eyes slightly red, and the corners of the mouth pursed down.

[0060] In this embodiment of the invention, a three-dimensional scene image is constructed by using the target script and customer emotional data. This combines the narrative script and the characters' emotions, allowing the textual plot and the three-dimensional visual emotions to rise and fall synchronously. This weakens the customer's resistance to insurance and effectively improves the success rate of product recommendations.

[0061] S4. Based on the key nodes of the target script, perform scene interaction expansion on the 3D scene image to generate an expanded scene video.

[0062] In this embodiment of the invention, the key node can be a branching point of the three-dimensional scene image. At the key node, the customer's choice will significantly change the subsequent plot of the three-dimensional scene image, such as whether to purchase insurance and the amount of insurance purchased.

[0063] In detail, participate Figure 7 As shown, the step of extending the 3D scene image based on the key nodes of the target script to generate an extended scene video includes: S71. Extract scene keywords from the target script and determine key nodes based on the scene keywords; S72. Based on the key nodes, perform scene interaction expansion on the three-dimensional scene image to obtain the multi-ending scripts corresponding to the key nodes; S73. Generate multi-ending scene images based on the multi-ending script; S74. Render the multi-ending scene images into a branch image frame sequence, and encode the multi-ending scene images into an extended scene video based on the branch image frame sequence.

[0064] In this embodiment of the invention, keywords representing turning points and endings in the target script can be identified based on a semantic extraction model specifically designed for the financial and insurance fields. These keywords include: clarifying claims rules, eliminating misunderstandings, understanding risks, resolving inner conflicts, and dispelling concerns. Additionally, keywords representing plot transitions, such as when the story reaches a decision point, when a choice is made, and when considering how to protect family members, can be identified.

[0065] In detail, the target semantic boundaries of scene keywords are predefined. For example, the semantics of transition and conclusion are that the character's doubts, misunderstandings, and psychological conflicts about insurance, claims, and risks are resolved; information communication is completed, and a stage conclusion is reached in cognition. Scene keywords can be: clarifying the claims rules, eliminating misunderstandings, seeing the risks clearly, resolving inner conflicts, dispelling concerns, eliminating doubts, untying knots, clarifying risk cognition, and letting go of worries.

[0066] The semantic transition in the plot signifies that the narrative enters a new phase, where characters are about to think, weigh options, and make choices, placing them in a pre-decision state, with the conflict yet to be resolved. Scene keywords could include moments of choice, making a decision, considering how to protect family members, beginning to weigh the pros and cons, facing trade-offs, and carefully considering protection plans.

[0067] Specifically, a pre-trained semantic extraction model identifies the semantics of different words in the target script and determines scene keywords based on the semantics.

[0068] Furthermore, the large language model in the aforementioned target script generation can be used to generate multiple ending scripts corresponding to key nodes in real time. For example, when the key node is whether to purchase insurance, the multiple ending scripts can include what happens if there is no insurance, the coverage after an accident with insurance, etc.; when the key node is the amount of insurance purchased, the multiple ending scripts can include results with high insurance amounts and results with low insurance amounts. By extending the scene interaction of the 3D scene image, multiple ending scenes can be automatically redrawn according to customer selection, generating personalized interactive content for different customers in batches, saving the labor cost of manually creating multiple versions of branch materials and significantly improving material production efficiency.

[0069] In another embodiment of the present invention, a corresponding scene depth map can be generated for each ending in a multi-ending script, and 6 to 12 fixed-view reference images (front, left and right sides, distant view, and top view) can be generated in batches. Each image carries corresponding depth information, and the scene depth maps from multiple perspectives are fused to generate a sparse 3D point cloud skeleton. Further, a neural radiance field (NeRF) is used to perform 3D reconstruction of the sparse 3D point cloud skeleton to generate scene images for each ending, thus obtaining multi-ending scene images.

[0070] In detail, multiple sets of branch image frame sequences are generated for each ending's scene image rendering. Each set of branch image frame sequences corresponds to one ending script, meaning multiple ending scripts correspond to multiple sets of branch image frame sequences. An index mapping relationship is established between image frames and the starting frames of each ending's scene image. The image frames are then encoded into a basic video stream, and the multiple ending scene images and index mapping relationship are written into the video metadata. This encapsulation results in an extended scene video that supports pop-up selection and switching of branch screens during playback.

[0071] S5. Enhance the emotion and adjust the color of the extended scene video to obtain the target rendering video, and use the target rendering video for product recommendation.

[0072] In this embodiment of the invention, emotion enhancement involves adding background music and voice-over to the extended scene video, while color adjustment involves adjusting the picture parameters of the extended scene video to match the best color scheme.

[0073] For details, please refer to Figure 8 As shown, the process of enhancing the emotion and adjusting the color of the extended scene video to obtain the target rendered video includes: S81. Generate video audio and background audio based on the multi-ending script of the extended scene video; S82. Map the video audio and the background audio to the frame number of the extended scene video; S83. Encapsulate the video audio and the background audio into the extended scene video according to the frame number to generate an emotion-enhanced video; S84. Analyze the sentiment tags corresponding to the multi-ending script, and adjust the picture parameters of the sentiment-enhanced video according to the sentiment tags to obtain the target rendered video.

[0074] In this embodiment of the invention, corresponding character text is generated based on the characters in the multi-ending script. For example, the character text for character 1 is: "I'm going to work, you stay home and take care of our daughter," and the character text for character 2 is: "When will my insurance claim result be released?" Then, voice-over with emotions such as sobbing and pauses is generated based on the corresponding character's emotional tags to obtain the video audio. Using a client-specific emotional control vector as the global condition for audio generation, the voice emotion dynamically follows the quantized emotional peak curve, resulting in more accurate video audio.

[0075] In detail, the character text and the corresponding emotional tags of the characters are encoded into text features and emotional condition features corresponding to the emotional tags. After being fused according to the timbre features of each character, they are encoded into Mel spectrum, and the time-domain speech waveform is reconstructed to obtain the video speech.

[0076] Furthermore, background audio is automatically matched from a preset background audio library based on the plot rhythm of each ending script. For example, the opening is accompanied by light, everyday music; after the incident, it becomes tense and oppressive; after the compensation, it becomes a soothing and warm piano piece.

[0077] In another embodiment of the present invention, each ending script is converted into a corresponding extended scene video playback start and end timestamp, the frame number is determined according to the timestamp, the speech rate, pitch, loudness and aeroacoustic parameters of the video speech and background speech are dynamically adjusted, and matching frame numbers are generated in segments to encapsulate the video speech and background speech into the extended scene video, so as to map the video speech and background speech to the frame number of the extended scene video to obtain the emotion-enhanced video.

[0078] Preferably, the facial lip shape of each person can be updated synchronously with the video and audio of each person, making the image of the person more accurate.

[0079] In detail, the aforementioned method for calculating customer sentiment tags can be used to calculate the sentiment tags corresponding to each script in the multi-ending script. Based on the sentiment tags, a preset color psychology model is used to match the optimal color scheme, and the rendering engine dynamically adjusts the screen parameters in real time. For example, in everyday scenes, the screen parameters are adjusted to a highly saturated warm yellow tone, while in the event of an accident, the screen parameters are adjusted to a low-brightness, high-contrast cool blue.

[0080] In this embodiment of the invention, a customized insurance application portal is displayed at the emotional peak of the target customer through a targeted rendered video. That is, a corresponding targeted rendered video clip is shown to the target customer. For example, when the target customer inquires about the results of an insurance claim, the video displays the claim payment scenario and pops up a message: "Upgraded coverage can further cover rehabilitation costs and children's education funds," demonstrating the possibilities under different interactive expansion scenarios. This improves case comprehension, personalizes the content, and generates high user resonance, effectively improving the accuracy and efficiency of product recommendations.

[0081] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0082] like Figure 9 The diagram shown is a functional block diagram of a product output device based on historical data provided in an embodiment of the present invention.

[0083] In this embodiment of the disclosure, a product output device based on historical data is provided, which corresponds one-to-one with the product output method based on historical data described in the above embodiments. For example... Figure 9 As shown, this product output device 100 based on historical data can be installed in an electronic device. According to its functions, the product output device 100 based on historical data includes a historical data matching module 101, a customer sentiment data generation module 102, a 3D scene image construction module 103, a scene interaction extension module 104, and a product recommendation module 105. Detailed descriptions of each functional module are as follows: The historical data matching module 101 is used to construct a corresponding customer profile based on the customer information of the target customer to be acquired, and to match the customer profile with the preset customer history database to generate a matching customer case. The customer sentiment data generation module 102 is used to generate the target script of the target customer based on the matched customer case, and to generate customer sentiment data based on the customer profile. The 3D scene image construction module 103 is used to construct a 3D scene image based on the target script and the customer emotion data. The scene interaction extension module 104 is used to extend the scene interaction of the three-dimensional scene image according to the key nodes of the target script, and generate an extended scene video. The product recommendation module 105 is used to enhance the emotion and adjust the color of the extended scene video to obtain a target rendering video, and then use the target rendering video to recommend products.

[0084] In one embodiment, when the historical data matching module 101 constructs a corresponding customer profile based on the customer information of the pre-acquired target customer, it is used to: Extract attribute features, behavioral features, and historical interaction features from the customer information; Construct feature triples based on the attribute features, the behavioral features, and the historical interaction features; A customer profile of the target customer is generated based on the feature triples.

[0085] In one embodiment, when the customer sentiment data generation module 102 executes the target script to generate the target customer based on the matched customer case, it is used to: Extract the structured customer profile and unstructured data from the matched customer cases; Semantic information is extracted from the structured customer profile and the unstructured data respectively, and multi-dimensional semantic information is generated based on the semantic information; Multi-head self-attention computation is performed on the multi-dimensional semantic information to obtain comprehensive semantic features; The comprehensive semantic features are mapped to text to obtain the target script for the target customer.

[0086] In one embodiment, when the customer sentiment data generation module 102 generates customer sentiment data based on the customer profile, it is used to: Identify the behavioral status of the target customer based on the behavioral characteristics of the customer profile; Calculate the customer sentiment tag corresponding to the customer profile based on the behavioral state; Customer sentiment data is constructed based on the aforementioned customer sentiment tags.

[0087] In one embodiment, when the 3D scene image construction module 103 executes the construction of a 3D scene image based on the target script and the customer emotion data, it is used to: The core entities of the scene are determined based on the target script, and a scene depth map is generated based on the core entities of the scene. A sparse 3D point cloud skeleton is constructed based on the scene depth map, and the sparse 3D point cloud skeleton is used for 3D reconstruction of neural radiation field to obtain a 3D scene. Based on the customer's emotional data, the characters in the 3D scene are modeled emotionally to obtain a 3D scene image.

[0088] In one embodiment, when the scene interaction extension module 104 performs scene interaction extension on the 3D scene image according to the key nodes of the target script to generate an extended scene video, it is used to: Extract scene keywords from the target script, and determine key nodes based on the scene keywords; Based on the key nodes, the scene interaction of the 3D scene image is expanded to obtain the multi-ending scripts corresponding to the key nodes; Generate multi-ending scene images based on the multi-ending script; The multi-ending scene images are rendered into a sequence of branch image frames, and the multi-ending scene images are encoded into an extended scene video based on the branch image frame sequence.

[0089] In one embodiment, when the product recommendation module 105 performs emotion enhancement and color adjustment on the extended scene video to obtain the target rendered video, it is used to: Generate video audio and background audio based on the multi-ending script of the extended scene video; Map the video audio and the background audio to the frame numbers of the extended scene video; The video audio and the background audio are encapsulated into the extended scene video according to the frame number to generate an emotion-enhanced video. The emotional tags corresponding to the multi-ending script are analyzed, and the picture parameters of the emotional enhancement video are adjusted according to the emotional tags to obtain the target rendered video.

[0090] In this invention, the specific limitations of a product output device based on historical data can be found in the above-described limitations of the product output method based on historical data, and will not be repeated here. Each module in the aforementioned product output device based on historical data can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0091] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of the server-side product output method based on historical data.

[0092] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 11As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of the client-side product output method based on historical data.

[0093] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Based on the customer information of the target customers to be acquired, a corresponding customer profile is constructed, and the customer profile is matched with a preset customer history database to generate a matching customer case. Based on the matched customer cases, a target script for the target customer is generated, and customer sentiment data is generated based on the customer profile. A 3D scene image is constructed based on the target script and the customer emotion data; Based on the key nodes of the target script, the 3D scene image is extended with scene interaction to generate an extended scene video; The extended scene video is enhanced with emotion and its color is adjusted to obtain a target rendered video, which is then used for product recommendations.

[0094] In the several embodiments provided by this invention, it should be understood that the disclosed devices and apparatuses can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0095] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0096] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0097] In some embodiments of this example, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the above embodiments.

[0098] The readable storage medium of the present invention stores a computer program, which, when executed by a processor of an electronic device, can perform the following: Based on the customer information of the target customers to be acquired, a corresponding customer profile is constructed, and the customer profile is matched with a preset customer history database to generate a matching customer case. Based on the matched customer cases, a target script for the target customer is generated, and customer sentiment data is generated based on the customer profile. A 3D scene image is constructed based on the target script and the customer emotion data; Based on the key nodes of the target script, the 3D scene image is extended with scene interaction to generate an extended scene video; The extended scene video is enhanced with emotion and its color is adjusted to obtain a target rendered video, which is then used for product recommendations.

[0099] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0100] Computer-readable storage media may also store at least one computer-executable program / instruction, such as computer-readable instructions. Computer-readable storage media include, but are not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Computer-readable storage media may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions stored on the computer-readable storage medium, the various methods described above can be performed.

[0101] In addition, the computer device may include (but is not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (e.g., keyboard, mouse, speakers, etc.).

[0102] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product / computer program product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.

[0103] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0104] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0105] In the embodiments provided in this disclosure, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this disclosure. 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 marked in the blocks may occur in a different order than those marked in the drawings. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0106] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

[0107] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

Claims

1. A product output method based on historical data, characterized in that, The method includes: Based on the customer information of the target customers to be acquired, a corresponding customer profile is constructed, and the customer profile is matched with a preset customer history database to generate a matching customer case. Based on the matched customer cases, a target script for the target customer is generated, and customer sentiment data is generated based on the customer profile. A 3D scene image is constructed based on the target script and the customer emotion data; Based on the key nodes of the target script, the 3D scene image is extended with scene interaction to generate an extended scene video; The extended scene video is enhanced with emotion and its color is adjusted to obtain a target rendered video, which is then used for product recommendations.

2. The product output method based on historical data as described in claim 1, characterized in that, Based on the customer information of the target customers to be acquired, construct corresponding customer profiles, including: Extract attribute features, behavioral features, and historical interaction features from the customer information; Construct feature triples based on the attribute features, the behavioral features, and the historical interaction features; A customer profile of the target customer is generated based on the feature triples.

3. The product output method based on historical data as described in claim 1, characterized in that, The step of generating the target script for the target customer based on the matched customer case includes: Extract the structured customer profile and unstructured data from the matched customer cases; Semantic information is extracted from the structured customer profile and the unstructured data respectively, and multi-dimensional semantic information is generated based on the semantic information; Multi-head self-attention computation is performed on the multi-dimensional semantic information to obtain comprehensive semantic features; The comprehensive semantic features are mapped to text to generate the target script for the target customer.

4. The product output method based on historical data as described in claim 1, characterized in that, The step of generating customer sentiment data based on the customer profile includes: Identify the behavioral status of the target customer based on the behavioral characteristics of the customer profile; Calculate the customer sentiment tag corresponding to the customer profile based on the behavioral state; Customer sentiment data is constructed based on the aforementioned customer sentiment tags.

5. The product output method based on historical data as described in claim 1, characterized in that, The step of constructing a 3D scene image based on the target script and the customer sentiment data includes: The core entities of the scene are determined based on the target script, and a scene depth map is generated based on the core entities of the scene. A sparse 3D point cloud skeleton is constructed based on the scene depth map, and a neural radiation field 3D reconstruction is performed on the sparse 3D point cloud skeleton to obtain a 3D scene. Based on the customer's emotional data, the characters in the 3D scene are modeled emotionally to generate a 3D scene image.

6. The product output method based on historical data as described in claim 1, characterized in that, The step of extending the 3D scene image based on the key nodes of the target script to obtain an extended scene video includes: Extract scene keywords from the target script, and determine key nodes based on the scene keywords; Based on the key nodes, the scene interaction of the 3D scene image is expanded to obtain the multi-ending scripts corresponding to the key nodes; Generate multi-ending scene images based on the multi-ending script; The multi-ending scene images are rendered into a sequence of branch image frames, and the multi-ending scene images are encoded into an extended scene video based on the branch image frame sequence.

7. The product output method based on historical data as described in claim 1, characterized in that, The process of enhancing the emotion and adjusting the color of the extended scene video to obtain the target rendered video includes: Generate video audio and background audio based on the multi-ending script of the extended scene video; Map the video audio and the background audio to the frame numbers of the extended scene video; The video audio and the background audio are encapsulated into the extended scene video according to the frame number to generate an emotion-enhanced video. The emotional tags corresponding to the multi-ending script are analyzed, and the picture parameters of the emotional enhancement video are adjusted according to the emotional tags to obtain the target rendered video.

8. A product output device based on historical data, characterized in that, The device includes: The historical data matching module is used to construct a corresponding customer profile based on the customer information of the target customer to be acquired, and to match the customer profile with the preset customer history database to generate a matching customer case. The customer sentiment data generation module is used to generate a target script for the target customer based on the matched customer case, and to generate customer sentiment data based on the customer profile. A 3D scene image construction module is used to construct a 3D scene image based on the target script and the customer emotion data. The scene interaction extension module is used to extend the 3D scene image based on the key nodes of the target script to generate an extended scene video. The product recommendation module is used to enhance the emotion and adjust the color of the extended scene video to obtain a target rendered video, and then use the target rendered video to recommend products.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the product output method based on historical data as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the product output method based on historical data as described in any one of claims 1 to 7.

Images